diff --git "a/data/vsibench_train/qa_mobj_3dDetect_json_conversation.json" "b/data/vsibench_train/qa_mobj_3dDetect_json_conversation.json" deleted file mode 100644--- "a/data/vsibench_train/qa_mobj_3dDetect_json_conversation.json" +++ /dev/null @@ -1,20354 +0,0 @@ -[ - { - "id": 0, - "data_source": "Mono3DRefer", - "scene_name": "000006", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [548.0, 171.33, 572.4, 194.42], \"box_3d\": [1.48, 1.56, 3.62, -2.72, 0.82, 48.22], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": -1.21, \"box_2d\": [505.25, 168.37, 575.44, 209.18], \"box_3d\": [1.67, 1.64, 4.32, -2.61, 1.13, 31.73], \"ry\": -1.3}, {\"category\": \"Car\", \"angle\": 0.15, \"box_2d\": [49.7, 185.65, 227.42, 246.96], \"box_3d\": [1.5, 1.62, 3.88, -12.54, 1.64, 19.72], \"ry\": -0.42}, {\"category\": \"Car\", \"angle\": 2.05, \"box_2d\": [328.67, 170.65, 397.24, 204.16], \"box_3d\": [1.68, 1.67, 4.29, -12.66, 1.13, 38.44], \"ry\": 1.73}]\n```" - } - ] - }, - { - "id": 1, - "data_source": "Mono3DRefer", - "scene_name": "000007", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [564.62, 174.59, 616.43, 224.74], \"box_3d\": [1.61, 1.66, 3.2, -0.69, 1.69, 25.01], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [481.59, 180.09, 512.55, 202.42], \"box_3d\": [1.4, 1.51, 3.7, -7.43, 1.88, 47.55], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [542.05, 175.55, 565.27, 193.79], \"box_3d\": [1.46, 1.66, 4.05, -4.71, 1.71, 60.52], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 2, - "data_source": "Mono3DRefer", - "scene_name": "000008", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.04, \"box_2d\": [334.85, 178.94, 624.5, 372.04], \"box_3d\": [1.57, 1.5, 3.68, -1.17, 1.65, 7.86], \"ry\": 1.9}, {\"category\": \"Car\", \"angle\": -1.84, \"box_2d\": [937.29, 197.39, 1241.0, 374.0], \"box_3d\": [1.39, 1.44, 3.08, 3.81, 1.64, 6.15], \"ry\": -1.31}, {\"category\": \"Car\", \"angle\": -1.33, \"box_2d\": [597.59, 176.18, 720.9, 261.14], \"box_3d\": [1.47, 1.6, 3.66, 1.07, 1.55, 14.44], \"ry\": -1.25}, {\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [741.18, 168.83, 792.25, 208.43], \"box_3d\": [1.7, 1.63, 4.08, 7.24, 1.55, 33.2], \"ry\": 1.95}, {\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [884.52, 178.31, 956.41, 240.18], \"box_3d\": [1.59, 1.59, 2.47, 8.48, 1.75, 19.96], \"ry\": -1.25}]\n```" - } - ] - }, - { - "id": 3, - "data_source": "Mono3DRefer", - "scene_name": "000009", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.5, \"box_2d\": [601.96, 177.01, 659.15, 229.51], \"box_3d\": [1.61, 1.66, 3.2, 0.7, 1.76, 23.88], \"ry\": -1.48}]\n```" - } - ] - }, - { - "id": 4, - "data_source": "Mono3DRefer", - "scene_name": "000013", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [455.7, 183.86, 533.81, 241.91], \"box_3d\": [1.45, 1.56, 3.47, -3.16, 1.77, 20.13], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 5, - "data_source": "Mono3DRefer", - "scene_name": "000015", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.29, \"box_2d\": [0.0, 194.7, 414.71, 373.0], \"box_3d\": [1.57, 1.67, 4.14, -2.75, 1.7, 4.1], \"ry\": 1.72}]\n```" - } - ] - }, - { - "id": 6, - "data_source": "Mono3DRefer", - "scene_name": "000016", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.45, \"box_2d\": [661.23, 174.85, 694.38, 197.63], \"box_3d\": [1.32, 1.48, 4.44, 4.19, 1.46, 44.41], \"ry\": -1.35}, {\"category\": \"Car\", \"angle\": -1.4, \"box_2d\": [684.49, 172.19, 733.85, 205.26], \"box_3d\": [1.42, 1.56, 4.16, 4.58, 1.41, 33.13], \"ry\": -1.27}, {\"category\": \"Car\", \"angle\": -0.72, \"box_2d\": [0.0, 193.56, 386.2, 374.0], \"box_3d\": [1.45, 1.59, 3.44, -3.67, 1.67, 6.05], \"ry\": -1.24}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [494.45, 181.84, 527.48, 205.27], \"box_3d\": [1.44, 1.35, 3.23, -6.41, 2.04, 46.74], \"ry\": 1.68}]\n```" - } - ] - }, - { - "id": 7, - "data_source": "Mono3DRefer", - "scene_name": "000017", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.23, \"box_2d\": [728.25, 173.97, 791.57, 200.2], \"box_3d\": [1.62, 1.72, 4.35, 9.71, 1.73, 47.16], \"ry\": -2.03}]\n```" - } - ] - }, - { - "id": 8, - "data_source": "Mono3DRefer", - "scene_name": "000018", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [523.41, 194.53, 550.65, 219.39], \"box_3d\": [1.44, 1.66, 4.53, -4.82, 2.9, 47.02], \"ry\": -1.69}]\n```" - } - ] - }, - { - "id": 9, - "data_source": "Mono3DRefer", - "scene_name": "000019", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [742.41, 184.49, 944.56, 321.39], \"box_3d\": [1.46, 1.6, 3.71, 2.84, 1.63, 9.72], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.48, \"box_2d\": [551.01, 184.06, 575.42, 204.29], \"box_3d\": [1.58, 1.65, 3.91, -3.85, 2.52, 59.59], \"ry\": -1.55}]\n```" - } - ] - }, - { - "id": 10, - "data_source": "Mono3DRefer", - "scene_name": "000020", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [685.05, 181.43, 804.68, 258.21], \"box_3d\": [1.4, 1.61, 4.37, 2.69, 1.6, 15.58], \"ry\": -1.61}]\n```" - } - ] - }, - { - "id": 11, - "data_source": "Mono3DRefer", - "scene_name": "000021", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [359.43, 179.3, 516.3, 270.97], \"box_3d\": [1.44, 1.64, 3.78, -3.03, 1.57, 13.3], \"ry\": 1.68}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [444.13, 163.92, 559.2, 247.04], \"box_3d\": [1.77, 1.71, 4.28, -2.52, 1.61, 17.6], \"ry\": 1.68}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [724.49, 171.09, 790.5, 215.13], \"box_3d\": [1.45, 1.71, 3.93, 5.18, 1.42, 25.97], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [534.25, 166.2, 596.08, 216.76], \"box_3d\": [1.73, 1.65, 3.96, -1.62, 1.53, 26.83], \"ry\": 1.67}, {\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [709.81, 162.43, 765.7, 205.68], \"box_3d\": [1.78, 1.89, 4.37, 5.57, 1.38, 32.13], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [562.02, 166.42, 605.57, 206.89], \"box_3d\": [1.68, 1.51, 3.29, -1.14, 1.43, 31.78], \"ry\": 1.66}]\n```" - } - ] - }, - { - "id": 12, - "data_source": "Mono3DRefer", - "scene_name": "000022", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.11, \"box_2d\": [128.66, 199.48, 218.9, 237.2], \"box_3d\": [1.42, 1.7, 4.14, -19.08, 2.65, 31.72], \"ry\": -1.65}, {\"category\": \"Car\", \"angle\": -1.04, \"box_2d\": [60.72, 198.17, 159.92, 238.35], \"box_3d\": [1.52, 1.71, 4.11, -21.79, 2.69, 31.61], \"ry\": -1.64}]\n```" - } - ] - }, - { - "id": 13, - "data_source": "Mono3DRefer", - "scene_name": "000023", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [372.95, 182.64, 412.21, 205.68], \"box_3d\": [1.67, 1.87, 3.69, -16.57, 2.43, 55.08], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 14, - "data_source": "Mono3DRefer", - "scene_name": "000025", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.29, \"box_2d\": [351.84, 183.19, 537.77, 308.64], \"box_3d\": [1.47, 1.6, 3.66, -2.21, 1.63, 10.42], \"ry\": -1.49}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [562.48, 173.46, 618.49, 217.36], \"box_3d\": [1.7, 1.63, 4.08, -0.78, 1.75, 30.18], \"ry\": 1.72}, {\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [724.21, 178.91, 805.39, 249.94], \"box_3d\": [1.59, 1.59, 2.47, 3.64, 1.75, 17.48], \"ry\": -1.49}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [720.81, 187.01, 779.98, 236.22], \"box_3d\": [1.37, 1.59, 3.22, 4.23, 1.83, 22.3], \"ry\": -1.44}]\n```" - } - ] - }, - { - "id": 15, - "data_source": "Mono3DRefer", - "scene_name": "000029", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [652.31, 174.94, 690.16, 204.97], \"box_3d\": [1.64, 1.69, 3.3, 3.46, 1.78, 41.53], \"ry\": -1.63}]\n```" - } - ] - }, - { - "id": 16, - "data_source": "Mono3DRefer", - "scene_name": "000033", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.17, \"box_2d\": [0.0, 189.66, 234.52, 322.3], \"box_3d\": [1.48, 1.64, 3.59, -7.09, 1.75, 10.15], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [582.63, 175.1, 601.14, 189.77], \"box_3d\": [1.26, 1.6, 3.56, -1.64, 1.48, 64.47], \"ry\": -1.6}]\n```" - } - ] - }, - { - "id": 17, - "data_source": "Mono3DRefer", - "scene_name": "000035", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [587.01, 173.33, 614.12, 200.12], \"box_3d\": [1.65, 1.67, 3.64, -0.65, 1.71, 46.7], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.4, \"box_2d\": [425.44, 176.21, 529.23, 259.84], \"box_3d\": [1.61, 1.51, 4.5, -2.88, 1.71, 16.31], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [234.86, 185.44, 320.32, 231.58], \"box_3d\": [1.61, 1.72, 3.65, -12.69, 2.12, 27.71], \"ry\": 1.53}]\n```" - } - ] - }, - { - "id": 18, - "data_source": "Mono3DRefer", - "scene_name": "000036", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [553.16, 178.73, 693.67, 311.88], \"box_3d\": [1.55, 1.63, 3.32, 0.11, 1.64, 10.13], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [341.05, 194.66, 390.28, 218.73], \"box_3d\": [1.37, 1.82, 3.96, -15.54, 2.81, 46.01], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 2.03, \"box_2d\": [286.23, 196.48, 337.13, 220.85], \"box_3d\": [1.38, 1.79, 2.94, -18.46, 2.88, 44.69], \"ry\": 1.64}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [412.76, 190.23, 455.28, 212.54], \"box_3d\": [1.4, 1.75, 4.75, -12.25, 2.66, 50.4], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 2.67, \"box_2d\": [960.19, 183.87, 1104.42, 229.69], \"box_3d\": [1.59, 1.73, 4.26, 15.49, 2.03, 26.68], \"ry\": -3.09}, {\"category\": \"Car\", \"angle\": 2.53, \"box_2d\": [1154.43, 178.14, 1241.0, 222.29], \"box_3d\": [1.55, 1.76, 3.59, 22.52, 1.76, 26.55], \"ry\": -3.06}]\n```" - } - ] - }, - { - "id": 19, - "data_source": "Mono3DRefer", - "scene_name": "000039", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [114.51, 193.47, 466.61, 374.0], \"box_3d\": [1.48, 1.67, 4.3, -2.7, 1.73, 7.17], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [683.15, 184.41, 776.5, 256.99], \"box_3d\": [1.46, 1.54, 3.62, 2.61, 1.74, 16.67], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [533.53, 177.81, 574.19, 212.8], \"box_3d\": [1.52, 1.57, 4.42, -2.6, 1.77, 33.82], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.46, \"box_2d\": [653.23, 180.26, 702.93, 220.26], \"box_3d\": [1.51, 1.58, 3.92, 2.68, 1.82, 29.48], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.58, \"box_2d\": [561.27, 178.49, 585.9, 201.02], \"box_3d\": [1.46, 1.6, 3.33, -2.47, 1.85, 48.95], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [639.85, 175.24, 667.59, 200.41], \"box_3d\": [1.64, 1.7, 3.54, 2.93, 1.82, 49.29], \"ry\": -1.56}]\n```" - } - ] - }, - { - "id": 20, - "data_source": "Mono3DRefer", - "scene_name": "000040", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [474.58, 179.03, 516.72, 210.71], \"box_3d\": [1.45, 1.62, 3.83, -5.58, 1.77, 35.37], \"ry\": 1.52}]\n```" - } - ] - }, - { - "id": 21, - "data_source": "Mono3DRefer", - "scene_name": "000041", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.96, \"box_2d\": [439.52, 178.17, 508.93, 205.17], \"box_3d\": [1.49, 1.64, 3.7, -7.84, 1.82, 41.75], \"ry\": 3.14}]\n```" - } - ] - }, - { - "id": 22, - "data_source": "Mono3DRefer", - "scene_name": "000042", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [576.34, 174.48, 604.86, 202.67], \"box_3d\": [1.65, 1.67, 3.64, -1.21, 1.77, 44.49], \"ry\": -1.59}]\n```" - } - ] - }, - { - "id": 23, - "data_source": "Mono3DRefer", - "scene_name": "000045", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.98, \"box_2d\": [269.01, 176.97, 375.72, 242.36], \"box_3d\": [1.8, 1.61, 3.83, -8.66, 1.94, 21.96], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [346.6, 186.7, 419.98, 229.04], \"box_3d\": [1.45, 1.45, 3.4, -8.46, 1.99, 27.06], \"ry\": 1.65}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [387.31, 178.52, 455.56, 221.43], \"box_3d\": [1.75, 1.71, 4.31, -8.27, 2.02, 31.88], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [455.95, 181.43, 500.99, 209.72], \"box_3d\": [1.53, 1.66, 4.01, -7.54, 2.03, 41.53], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [480.03, 181.4, 517.92, 205.67], \"box_3d\": [1.48, 1.6, 3.9, -7.16, 2.05, 46.7], \"ry\": 1.64}, {\"category\": \"Car\", \"angle\": 2.12, \"box_2d\": [75.43, 192.52, 273.65, 277.52], \"box_3d\": [1.47, 1.63, 4.11, -8.95, 1.92, 15.21], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 24, - "data_source": "Mono3DRefer", - "scene_name": "000046", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [578.47, 176.02, 625.98, 221.24], \"box_3d\": [1.54, 1.63, 3.67, -0.32, 1.68, 26.76], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [597.59, 168.75, 634.85, 207.61], \"box_3d\": [1.77, 1.69, 4.32, 0.26, 1.61, 35.33], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.89, \"box_2d\": [769.21, 151.69, 1087.61, 374.0], \"box_3d\": [1.74, 1.62, 4.13, 2.87, 1.61, 7.64], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [706.06, 174.92, 807.35, 254.28], \"box_3d\": [1.53, 1.46, 3.26, 3.03, 1.59, 15.65], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [660.18, 171.93, 703.11, 210.02], \"box_3d\": [1.6, 1.56, 3.45, 3.13, 1.59, 32.3], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [646.4, 174.21, 684.26, 204.82], \"box_3d\": [1.52, 1.53, 4.07, 2.85, 1.62, 38.2], \"ry\": -1.61}]\n```" - } - ] - }, - { - "id": 25, - "data_source": "Mono3DRefer", - "scene_name": "000047", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.86, 718.86), (cx, cy) = (607.19, 185.22). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [572.31, 171.6, 624.11, 218.82], \"box_3d\": [1.6, 1.76, 3.84, -0.38, 1.13, 26.54], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -0.92, \"box_2d\": [0.0, 190.67, 171.41, 313.27], \"box_3d\": [1.47, 1.54, 3.51, -8.37, 1.59, 10.68], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.22, \"box_2d\": [214.99, 165.2, 423.55, 292.98], \"box_3d\": [1.78, 1.81, 4.32, -4.65, 1.49, 12.25], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.24, \"box_2d\": [296.94, 182.19, 393.12, 231.61], \"box_3d\": [1.45, 1.61, 4.3, -8.54, 1.37, 23.68], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.38, \"box_2d\": [428.19, 180.81, 498.01, 226.24], \"box_3d\": [1.45, 1.67, 3.88, -5.0, 1.3, 25.13], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.49, \"box_2d\": [522.27, 176.55, 552.71, 202.38], \"box_3d\": [1.63, 1.67, 3.92, -4.64, 1.07, 47.7], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -0.96, \"box_2d\": [0.0, 185.88, 180.39, 253.13], \"box_3d\": [1.42, 1.57, 4.23, -12.64, 1.46, 17.75], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.12, \"box_2d\": [193.59, 181.29, 294.71, 229.39], \"box_3d\": [1.58, 1.64, 3.96, -12.99, 1.45, 25.9], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 26, - "data_source": "Mono3DRefer", - "scene_name": "000049", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [287.85, 175.62, 379.41, 228.84], \"box_3d\": [1.45, 1.47, 3.52, -8.24, 1.56, 21.69], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 2.23, \"box_2d\": [0.0, 170.48, 111.64, 240.66], \"box_3d\": [1.53, 1.53, 3.45, -14.04, 1.5, 17.43], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [595.5, 177.15, 657.31, 227.34], \"box_3d\": [1.36, 1.69, 3.38, 0.45, 1.51, 21.55], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -2.24, \"box_2d\": [305.23, 179.59, 494.97, 263.68], \"box_3d\": [1.4, 1.6, 3.55, -4.1, 1.55, 13.75], \"ry\": -2.52}, {\"category\": \"Car\", \"angle\": -2.14, \"box_2d\": [410.09, 176.18, 514.77, 228.78], \"box_3d\": [1.46, 1.49, 3.4, -4.55, 1.58, 21.82], \"ry\": -2.34}, {\"category\": \"Car\", \"angle\": -2.25, \"box_2d\": [435.96, 173.6, 537.84, 218.33], \"box_3d\": [1.43, 1.62, 3.47, -4.31, 1.48, 24.83], \"ry\": -2.43}, {\"category\": \"Car\", \"angle\": -2.24, \"box_2d\": [473.64, 171.53, 556.16, 203.86], \"box_3d\": [1.37, 1.55, 3.96, -4.36, 1.34, 32.49], \"ry\": -2.37}, {\"category\": \"Car\", \"angle\": -1.98, \"box_2d\": [823.97, 189.33, 1083.14, 327.79], \"box_3d\": [1.41, 1.61, 3.71, 4.16, 1.66, 9.53], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.83, \"box_2d\": [753.34, 185.35, 885.42, 263.59], \"box_3d\": [1.36, 1.59, 3.85, 4.1, 1.64, 14.89], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [719.83, 169.31, 816.44, 238.11], \"box_3d\": [1.72, 1.79, 3.94, 4.26, 1.66, 20.2], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -0.38, \"box_2d\": [776.77, 170.97, 882.28, 211.7], \"box_3d\": [1.7, 1.77, 4.06, 9.6, 1.65, 31.83], \"ry\": -0.09}, {\"category\": \"Car\", \"angle\": 2.26, \"box_2d\": [0.0, 178.64, 252.92, 280.6], \"box_3d\": [1.41, 1.58, 3.9, -7.94, 1.52, 12.04], \"ry\": 1.69}, {\"category\": \"Car\", \"angle\": 2.19, \"box_2d\": [194.14, 174.89, 341.99, 241.74], \"box_3d\": [1.44, 1.58, 3.25, -8.06, 1.5, 17.23], \"ry\": 1.75}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [603.49, 171.71, 639.3, 204.36], \"box_3d\": [1.52, 1.67, 3.61, 0.55, 1.49, 35.67], \"ry\": -1.54}]\n```" - } - ] - }, - { - "id": 27, - "data_source": "Mono3DRefer", - "scene_name": "000051", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [713.9, 174.69, 744.43, 204.49], \"box_3d\": [1.55, 1.57, 3.64, 6.5, 1.68, 39.76], \"ry\": -1.43}]\n```" - } - ] - }, - { - "id": 28, - "data_source": "Mono3DRefer", - "scene_name": "000055", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.24, \"box_2d\": [31.61, 193.32, 129.71, 230.77], \"box_3d\": [1.48, 1.35, 3.93, -23.47, 2.44, 32.12], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 2.14, \"box_2d\": [104.42, 188.01, 188.34, 220.56], \"box_3d\": [1.5, 1.62, 4.08, -23.45, 2.3, 36.62], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 2.08, \"box_2d\": [183.23, 187.33, 245.86, 217.54], \"box_3d\": [1.65, 1.57, 3.82, -23.27, 2.53, 42.56], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 2.04, \"box_2d\": [227.36, 189.17, 281.98, 211.96], \"box_3d\": [1.37, 1.36, 4.44, -23.4, 2.48, 47.64], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 29, - "data_source": "Mono3DRefer", - "scene_name": "000057", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [514.51, 193.66, 563.52, 231.99], \"box_3d\": [1.42, 1.66, 4.11, -2.97, 2.35, 30.55], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 30, - "data_source": "Mono3DRefer", - "scene_name": "000059", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [558.06, 180.83, 599.17, 213.45], \"box_3d\": [1.26, 1.6, 3.56, -1.33, 1.6, 30.01], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.48, \"box_2d\": [509.12, 180.43, 543.44, 205.58], \"box_3d\": [1.37, 1.63, 3.57, -4.82, 1.82, 41.7], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 2.4, \"box_2d\": [969.34, 195.19, 1100.37, 246.63], \"box_3d\": [1.38, 1.35, 3.3, 12.55, 2.08, 21.41], \"ry\": 2.92}, {\"category\": \"Car\", \"angle\": 2.08, \"box_2d\": [1129.34, 194.18, 1241.0, 249.48], \"box_3d\": [1.59, 1.73, 4.02, 19.12, 2.33, 23.51], \"ry\": 2.75}]\n```" - } - ] - }, - { - "id": 31, - "data_source": "Mono3DRefer", - "scene_name": "000060", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [550.84, 173.54, 614.08, 236.57], \"box_3d\": [1.62, 1.63, 4.5, -0.78, 1.67, 21.02], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [305.57, 177.91, 341.6, 196.33], \"box_3d\": [1.46, 1.31, 4.15, -23.72, 1.89, 59.83], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 32, - "data_source": "Mono3DRefer", - "scene_name": "000061", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.58, \"box_2d\": [69.7, 217.73, 112.22, 254.15], \"box_3d\": [1.39, 1.52, 3.56, -23.89, 3.55, 33.21], \"ry\": 0.96}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [0.0, 218.4, 93.89, 297.9], \"box_3d\": [1.46, 1.59, 3.35, -13.06, 2.61, 16.63], \"ry\": 1.01}]\n```" - } - ] - }, - { - "id": 33, - "data_source": "Mono3DRefer", - "scene_name": "000064", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.83, \"box_2d\": [715.18, 175.63, 762.77, 203.9], \"box_3d\": [1.44, 1.56, 3.96, 6.91, 1.61, 39.17], \"ry\": -1.65}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [657.65, 179.93, 709.86, 219.92], \"box_3d\": [1.38, 1.49, 3.32, 2.65, 1.65, 26.82], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.85, \"box_2d\": [731.51, 180.39, 882.28, 275.8], \"box_3d\": [1.5, 1.62, 3.89, 3.43, 1.65, 13.46], \"ry\": -1.6}]\n```" - } - ] - }, - { - "id": 34, - "data_source": "Mono3DRefer", - "scene_name": "000066", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.53, \"box_2d\": [671.48, 168.74, 710.12, 206.53], \"box_3d\": [1.57, 1.56, 2.62, 3.54, 1.42, 31.58], \"ry\": -1.42}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [591.73, 172.89, 633.98, 199.59], \"box_3d\": [1.43, 1.68, 4.19, 0.19, 1.46, 40.99], \"ry\": 1.75}]\n```" - } - ] - }, - { - "id": 35, - "data_source": "Mono3DRefer", - "scene_name": "000068", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [69.26, 200.28, 435.08, 374.0], \"box_3d\": [1.46, 1.62, 4.04, -3.0, 1.79, 6.98], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 2.26, \"box_2d\": [0.0, 209.2, 198.59, 374.0], \"box_3d\": [1.46, 1.58, 3.72, -5.44, 1.85, 6.22], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [478.18, 187.68, 549.54, 249.43], \"box_3d\": [1.57, 1.6, 3.99, -2.73, 2.03, 20.96], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [530.03, 187.79, 573.1, 226.09], \"box_3d\": [1.52, 1.54, 3.68, -2.53, 2.2, 31.5], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -0.16, \"box_2d\": [667.74, 182.35, 790.82, 230.38], \"box_3d\": [1.62, 1.65, 4.14, 4.19, 1.99, 25.95], \"ry\": 0.0}, {\"category\": \"Car\", \"angle\": -0.11, \"box_2d\": [657.37, 184.48, 763.34, 221.64], \"box_3d\": [1.55, 1.66, 4.47, 4.35, 2.1, 32.0], \"ry\": 0.02}, {\"category\": \"Car\", \"angle\": -0.01, \"box_2d\": [637.45, 180.34, 714.44, 212.34], \"box_3d\": [1.69, 1.76, 4.12, 3.59, 2.12, 39.79], \"ry\": 0.08}]\n```" - } - ] - }, - { - "id": 36, - "data_source": "Mono3DRefer", - "scene_name": "000069", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.73, \"box_2d\": [130.0, 186.92, 355.47, 257.87], \"box_3d\": [1.47, 1.74, 4.26, -8.31, 1.82, 16.53], \"ry\": 3.1}, {\"category\": \"Car\", \"angle\": 0.21, \"box_2d\": [501.21, 177.91, 556.47, 194.32], \"box_3d\": [1.36, 1.68, 4.49, -7.02, 1.82, 62.51], \"ry\": 0.1}]\n```" - } - ] - }, - { - "id": 37, - "data_source": "Mono3DRefer", - "scene_name": "000071", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [596.84, 174.1, 664.34, 239.31], \"box_3d\": [1.61, 1.66, 3.2, 0.51, 1.67, 19.58], \"ry\": -1.55}]\n```" - } - ] - }, - { - "id": 38, - "data_source": "Mono3DRefer", - "scene_name": "000078", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [615.97, 168.04, 651.65, 189.21], \"box_3d\": [1.43, 1.68, 4.19, 1.72, 1.13, 51.25], \"ry\": 1.82}, {\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [222.78, 166.78, 454.29, 308.44], \"box_3d\": [1.7, 1.82, 4.45, -3.82, 1.66, 10.96], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.47, \"box_2d\": [658.66, 166.95, 697.19, 201.46], \"box_3d\": [1.57, 1.56, 2.62, 3.24, 1.32, 34.46], \"ry\": -1.38}]\n```" - } - ] - }, - { - "id": 39, - "data_source": "Mono3DRefer", - "scene_name": "000079", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.45, \"box_2d\": [669.49, 174.52, 700.47, 196.1], \"box_3d\": [1.32, 1.48, 4.44, 4.88, 1.45, 46.78], \"ry\": -1.35}, {\"category\": \"Car\", \"angle\": -1.44, \"box_2d\": [701.01, 171.82, 742.42, 201.29], \"box_3d\": [1.42, 1.56, 4.16, 5.75, 1.39, 36.99], \"ry\": -1.29}, {\"category\": \"Car\", \"angle\": -1.08, \"box_2d\": [357.22, 182.05, 567.87, 301.78], \"box_3d\": [1.45, 1.59, 3.44, -2.04, 1.61, 10.74], \"ry\": -1.26}, {\"category\": \"Car\", \"angle\": -0.7, \"box_2d\": [0.0, 183.03, 326.6, 330.77], \"box_3d\": [1.51, 1.61, 4.31, -6.07, 1.67, 9.82], \"ry\": -1.24}, {\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [483.32, 182.25, 518.76, 207.09], \"box_3d\": [1.44, 1.35, 3.23, -6.67, 2.03, 44.23], \"ry\": 1.68}]\n```" - } - ] - }, - { - "id": 40, - "data_source": "Mono3DRefer", - "scene_name": "000080", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [457.71, 176.44, 528.64, 229.39], \"box_3d\": [1.58, 1.6, 4.11, -3.79, 1.72, 23.79], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [536.11, 176.89, 565.32, 202.72], \"box_3d\": [1.52, 1.55, 3.79, -3.68, 1.79, 44.92], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [589.68, 172.46, 633.43, 216.04], \"box_3d\": [1.62, 1.63, 4.5, 0.02, 1.64, 29.39], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 41, - "data_source": "Mono3DRefer", - "scene_name": "000083", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.92, \"box_2d\": [0.0, 202.73, 361.84, 373.0], \"box_3d\": [1.47, 1.59, 4.03, -2.57, 1.64, 3.15], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.35, \"box_2d\": [368.92, 183.19, 504.02, 279.22], \"box_3d\": [1.41, 1.54, 3.36, -2.74, 1.45, 12.45], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [587.82, 169.53, 643.17, 220.02], \"box_3d\": [1.52, 1.67, 4.38, 0.43, 1.16, 24.12], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.47, \"box_2d\": [519.33, 170.45, 571.86, 215.65], \"box_3d\": [1.56, 1.59, 3.65, -2.04, 1.18, 26.91], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [463.38, 174.76, 511.89, 205.66], \"box_3d\": [1.42, 1.56, 4.16, -5.56, 1.11, 35.5], \"ry\": 1.61}]\n```" - } - ] - }, - { - "id": 42, - "data_source": "Mono3DRefer", - "scene_name": "000084", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.55, \"box_2d\": [459.33, 179.71, 505.68, 227.3], \"box_3d\": [1.68, 1.63, 3.9, -4.99, 1.96, 27.84], \"ry\": 1.37}, {\"category\": \"Car\", \"angle\": 1.46, \"box_2d\": [448.2, 181.65, 486.43, 208.6], \"box_3d\": [1.47, 1.7, 4.39, -8.4, 1.99, 42.1], \"ry\": 1.26}]\n```" - } - ] - }, - { - "id": 43, - "data_source": "Mono3DRefer", - "scene_name": "000085", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.42, \"box_2d\": [0.0, 179.86, 156.79, 244.69], \"box_3d\": [1.44, 1.62, 3.91, -13.19, 1.62, 17.05], \"ry\": -3.07}]\n```" - } - ] - }, - { - "id": 44, - "data_source": "Mono3DRefer", - "scene_name": "000087", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.2, \"box_2d\": [23.14, 183.9, 225.74, 258.36], \"box_3d\": [1.45, 1.59, 4.28, -10.88, 1.72, 16.52], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [694.9, 174.26, 735.64, 207.53], \"box_3d\": [1.43, 1.59, 4.21, 4.79, 1.52, 33.48], \"ry\": -1.49}, {\"category\": \"Car\", \"angle\": 3.11, \"box_2d\": [797.32, 169.25, 844.72, 185.72], \"box_3d\": [1.41, 1.59, 4.01, 18.87, 1.13, 64.7], \"ry\": -2.89}, {\"category\": \"Car\", \"angle\": 3.04, \"box_2d\": [886.07, 166.52, 939.02, 185.61], \"box_3d\": [1.59, 1.64, 4.08, 26.27, 1.09, 62.79], \"ry\": -2.85}]\n```" - } - ] - }, - { - "id": 45, - "data_source": "Mono3DRefer", - "scene_name": "000088", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.49, \"box_2d\": [175.86, 191.58, 373.02, 247.37], \"box_3d\": [1.36, 1.68, 4.49, -8.88, 1.89, 19.19], \"ry\": 0.06}]\n```" - } - ] - }, - { - "id": 46, - "data_source": "Mono3DRefer", - "scene_name": "000089", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.33, \"box_2d\": [317.74, 196.5, 504.71, 314.03], \"box_3d\": [1.33, 1.71, 4.09, -2.79, 1.75, 10.91], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [657.48, 176.96, 720.62, 224.75], \"box_3d\": [1.49, 1.56, 4.34, 2.6, 1.65, 24.94], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.41, \"box_2d\": [458.28, 184.16, 538.75, 243.43], \"box_3d\": [1.42, 1.53, 4.12, -3.0, 1.75, 19.85], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [704.25, 177.37, 853.41, 287.79], \"box_3d\": [1.59, 1.59, 3.89, 2.68, 1.69, 12.49], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [635.02, 178.02, 665.21, 203.12], \"box_3d\": [1.38, 1.64, 3.51, 2.33, 1.69, 41.73], \"ry\": -1.49}]\n```" - } - ] - }, - { - "id": 47, - "data_source": "Mono3DRefer", - "scene_name": "000090", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.82, \"box_2d\": [5.08, 199.56, 126.68, 269.46], \"box_3d\": [1.57, 1.72, 3.52, -14.21, 2.3, 18.74], \"ry\": -2.46}]\n```" - } - ] - }, - { - "id": 48, - "data_source": "Mono3DRefer", - "scene_name": "000091", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [709.26, 175.05, 738.72, 203.83], \"box_3d\": [1.55, 1.57, 3.64, 6.46, 1.7, 41.1], \"ry\": -1.42}]\n```" - } - ] - }, - { - "id": 49, - "data_source": "Mono3DRefer", - "scene_name": "000093", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.1, \"box_2d\": [0.0, 187.51, 326.99, 374.0], \"box_3d\": [1.61, 1.74, 4.06, -4.56, 1.79, 7.29], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [722.86, 168.15, 812.89, 237.29], \"box_3d\": [1.77, 1.77, 5.1, 4.44, 1.69, 21.21], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [246.99, 194.92, 425.43, 297.77], \"box_3d\": [1.39, 1.55, 3.65, -4.46, 1.81, 12.2], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [359.29, 185.63, 479.38, 258.98], \"box_3d\": [1.46, 1.63, 3.92, -4.32, 1.79, 16.8], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [550.61, 176.8, 587.13, 202.37], \"box_3d\": [1.47, 1.66, 3.76, -2.47, 1.72, 43.68], \"ry\": 1.66}]\n```" - } - ] - }, - { - "id": 50, - "data_source": "Mono3DRefer", - "scene_name": "000095", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [599.12, 182.07, 667.54, 247.13], \"box_3d\": [1.56, 1.65, 3.69, 0.52, 1.82, 19.42], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.11, \"box_2d\": [774.94, 172.0, 963.28, 270.83], \"box_3d\": [1.54, 1.72, 3.88, 4.81, 1.54, 13.24], \"ry\": -0.77}, {\"category\": \"Car\", \"angle\": -0.98, \"box_2d\": [705.57, 179.11, 802.71, 228.95], \"box_3d\": [1.52, 1.49, 3.46, 4.83, 1.75, 24.11], \"ry\": -0.79}, {\"category\": \"Car\", \"angle\": -0.91, \"box_2d\": [673.38, 181.32, 757.07, 216.77], \"box_3d\": [1.39, 1.61, 3.65, 4.49, 1.77, 30.76], \"ry\": -0.76}, {\"category\": \"Car\", \"angle\": 2.03, \"box_2d\": [146.16, 192.89, 340.09, 290.07], \"box_3d\": [1.52, 1.65, 4.1, -6.9, 1.96, 14.01], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [452.35, 183.18, 509.3, 222.21], \"box_3d\": [1.6, 1.6, 3.77, -5.69, 2.07, 31.97], \"ry\": 1.64}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [527.28, 181.12, 563.95, 206.6], \"box_3d\": [1.61, 1.7, 3.97, -4.28, 2.18, 48.2], \"ry\": 1.67}, {\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [241.74, 191.35, 355.4, 242.41], \"box_3d\": [1.45, 1.65, 4.63, -10.25, 2.12, 24.07], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [327.5, 181.26, 399.59, 228.12], \"box_3d\": [1.81, 1.59, 3.93, -10.26, 2.18, 30.22], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [410.89, 183.43, 465.31, 213.35], \"box_3d\": [1.56, 1.62, 4.19, -9.63, 2.18, 40.62], \"ry\": 1.69}, {\"category\": \"Car\", \"angle\": 2.0, \"box_2d\": [486.69, 184.25, 533.31, 204.92], \"box_3d\": [1.42, 1.67, 4.53, -7.35, 2.29, 53.34], \"ry\": 1.86}, {\"category\": \"Car\", \"angle\": 2.27, \"box_2d\": [71.71, 187.44, 194.72, 239.44], \"box_3d\": [1.47, 1.48, 3.21, -14.81, 1.95, 22.53], \"ry\": 1.69}, {\"category\": \"Car\", \"angle\": 2.45, \"box_2d\": [87.38, 185.27, 204.23, 220.99], \"box_3d\": [1.47, 1.78, 4.34, -21.08, 2.07, 32.94], \"ry\": 1.88}]\n```" - } - ] - }, - { - "id": 51, - "data_source": "Mono3DRefer", - "scene_name": "000096", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.64, \"box_2d\": [783.79, 184.68, 1176.76, 339.87], \"box_3d\": [1.41, 1.62, 3.6, 3.91, 1.57, 7.94], \"ry\": -0.19}, {\"category\": \"Car\", \"angle\": -0.48, \"box_2d\": [661.39, 160.21, 1003.28, 300.49], \"box_3d\": [1.77, 1.96, 4.39, 3.11, 1.63, 10.63], \"ry\": -0.21}, {\"category\": \"Car\", \"angle\": 1.43, \"box_2d\": [505.85, 185.26, 557.04, 219.6], \"box_3d\": [1.37, 1.6, 4.5, -3.59, 1.94, 31.97], \"ry\": 1.32}, {\"category\": \"Car\", \"angle\": -0.41, \"box_2d\": [639.45, 187.58, 879.59, 275.73], \"box_3d\": [1.4, 1.51, 3.98, 2.62, 1.69, 13.09], \"ry\": -0.22}]\n```" - } - ] - }, - { - "id": 52, - "data_source": "Mono3DRefer", - "scene_name": "000097", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [536.07, 176.04, 569.72, 194.78], \"box_3d\": [1.45, 1.59, 4.28, -4.6, 1.73, 58.47], \"ry\": 1.77}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [717.03, 174.45, 738.76, 193.71], \"box_3d\": [1.43, 1.59, 4.21, 9.2, 1.58, 56.31], \"ry\": -1.4}]\n```" - } - ] - }, - { - "id": 53, - "data_source": "Mono3DRefer", - "scene_name": "000100", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.46, \"box_2d\": [19.88, 179.87, 206.64, 238.0], \"box_3d\": [1.44, 1.62, 3.91, -13.04, 1.64, 18.98], \"ry\": -3.05}]\n```" - } - ] - }, - { - "id": 54, - "data_source": "Mono3DRefer", - "scene_name": "000101", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [569.23, 177.21, 636.06, 240.52], \"box_3d\": [1.55, 1.63, 3.32, -0.22, 1.68, 19.47], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.22, \"box_2d\": [0.0, 212.95, 108.77, 265.03], \"box_3d\": [1.4, 1.85, 4.11, -19.15, 2.85, 24.13], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 2.09, \"box_2d\": [106.33, 200.09, 220.09, 253.47], \"box_3d\": [1.62, 1.85, 3.57, -15.54, 2.63, 25.25], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 55, - "data_source": "Mono3DRefer", - "scene_name": "000104", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.51, \"box_2d\": [818.86, 180.55, 882.54, 238.79], \"box_3d\": [1.65, 1.67, 3.64, 7.57, 1.91, 22.62], \"ry\": -1.19}, {\"category\": \"Car\", \"angle\": 2.13, \"box_2d\": [362.79, 181.57, 534.13, 261.63], \"box_3d\": [1.47, 1.57, 4.24, -3.36, 1.68, 15.69], \"ry\": 1.93}, {\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [610.21, 182.8, 684.9, 222.12], \"box_3d\": [1.4, 1.68, 4.45, 1.55, 1.82, 28.61], \"ry\": 1.93}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [683.05, 182.45, 729.23, 213.01], \"box_3d\": [1.45, 1.62, 3.77, 4.93, 1.96, 36.89], \"ry\": 1.9}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [713.77, 181.24, 750.23, 207.65], \"box_3d\": [1.53, 1.46, 4.18, 7.56, 2.07, 44.59], \"ry\": 1.92}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [739.72, 181.11, 769.04, 203.92], \"box_3d\": [1.59, 1.57, 3.39, 10.52, 2.2, 52.53], \"ry\": 1.92}, {\"category\": \"Car\", \"angle\": -1.53, \"box_2d\": [855.51, 185.67, 886.83, 212.32], \"box_3d\": [1.5, 1.61, 4.07, 15.93, 2.31, 43.98], \"ry\": -1.18}]\n```" - } - ] - }, - { - "id": 56, - "data_source": "Mono3DRefer", - "scene_name": "000106", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [603.07, 177.15, 630.68, 204.3], \"box_3d\": [1.65, 1.67, 3.64, 0.38, 1.94, 46.12], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 2.18, \"box_2d\": [0.0, 202.63, 204.81, 291.54], \"box_3d\": [1.33, 1.46, 3.87, -9.63, 1.97, 13.8], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 2.07, \"box_2d\": [167.11, 187.32, 280.23, 244.78], \"box_3d\": [1.63, 1.5, 3.83, -12.2, 2.11, 22.99], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [449.26, 183.72, 483.4, 204.87], \"box_3d\": [1.46, 1.58, 4.32, -10.56, 2.28, 53.19], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 57, - "data_source": "Mono3DRefer", - "scene_name": "000107", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.08, \"box_2d\": [812.76, 184.44, 1241.0, 374.0], \"box_3d\": [1.46, 1.56, 4.11, 2.54, 1.55, 4.31], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.85, \"box_2d\": [733.06, 182.54, 934.34, 317.95], \"box_3d\": [1.45, 1.49, 3.79, 2.73, 1.6, 9.79], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [683.98, 178.38, 792.24, 258.45], \"box_3d\": [1.5, 1.53, 4.13, 2.6, 1.64, 15.74], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.44, \"box_2d\": [470.09, 178.3, 539.83, 233.3], \"box_3d\": [1.58, 1.65, 3.91, -3.28, 1.77, 22.89], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [671.73, 179.32, 733.51, 228.42], \"box_3d\": [1.47, 1.57, 4.09, 2.94, 1.69, 23.86], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [650.51, 179.32, 692.59, 214.58], \"box_3d\": [1.45, 1.58, 3.53, 2.63, 1.75, 31.83], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [634.97, 177.56, 666.89, 203.78], \"box_3d\": [1.46, 1.67, 4.4, 2.36, 1.76, 42.96], \"ry\": -1.56}]\n```" - } - ] - }, - { - "id": 58, - "data_source": "Mono3DRefer", - "scene_name": "000108", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [568.94, 172.64, 674.17, 273.99], \"box_3d\": [1.66, 1.73, 3.05, 0.14, 1.68, 13.49], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 59, - "data_source": "Mono3DRefer", - "scene_name": "000110", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.5, \"box_2d\": [862.65, 129.39, 1241.0, 304.96], \"box_3d\": [1.73, 1.74, 4.71, 5.5, 1.3, 8.19], \"ry\": 3.07}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [346.64, 175.63, 449.93, 248.9], \"box_3d\": [1.58, 1.76, 4.18, -5.13, 1.67, 17.86], \"ry\": 1.46}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [420.44, 170.72, 540.83, 256.12], \"box_3d\": [1.65, 1.88, 4.45, -2.78, 1.64, 16.3], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -0.35, \"box_2d\": [815.59, 143.96, 962.82, 198.54], \"box_3d\": [1.9, 1.78, 4.72, 10.19, 0.9, 26.65], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": -2.09, \"box_2d\": [966.1, 144.74, 1039.76, 182.96], \"box_3d\": [1.8, 1.65, 3.55, 19.49, 0.49, 35.99], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -2.13, \"box_2d\": [1004.98, 144.16, 1087.13, 178.96], \"box_3d\": [1.64, 1.7, 3.91, 21.91, 0.3, 36.47], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [407.73, 178.44, 487.07, 230.28], \"box_3d\": [1.55, 1.71, 4.5, -5.35, 1.76, 24.13], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.45, \"box_2d\": [657.19, 166.33, 702.65, 198.71], \"box_3d\": [1.5, 1.71, 4.44, 3.39, 1.22, 35.96], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": -1.46, \"box_2d\": [599.3, 171.76, 631.96, 197.12], \"box_3d\": [1.58, 1.71, 3.75, 0.39, 1.54, 47.31], \"ry\": -1.45}]\n```" - } - ] - }, - { - "id": 60, - "data_source": "Mono3DRefer", - "scene_name": "000111", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [372.76, 177.51, 404.38, 199.68], \"box_3d\": [1.47, 1.29, 2.95, -15.21, 1.8, 49.6], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.43, \"box_2d\": [135.31, 178.56, 175.28, 203.7], \"box_3d\": [1.48, 1.59, 3.66, -28.28, 1.86, 44.78], \"ry\": 0.87}, {\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [591.97, 174.6, 640.35, 220.36], \"box_3d\": [1.56, 1.65, 3.69, 0.16, 1.65, 26.69], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [386.09, 185.25, 455.28, 225.54], \"box_3d\": [1.37, 1.64, 3.92, -7.11, 1.86, 27.32], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": -1.01, \"box_2d\": [738.48, 169.58, 918.26, 253.0], \"box_3d\": [1.49, 1.59, 4.09, 4.54, 1.44, 14.81], \"ry\": -0.72}, {\"category\": \"Car\", \"angle\": -0.96, \"box_2d\": [684.2, 174.95, 809.64, 227.09], \"box_3d\": [1.4, 1.63, 4.05, 4.1, 1.47, 21.44], \"ry\": -0.78}, {\"category\": \"Car\", \"angle\": -0.93, \"box_2d\": [682.09, 169.25, 770.32, 211.71], \"box_3d\": [1.61, 1.67, 3.65, 4.72, 1.49, 29.23], \"ry\": -0.78}, {\"category\": \"Car\", \"angle\": -0.87, \"box_2d\": [646.27, 173.34, 712.42, 199.45], \"box_3d\": [1.49, 1.63, 4.16, 4.17, 1.53, 43.22], \"ry\": -0.78}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [393.82, 175.82, 444.37, 206.32], \"box_3d\": [1.67, 1.68, 4.49, -11.09, 1.86, 42.11], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [425.76, 175.42, 465.53, 200.91], \"box_3d\": [1.63, 1.58, 4.2, -11.03, 1.82, 48.57], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 61, - "data_source": "Mono3DRefer", - "scene_name": "000113", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [924.07, 136.19, 1241.0, 374.0], \"box_3d\": [1.75, 1.84, 4.43, 4.23, 1.53, 7.09], \"ry\": -1.23}, {\"category\": \"Car\", \"angle\": -1.33, \"box_2d\": [318.14, 182.8, 401.33, 229.76], \"box_3d\": [1.48, 1.71, 4.65, -8.76, 1.84, 25.53], \"ry\": -1.66}, {\"category\": \"Car\", \"angle\": -1.18, \"box_2d\": [214.55, 166.45, 318.87, 230.67], \"box_3d\": [2.04, 1.66, 4.62, -11.97, 1.87, 25.43], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": -1.51, \"box_2d\": [455.9, 171.22, 499.06, 205.38], \"box_3d\": [1.42, 1.64, 3.95, -5.92, 1.38, 32.39], \"ry\": -1.69}, {\"category\": \"Car\", \"angle\": -1.85, \"box_2d\": [692.53, 171.83, 770.72, 215.35], \"box_3d\": [1.51, 1.72, 4.5, 4.5, 1.51, 27.59], \"ry\": -1.69}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [573.97, 163.73, 607.83, 193.3], \"box_3d\": [1.7, 1.8, 4.5, -1.24, 1.2, 44.05], \"ry\": -1.65}, {\"category\": \"Car\", \"angle\": -1.9, \"box_2d\": [745.82, 166.87, 827.42, 212.81], \"box_3d\": [1.64, 1.73, 4.43, 6.77, 1.46, 28.28], \"ry\": -1.66}, {\"category\": \"Car\", \"angle\": -1.02, \"box_2d\": [37.15, 187.2, 172.25, 237.32], \"box_3d\": [1.53, 1.61, 4.73, -17.54, 2.06, 25.31], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": -0.98, \"box_2d\": [0.0, 192.82, 32.87, 233.67], \"box_3d\": [1.4, 1.56, 3.48, -22.11, 2.13, 25.07], \"ry\": -1.7}]\n```" - } - ] - }, - { - "id": 62, - "data_source": "Mono3DRefer", - "scene_name": "000116", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [495.51, 181.48, 548.24, 222.91], \"box_3d\": [1.55, 1.64, 3.7, -3.54, 1.91, 29.23], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.44, \"box_2d\": [563.32, 177.53, 606.36, 208.9], \"box_3d\": [1.63, 1.81, 4.54, -1.36, 1.91, 40.16], \"ry\": -1.48}, {\"category\": \"Car\", \"angle\": 1.18, \"box_2d\": [793.39, 146.43, 1130.33, 369.18], \"box_3d\": [1.77, 1.63, 4.1, 3.31, 1.59, 7.85], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [682.29, 172.91, 753.25, 230.45], \"box_3d\": [1.69, 1.64, 4.21, 3.38, 1.73, 23.57], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 63, - "data_source": "Mono3DRefer", - "scene_name": "000119", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [381.63, 186.6, 479.85, 256.4], \"box_3d\": [1.4, 1.45, 3.3, -4.05, 1.74, 16.55], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [496.82, 183.63, 587.26, 252.05], \"box_3d\": [1.4, 1.63, 3.89, -1.53, 1.67, 17.03], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [519.34, 180.25, 550.61, 206.65], \"box_3d\": [1.38, 1.33, 4.32, -4.17, 1.8, 40.28], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 64, - "data_source": "Mono3DRefer", - "scene_name": "000121", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.08, \"box_2d\": [340.31, 183.86, 415.2, 224.13], \"box_3d\": [1.44, 1.35, 3.23, -8.97, 1.89, 27.99], \"ry\": 1.77}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [487.45, 180.67, 524.76, 205.44], \"box_3d\": [1.46, 1.37, 3.45, -6.43, 1.96, 44.77], \"ry\": 1.71}, {\"category\": \"Car\", \"angle\": -1.51, \"box_2d\": [721.67, 170.45, 752.7, 196.14], \"box_3d\": [1.45, 1.59, 3.44, 7.57, 1.34, 42.88], \"ry\": -1.33}, {\"category\": \"Car\", \"angle\": -1.36, \"box_2d\": [637.5, 172.08, 683.95, 202.02], \"box_3d\": [1.51, 1.61, 4.31, 2.76, 1.5, 38.75], \"ry\": -1.29}]\n```" - } - ] - }, - { - "id": 65, - "data_source": "Mono3DRefer", - "scene_name": "000122", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [581.54, 177.91, 650.72, 244.53], \"box_3d\": [1.66, 1.73, 3.05, 0.11, 1.82, 19.73], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 66, - "data_source": "Mono3DRefer", - "scene_name": "000123", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [602.41, 176.05, 634.63, 198.55], \"box_3d\": [1.44, 1.68, 4.39, 0.6, 1.67, 48.73], \"ry\": 1.7}]\n```" - } - ] - }, - { - "id": 67, - "data_source": "Mono3DRefer", - "scene_name": "000125", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [518.63, 165.68, 740.8, 374.0], \"box_3d\": [1.55, 1.63, 3.32, 0.1, 1.52, 7.01], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -2.4, \"box_2d\": [207.64, 191.87, 314.06, 224.57], \"box_3d\": [1.46, 1.85, 4.52, -16.75, 2.4, 34.57], \"ry\": -2.85}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [376.83, 185.26, 418.16, 205.23], \"box_3d\": [1.37, 1.82, 3.96, -15.63, 2.31, 53.18], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -2.27, \"box_2d\": [62.38, 201.43, 144.37, 232.69], \"box_3d\": [1.52, 1.8, 3.3, -26.45, 3.04, 37.65], \"ry\": -2.88}, {\"category\": \"Car\", \"angle\": -2.58, \"box_2d\": [327.61, 183.69, 399.92, 214.87], \"box_3d\": [1.6, 1.6, 3.29, -13.14, 2.2, 38.47], \"ry\": -2.91}, {\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [335.04, 187.99, 374.41, 208.15], \"box_3d\": [1.38, 1.79, 2.94, -18.58, 2.51, 52.57], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [425.88, 181.46, 464.31, 201.44], \"box_3d\": [1.4, 1.75, 4.75, -12.29, 2.06, 53.93], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 68, - "data_source": "Mono3DRefer", - "scene_name": "000128", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.17, \"box_2d\": [6.2, 184.24, 193.57, 262.33], \"box_3d\": [1.84, 2.04, 4.33, -13.63, 2.18, 19.61], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 69, - "data_source": "Mono3DRefer", - "scene_name": "000129", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.15, \"box_2d\": [53.08, 185.99, 265.42, 288.46], \"box_3d\": [1.73, 1.6, 4.11, -8.86, 2.02, 14.58], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [301.7, 185.67, 399.78, 235.31], \"box_3d\": [1.47, 1.63, 4.11, -8.54, 1.92, 24.02], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [385.9, 175.79, 451.78, 221.27], \"box_3d\": [1.8, 1.61, 3.83, -8.09, 1.95, 30.74], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [429.43, 183.6, 476.76, 214.77], \"box_3d\": [1.45, 1.45, 3.4, -7.78, 2.0, 35.85], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [451.47, 178.0, 500.72, 211.1], \"box_3d\": [1.75, 1.71, 4.31, -7.51, 2.06, 40.69], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [497.01, 180.44, 532.54, 203.57], \"box_3d\": [1.53, 1.66, 4.01, -6.6, 2.06, 50.21], \"ry\": 1.64}]\n```" - } - ] - }, - { - "id": 70, - "data_source": "Mono3DRefer", - "scene_name": "000131", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.69, \"box_2d\": [205.19, 186.62, 358.07, 240.66], \"box_3d\": [1.35, 1.6, 3.86, -9.36, 1.78, 20.79], \"ry\": -1.11}, {\"category\": \"Car\", \"angle\": -0.93, \"box_2d\": [442.96, 178.65, 530.25, 217.98], \"box_3d\": [1.49, 1.53, 3.8, -5.01, 1.74, 29.57], \"ry\": -1.09}, {\"category\": \"Car\", \"angle\": 2.43, \"box_2d\": [208.95, 180.57, 327.36, 223.32], \"box_3d\": [1.49, 1.34, 4.25, -13.04, 1.81, 27.73], \"ry\": 1.99}, {\"category\": \"Car\", \"angle\": 2.07, \"box_2d\": [459.14, 174.29, 520.51, 205.65], \"box_3d\": [1.72, 1.53, 4.55, -6.95, 1.83, 42.04], \"ry\": 1.9}, {\"category\": \"Car\", \"angle\": -1.12, \"box_2d\": [576.15, 174.56, 629.6, 202.46], \"box_3d\": [1.49, 1.56, 3.65, -0.38, 1.6, 40.56], \"ry\": -1.13}, {\"category\": \"Car\", \"angle\": -1.3, \"box_2d\": [711.82, 173.54, 737.26, 190.13], \"box_3d\": [1.44, 1.46, 3.16, 10.3, 1.52, 64.8], \"ry\": -1.15}]\n```" - } - ] - }, - { - "id": 71, - "data_source": "Mono3DRefer", - "scene_name": "000137", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [595.95, 175.74, 627.85, 206.45], \"box_3d\": [1.66, 1.73, 3.05, 0.07, 1.85, 41.01], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [317.87, 186.6, 376.12, 218.12], \"box_3d\": [1.5, 1.8, 3.46, -13.5, 2.23, 37.14], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [426.78, 182.03, 458.71, 202.32], \"box_3d\": [1.54, 1.66, 3.72, -13.36, 2.29, 57.74], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 72, - "data_source": "Mono3DRefer", - "scene_name": "000138", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.95, \"box_2d\": [38.45, 181.56, 140.57, 224.81], \"box_3d\": [1.56, 1.72, 3.09, -20.09, 1.91, 27.95], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [382.74, 181.02, 491.72, 251.1], \"box_3d\": [1.46, 1.59, 4.12, -4.03, 1.67, 17.29], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.14, \"box_2d\": [269.6, 176.91, 317.85, 202.62], \"box_3d\": [1.57, 1.57, 3.32, -20.26, 1.85, 46.28], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": 1.97, \"box_2d\": [307.38, 175.77, 355.18, 199.29], \"box_3d\": [1.57, 1.2, 5.24, -19.72, 1.8, 51.21], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [547.16, 176.49, 571.08, 194.74], \"box_3d\": [1.41, 1.69, 4.36, -4.11, 1.73, 58.61], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [506.46, 172.23, 549.63, 209.98], \"box_3d\": [1.72, 1.66, 4.32, -3.99, 1.73, 35.41], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [586.18, 174.11, 623.76, 211.54], \"box_3d\": [1.62, 1.63, 4.5, -0.24, 1.71, 33.84], \"ry\": -1.56}]\n```" - } - ] - }, - { - "id": 73, - "data_source": "Mono3DRefer", - "scene_name": "000139", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [594.35, 177.97, 621.4, 204.02], \"box_3d\": [1.66, 1.73, 3.05, -0.16, 2.02, 48.08], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 74, - "data_source": "Mono3DRefer", - "scene_name": "000141", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [405.52, 184.59, 530.15, 263.84], \"box_3d\": [1.44, 1.68, 4.39, -2.94, 1.71, 15.64], \"ry\": 1.61}]\n```" - } - ] - }, - { - "id": 75, - "data_source": "Mono3DRefer", - "scene_name": "000142", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.07, \"box_2d\": [835.06, 166.73, 1241.0, 374.0], \"box_3d\": [1.57, 1.54, 3.22, 2.86, 1.56, 5.12], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [726.07, 154.71, 916.86, 306.71], \"box_3d\": [1.74, 1.6, 3.86, 2.7, 1.56, 10.24], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [688.52, 153.34, 806.1, 252.85], \"box_3d\": [1.83, 1.69, 4.44, 2.74, 1.5, 15.58], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [678.28, 170.36, 747.97, 231.25], \"box_3d\": [1.61, 1.62, 3.78, 2.9, 1.57, 21.17], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [530.82, 167.68, 560.49, 197.37], \"box_3d\": [1.86, 1.57, 3.83, -4.22, 1.56, 47.51], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [428.14, 178.08, 495.67, 225.49], \"box_3d\": [1.5, 1.46, 3.7, -5.05, 1.69, 24.84], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 2.21, \"box_2d\": [0.0, 197.78, 273.61, 374.0], \"box_3d\": [1.54, 1.54, 3.72, -5.07, 1.84, 7.13], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [197.05, 186.01, 398.12, 301.0], \"box_3d\": [1.57, 1.6, 4.24, -5.08, 1.82, 12.3], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [337.15, 176.21, 452.69, 252.15], \"box_3d\": [1.65, 1.61, 3.73, -5.17, 1.75, 17.71], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 76, - "data_source": "Mono3DRefer", - "scene_name": "000143", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.7, \"box_2d\": [940.46, 180.55, 1003.22, 202.15], \"box_3d\": [1.42, 1.58, 3.48, 24.71, 1.96, 49.38], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": 2.21, \"box_2d\": [4.58, 199.18, 128.63, 248.05], \"box_3d\": [1.49, 1.81, 3.36, -18.95, 2.47, 25.29], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.98, \"box_2d\": [120.97, 201.18, 178.0, 228.44], \"box_3d\": [1.41, 1.64, 3.22, -26.81, 3.12, 42.07], \"ry\": 1.42}]\n```" - } - ] - }, - { - "id": 77, - "data_source": "Mono3DRefer", - "scene_name": "000144", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.18, \"box_2d\": [0.0, 202.35, 277.49, 374.0], \"box_3d\": [1.5, 1.46, 3.7, -5.07, 1.86, 7.38], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -2.17, \"box_2d\": [919.98, 162.0, 1241.0, 374.0], \"box_3d\": [1.61, 1.62, 3.78, 2.96, 1.58, 3.44], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [472.82, 172.55, 524.05, 220.8], \"box_3d\": [1.86, 1.57, 3.83, -4.6, 1.88, 29.98], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 78, - "data_source": "Mono3DRefer", - "scene_name": "000145", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (707.05, 707.05), (cx, cy) = (604.08, 180.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.16, \"box_2d\": [960.87, 180.9, 1069.16, 229.09], \"box_3d\": [1.48, 1.56, 3.33, 13.75, 1.51, 23.96], \"ry\": -1.64}, {\"category\": \"Car\", \"angle\": -2.24, \"box_2d\": [1034.19, 183.18, 1155.16, 229.33], \"box_3d\": [1.4, 1.59, 3.35, 16.37, 1.51, 23.86], \"ry\": -1.65}, {\"category\": \"Car\", \"angle\": 0.89, \"box_2d\": [1097.59, 183.53, 1223.0, 230.26], \"box_3d\": [1.44, 1.5, 3.8, 19.16, 1.57, 24.48], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [601.61, 171.69, 637.7, 199.9], \"box_3d\": [1.39, 1.64, 4.51, 0.72, 0.95, 37.7], \"ry\": -1.61}]\n```" - } - ] - }, - { - "id": 79, - "data_source": "Mono3DRefer", - "scene_name": "000148", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.39, \"box_2d\": [782.12, 174.04, 860.47, 203.67], \"box_3d\": [1.62, 1.72, 4.35, 12.19, 1.72, 41.96], \"ry\": -2.11}]\n```" - } - ] - }, - { - "id": 80, - "data_source": "Mono3DRefer", - "scene_name": "000150", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [565.58, 175.41, 616.95, 225.13], \"box_3d\": [1.61, 1.66, 3.2, -0.66, 1.72, 25.21], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [453.46, 183.22, 494.9, 211.37], \"box_3d\": [1.4, 1.51, 3.7, -7.22, 1.97, 38.47], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [531.96, 177.58, 560.16, 199.7], \"box_3d\": [1.46, 1.66, 4.05, -4.43, 1.81, 50.24], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 81, - "data_source": "Mono3DRefer", - "scene_name": "000151", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (707.05, 707.05), (cx, cy) = (604.08, 180.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.86, \"box_2d\": [0.0, 200.41, 303.17, 369.0], \"box_3d\": [1.5, 1.78, 3.69, -3.17, 1.67, 3.35], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -0.39, \"box_2d\": [845.82, 161.32, 1004.89, 210.46], \"box_3d\": [1.45, 1.75, 4.23, 10.03, 0.89, 22.42], \"ry\": 0.03}]\n```" - } - ] - }, - { - "id": 82, - "data_source": "Mono3DRefer", - "scene_name": "000153", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (707.05, 707.05), (cx, cy) = (604.08, 180.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [669.6, 159.28, 704.34, 194.51], \"box_3d\": [1.91, 1.65, 4.4, 4.69, 0.75, 40.87], \"ry\": -1.53}]\n```" - } - ] - }, - { - "id": 83, - "data_source": "Mono3DRefer", - "scene_name": "000157", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.37, \"box_2d\": [1201.15, 177.29, 1241.0, 360.68], \"box_3d\": [1.48, 1.71, 4.65, 7.34, 1.5, 6.51], \"ry\": -0.55}, {\"category\": \"Car\", \"angle\": -1.04, \"box_2d\": [891.44, 125.59, 1231.34, 319.09], \"box_3d\": [2.04, 1.66, 4.62, 5.94, 1.54, 9.42], \"ry\": -0.49}, {\"category\": \"Car\", \"angle\": -0.75, \"box_2d\": [656.98, 173.08, 892.62, 262.76], \"box_3d\": [1.53, 1.61, 4.73, 3.32, 1.56, 14.36], \"ry\": -0.53}, {\"category\": \"Car\", \"angle\": -0.6, \"box_2d\": [581.09, 183.36, 728.9, 244.92], \"box_3d\": [1.4, 1.56, 3.48, 1.13, 1.69, 18.4], \"ry\": -0.54}, {\"category\": \"Car\", \"angle\": -1.82, \"box_2d\": [319.23, 192.29, 406.17, 238.46], \"box_3d\": [1.27, 1.66, 4.06, -8.01, 1.93, 23.03], \"ry\": -2.15}, {\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [219.4, 194.58, 279.65, 242.44], \"box_3d\": [1.48, 1.67, 4.5, -13.1, 2.31, 26.02], \"ry\": -2.09}]\n```" - } - ] - }, - { - "id": 84, - "data_source": "Mono3DRefer", - "scene_name": "000160", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [493.84, 183.02, 549.17, 226.05], \"box_3d\": [1.48, 1.62, 3.21, -3.27, 1.87, 26.83], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [558.47, 180.63, 580.7, 200.5], \"box_3d\": [1.55, 1.51, 3.75, -3.27, 2.19, 58.69], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [343.69, 179.05, 379.71, 203.45], \"box_3d\": [1.92, 1.58, 3.93, -20.4, 2.45, 59.38], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [591.6, 179.67, 633.03, 220.94], \"box_3d\": [1.62, 1.63, 4.5, 0.05, 1.93, 30.9], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 85, - "data_source": "Mono3DRefer", - "scene_name": "000162", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.93, \"box_2d\": [3.64, 173.55, 121.62, 220.64], \"box_3d\": [1.69, 1.74, 3.76, -21.18, 1.74, 28.06], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.03, \"box_2d\": [135.24, 175.95, 207.63, 209.55], \"box_3d\": [1.55, 1.39, 3.55, -21.47, 1.72, 35.41], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -0.76, \"box_2d\": [159.36, 166.92, 223.25, 195.21], \"box_3d\": [1.72, 1.76, 3.14, -26.5, 1.38, 45.71], \"ry\": -1.28}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [573.87, 171.69, 633.19, 230.82], \"box_3d\": [1.62, 1.63, 4.5, -0.26, 1.62, 22.26], \"ry\": -1.59}]\n```" - } - ] - }, - { - "id": 86, - "data_source": "Mono3DRefer", - "scene_name": "000163", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.77, \"box_2d\": [258.45, 164.94, 438.13, 248.04], \"box_3d\": [1.86, 1.69, 3.69, -6.23, 1.71, 17.33], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": -2.85, \"box_2d\": [321.99, 171.43, 472.77, 230.26], \"box_3d\": [1.68, 1.69, 4.07, -6.42, 1.67, 21.94], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [804.72, 152.49, 871.5, 197.29], \"box_3d\": [2.04, 1.99, 4.82, 11.1, 1.14, 35.55], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": -1.91, \"box_2d\": [857.19, 161.47, 917.02, 193.08], \"box_3d\": [1.49, 1.56, 3.88, 13.84, 0.98, 36.31], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -2.91, \"box_2d\": [374.76, 175.04, 495.42, 216.99], \"box_3d\": [1.48, 1.56, 4.12, 0.23, 1.59, 26.95], \"ry\": 3.14}, {\"category\": \"Car\", \"angle\": -2.95, \"box_2d\": [398.89, 175.33, 492.56, 214.08], \"box_3d\": [1.51, 1.59, 3.5, 0.0, 1.64, 29.62], \"ry\": 3.11}, {\"category\": \"Car\", \"angle\": -2.93, \"box_2d\": [404.63, 167.36, 515.88, 208.73], \"box_3d\": [1.74, 1.75, 4.57, -6.63, 1.54, 32.08], \"ry\": -3.13}]\n```" - } - ] - }, - { - "id": 87, - "data_source": "Mono3DRefer", - "scene_name": "000165", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.07, \"box_2d\": [356.62, 163.82, 486.03, 243.24], \"box_3d\": [1.83, 1.64, 3.71, -4.97, 1.65, 18.58], \"ry\": -2.32}, {\"category\": \"Car\", \"angle\": -2.27, \"box_2d\": [412.04, 177.09, 511.76, 221.61], \"box_3d\": [1.5, 1.58, 3.57, -5.42, 1.67, 26.07], \"ry\": -2.48}, {\"category\": \"Car\", \"angle\": -2.27, \"box_2d\": [443.08, 167.94, 536.35, 215.62], \"box_3d\": [1.8, 1.63, 3.79, -4.91, 1.64, 29.02], \"ry\": -2.43}, {\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [583.11, 173.29, 617.63, 204.61], \"box_3d\": [1.52, 1.67, 3.61, -0.48, 1.57, 37.12], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [581.95, 178.78, 636.65, 223.24], \"box_3d\": [1.36, 1.69, 3.38, -0.06, 1.58, 24.12], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -2.14, \"box_2d\": [287.29, 164.14, 471.26, 259.37], \"box_3d\": [1.76, 1.74, 4.16, -5.02, 1.63, 15.29], \"ry\": -2.45}]\n```" - } - ] - }, - { - "id": 88, - "data_source": "Mono3DRefer", - "scene_name": "000166", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.3, \"box_2d\": [0.0, 175.12, 44.6, 219.59], \"box_3d\": [1.47, 1.82, 3.03, -21.13, 1.56, 24.21], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -1.17, \"box_2d\": [296.49, 172.8, 339.4, 192.06], \"box_3d\": [1.42, 1.68, 3.93, -22.58, 1.45, 55.87], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.1, \"box_2d\": [222.16, 157.13, 275.57, 189.98], \"box_3d\": [2.48, 1.81, 4.73, -28.61, 1.32, 57.29], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [570.67, 169.82, 616.94, 215.92], \"box_3d\": [1.62, 1.63, 4.5, -0.62, 1.55, 27.9], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 89, - "data_source": "Mono3DRefer", - "scene_name": "000168", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [553.89, 184.63, 598.47, 220.55], \"box_3d\": [1.26, 1.6, 3.56, -1.39, 1.73, 27.81], \"ry\": -1.64}, {\"category\": \"Car\", \"angle\": -1.49, \"box_2d\": [502.27, 183.99, 537.52, 210.78], \"box_3d\": [1.37, 1.63, 3.57, -4.95, 2.0, 39.72], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": 2.52, \"box_2d\": [837.66, 190.04, 921.14, 223.55], \"box_3d\": [1.38, 1.35, 3.3, 11.95, 2.18, 32.11], \"ry\": 2.87}, {\"category\": \"Car\", \"angle\": 2.24, \"box_2d\": [955.63, 189.82, 1046.94, 226.08], \"box_3d\": [1.59, 1.73, 4.02, 18.79, 2.44, 34.7], \"ry\": 2.73}]\n```" - } - ] - }, - { - "id": 90, - "data_source": "Mono3DRefer", - "scene_name": "000172", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [582.19, 170.67, 618.01, 204.75], \"box_3d\": [1.54, 1.63, 3.67, -0.5, 1.47, 34.89], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [596.79, 165.06, 626.58, 196.12], \"box_3d\": [1.77, 1.69, 4.32, 0.1, 1.35, 43.65], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -2.09, \"box_2d\": [822.23, 180.98, 1241.0, 374.0], \"box_3d\": [1.49, 1.55, 3.93, 2.78, 1.56, 4.83], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [727.09, 171.5, 890.13, 304.24], \"box_3d\": [1.54, 1.48, 3.17, 2.53, 1.55, 10.02], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [708.59, 168.99, 814.5, 253.27], \"box_3d\": [1.54, 1.55, 3.51, 2.99, 1.5, 15.07], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [658.22, 164.22, 711.5, 212.42], \"box_3d\": [1.74, 1.62, 4.13, 2.84, 1.46, 28.4], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 2.12, \"box_2d\": [0.0, 189.21, 251.43, 303.15], \"box_3d\": [1.51, 1.58, 4.16, -7.85, 1.82, 12.07], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [257.04, 181.46, 376.13, 243.67], \"box_3d\": [1.51, 1.65, 3.94, -7.91, 1.76, 19.75], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [637.51, 167.98, 661.53, 190.65], \"box_3d\": [1.6, 1.56, 3.45, 2.86, 1.28, 53.1], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 91, - "data_source": "Mono3DRefer", - "scene_name": "000176", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [500.99, 164.61, 585.54, 243.1], \"box_3d\": [1.91, 1.66, 4.22, -1.75, 1.73, 19.75], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [553.91, 169.52, 615.84, 225.81], \"box_3d\": [1.83, 1.69, 4.28, -0.84, 1.75, 25.71], \"ry\": 1.66}, {\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [624.61, 179.02, 669.6, 207.09], \"box_3d\": [1.51, 1.63, 4.03, 2.14, 1.87, 41.12], \"ry\": 1.86}, {\"category\": \"Car\", \"angle\": -2.15, \"box_2d\": [844.52, 209.22, 1241.0, 374.0], \"box_3d\": [1.48, 1.6, 3.79, 2.5, 1.74, 3.48], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [425.86, 182.0, 542.59, 273.89], \"box_3d\": [1.56, 1.55, 3.64, -2.38, 1.75, 14.22], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 92, - "data_source": "Mono3DRefer", - "scene_name": "000177", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.67, \"box_2d\": [347.96, 180.67, 438.56, 211.43], \"box_3d\": [1.44, 1.62, 3.91, -10.57, 1.84, 35.17], \"ry\": -2.96}]\n```" - } - ] - }, - { - "id": 93, - "data_source": "Mono3DRefer", - "scene_name": "000179", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [591.44, 175.51, 657.28, 239.12], \"box_3d\": [1.61, 1.66, 3.2, 0.38, 1.7, 20.03], \"ry\": -1.54}]\n```" - } - ] - }, - { - "id": 94, - "data_source": "Mono3DRefer", - "scene_name": "000180", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [541.04, 181.47, 566.39, 197.81], \"box_3d\": [1.37, 1.79, 4.16, -4.94, 2.14, 63.6], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 0.19, \"box_2d\": [439.69, 183.58, 507.67, 206.66], \"box_3d\": [1.56, 1.71, 4.5, -9.7, 2.35, 51.41], \"ry\": 0.01}]\n```" - } - ] - }, - { - "id": 95, - "data_source": "Mono3DRefer", - "scene_name": "000183", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.18, \"box_2d\": [884.54, 170.7, 1241.0, 374.0], \"box_3d\": [1.46, 1.67, 4.4, 2.8, 1.46, 3.31], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": -2.62, \"box_2d\": [51.28, 170.11, 365.03, 305.68], \"box_3d\": [1.8, 1.62, 3.7, -5.76, 1.79, 10.54], \"ry\": -3.11}, {\"category\": \"Car\", \"angle\": 1.45, \"box_2d\": [682.05, 167.43, 741.3, 220.29], \"box_3d\": [1.77, 1.63, 4.1, 3.61, 1.62, 26.48], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [662.5, 169.09, 696.78, 200.09], \"box_3d\": [1.69, 1.64, 4.21, 3.97, 1.52, 41.94], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [551.51, 176.39, 580.12, 200.97], \"box_3d\": [1.55, 1.64, 3.7, -2.92, 1.8, 47.83], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.45, \"box_2d\": [589.23, 172.87, 618.3, 193.97], \"box_3d\": [1.63, 1.81, 4.54, -0.48, 1.66, 58.69], \"ry\": -1.46}]\n```" - } - ] - }, - { - "id": 96, - "data_source": "Mono3DRefer", - "scene_name": "000184", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.4, \"box_2d\": [611.04, 177.87, 643.08, 201.09], \"box_3d\": [1.54, 1.63, 3.67, 1.2, 1.91, 50.21], \"ry\": -1.38}, {\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [443.04, 184.18, 500.28, 217.8], \"box_3d\": [1.56, 1.59, 3.69, -6.87, 2.14, 35.99], \"ry\": 1.74}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [474.35, 176.63, 523.09, 212.88], \"box_3d\": [1.93, 1.59, 4.01, -6.24, 2.16, 40.7], \"ry\": 1.72}, {\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [659.04, 175.49, 688.54, 202.59], \"box_3d\": [1.58, 1.67, 3.84, 3.87, 1.77, 44.49], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": 2.74, \"box_2d\": [962.47, 162.51, 1053.22, 192.8], \"box_3d\": [1.39, 1.51, 3.49, 19.05, 0.93, 34.71], \"ry\": -3.04}]\n```" - } - ] - }, - { - "id": 97, - "data_source": "Mono3DRefer", - "scene_name": "000185", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.28, \"box_2d\": [0.0, 206.52, 264.52, 374.0], \"box_3d\": [1.56, 1.55, 3.64, -3.09, 1.77, 3.03], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [181.07, 157.35, 454.31, 374.0], \"box_3d\": [1.91, 1.66, 4.22, -3.06, 1.8, 8.43], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [400.82, 173.04, 526.56, 281.06], \"box_3d\": [1.83, 1.69, 4.28, -2.79, 1.86, 14.46], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [544.45, 185.49, 606.56, 225.44], \"box_3d\": [1.51, 1.63, 4.03, -1.39, 2.06, 30.09], \"ry\": 1.77}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [586.23, 184.08, 635.02, 217.72], \"box_3d\": [1.57, 1.44, 3.6, 0.05, 2.16, 36.27], \"ry\": 1.87}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [630.37, 185.26, 673.44, 212.54], \"box_3d\": [1.47, 1.51, 3.5, 2.43, 2.2, 41.59], \"ry\": 1.92}]\n```" - } - ] - }, - { - "id": 98, - "data_source": "Mono3DRefer", - "scene_name": "000191", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.86, 718.86), (cx, cy) = (607.19, 185.22). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [577.55, 178.14, 614.2, 211.56], \"box_3d\": [1.6, 1.76, 3.84, -0.64, 1.25, 36.71], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -0.71, \"box_2d\": [0.0, 206.71, 198.09, 375.0], \"box_3d\": [1.48, 1.64, 3.81, -3.52, 1.64, 2.74], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.19, \"box_2d\": [233.74, 188.79, 362.78, 253.97], \"box_3d\": [1.39, 1.62, 3.73, -7.57, 1.51, 17.87], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.35, \"box_2d\": [373.86, 190.62, 501.68, 272.04], \"box_3d\": [1.37, 1.61, 4.21, -3.37, 1.52, 14.87], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.29, \"box_2d\": [358.25, 184.88, 432.5, 227.53], \"box_3d\": [1.45, 1.64, 3.86, -7.92, 1.46, 27.05], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.42, \"box_2d\": [462.02, 184.85, 518.69, 221.0], \"box_3d\": [1.38, 1.71, 4.49, -4.93, 1.39, 30.53], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -0.89, \"box_2d\": [0.0, 188.07, 124.51, 269.52], \"box_3d\": [1.49, 1.59, 4.32, -12.27, 1.58, 14.83], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.12, \"box_2d\": [206.69, 187.95, 303.2, 230.07], \"box_3d\": [1.33, 1.55, 3.83, -12.48, 1.46, 25.63], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.21, \"box_2d\": [289.34, 179.48, 361.92, 217.7], \"box_3d\": [1.63, 1.71, 4.13, -12.87, 1.38, 32.99], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 99, - "data_source": "Mono3DRefer", - "scene_name": "000192", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.86, 718.86), (cx, cy) = (607.19, 185.22). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [573.15, 179.59, 605.63, 209.19], \"box_3d\": [1.6, 1.76, 3.84, -1.07, 1.29, 41.19], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -0.93, \"box_2d\": [0.0, 188.97, 175.41, 325.79], \"box_3d\": [1.54, 1.65, 3.82, -7.94, 1.62, 9.88], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.1, \"box_2d\": [140.95, 187.68, 315.55, 268.37], \"box_3d\": [1.49, 1.67, 4.05, -8.15, 1.58, 15.82], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.42, \"box_2d\": [450.93, 185.93, 512.69, 229.68], \"box_3d\": [1.41, 1.58, 3.99, -4.5, 1.46, 25.97], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.44, \"box_2d\": [487.99, 184.1, 528.93, 213.69], \"box_3d\": [1.42, 1.61, 3.56, -5.09, 1.38, 37.03], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 100, - "data_source": "Mono3DRefer", - "scene_name": "000200", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [602.62, 178.85, 663.95, 238.07], \"box_3d\": [1.61, 1.66, 3.2, 0.64, 1.8, 21.4], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.9, \"box_2d\": [797.21, 186.52, 1018.27, 325.26], \"box_3d\": [1.43, 1.5, 3.38, 3.55, 1.63, 9.34], \"ry\": -1.55}]\n```" - } - ] - }, - { - "id": 101, - "data_source": "Mono3DRefer", - "scene_name": "000202", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [535.01, 180.59, 567.79, 204.85], \"box_3d\": [1.44, 1.72, 4.14, -3.68, 1.94, 45.59], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 102, - "data_source": "Mono3DRefer", - "scene_name": "000210", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.22, \"box_2d\": [0.0, 180.29, 155.87, 315.87], \"box_3d\": [1.8, 1.61, 3.83, -9.0, 1.93, 11.07], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 2.17, \"box_2d\": [124.67, 193.46, 286.04, 269.56], \"box_3d\": [1.45, 1.45, 3.4, -8.89, 1.94, 16.08], \"ry\": 1.67}, {\"category\": \"Car\", \"angle\": 1.98, \"box_2d\": [238.92, 179.34, 363.82, 247.43], \"box_3d\": [1.75, 1.71, 4.31, -8.8, 1.95, 20.9], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [376.25, 182.35, 444.03, 221.53], \"box_3d\": [1.53, 1.66, 4.01, -8.43, 1.94, 30.62], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [419.21, 181.52, 472.75, 213.59], \"box_3d\": [1.48, 1.6, 3.9, -8.11, 1.93, 35.82], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [498.77, 177.11, 530.95, 199.1], \"box_3d\": [1.61, 1.67, 4.56, -7.3, 1.96, 55.63], \"ry\": 1.62}]\n```" - } - ] - }, - { - "id": 103, - "data_source": "Mono3DRefer", - "scene_name": "000214", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.32, \"box_2d\": [277.11, 180.84, 412.58, 232.04], \"box_3d\": [1.52, 1.64, 3.7, -8.37, 1.8, 22.9], \"ry\": -0.03}, {\"category\": \"Car\", \"angle\": -2.76, \"box_2d\": [334.05, 182.0, 443.19, 222.67], \"box_3d\": [1.5, 1.62, 3.69, -8.56, 1.87, 27.94], \"ry\": -3.05}, {\"category\": \"Car\", \"angle\": 1.49, \"box_2d\": [872.38, 162.78, 902.54, 184.24], \"box_3d\": [1.51, 1.7, 4.45, 20.53, 0.82, 53.55], \"ry\": 1.85}]\n```" - } - ] - }, - { - "id": 104, - "data_source": "Mono3DRefer", - "scene_name": "000215", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.09, \"box_2d\": [215.24, 190.26, 586.08, 316.48], \"box_3d\": [1.45, 1.68, 4.53, -2.75, 1.73, 9.96], \"ry\": -0.17}, {\"category\": \"Car\", \"angle\": -0.77, \"box_2d\": [943.77, 188.46, 1241.0, 314.86], \"box_3d\": [1.35, 1.35, 3.38, 6.08, 1.57, 9.01], \"ry\": -0.19}]\n```" - } - ] - }, - { - "id": 105, - "data_source": "Mono3DRefer", - "scene_name": "000217", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [316.11, 191.89, 461.97, 282.37], \"box_3d\": [1.37, 1.64, 4.4, -4.07, 1.78, 13.88], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [475.46, 183.89, 525.42, 222.67], \"box_3d\": [1.4, 1.45, 3.3, -4.27, 1.85, 28.24], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.58, \"box_2d\": [562.18, 178.55, 594.26, 206.23], \"box_3d\": [1.4, 1.63, 3.89, -1.73, 1.72, 38.85], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [537.61, 179.82, 561.21, 200.59], \"box_3d\": [1.38, 1.33, 4.32, -4.24, 1.88, 50.62], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 106, - "data_source": "Mono3DRefer", - "scene_name": "000219", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [566.3, 176.56, 616.74, 225.38], \"box_3d\": [1.61, 1.66, 3.2, -0.68, 1.76, 25.65], \"ry\": -1.59}]\n```" - } - ] - }, - { - "id": 107, - "data_source": "Mono3DRefer", - "scene_name": "000221", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.87, \"box_2d\": [776.78, 185.07, 955.98, 301.67], \"box_3d\": [1.43, 1.5, 3.38, 3.57, 1.63, 10.76], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [603.24, 178.88, 664.34, 237.88], \"box_3d\": [1.61, 1.66, 3.2, 0.66, 1.8, 21.47], \"ry\": -1.54}]\n```" - } - ] - }, - { - "id": 108, - "data_source": "Mono3DRefer", - "scene_name": "000222", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [425.71, 181.46, 464.15, 201.44], \"box_3d\": [1.4, 1.75, 4.75, -12.3, 2.06, 53.93], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [518.65, 165.8, 740.84, 374.0], \"box_3d\": [1.55, 1.63, 3.32, 0.1, 1.52, 7.01], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -2.7, \"box_2d\": [344.82, 184.22, 457.2, 218.38], \"box_3d\": [1.46, 1.85, 4.52, -9.45, 2.01, 32.65], \"ry\": -2.98}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [376.56, 185.33, 418.02, 205.3], \"box_3d\": [1.37, 1.82, 3.96, -15.64, 2.31, 53.18], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -2.39, \"box_2d\": [175.97, 194.37, 259.89, 227.15], \"box_3d\": [1.52, 1.8, 3.3, -19.35, 2.61, 35.59], \"ry\": -2.88}, {\"category\": \"Car\", \"angle\": -2.76, \"box_2d\": [426.64, 178.59, 498.57, 210.48], \"box_3d\": [1.6, 1.6, 3.29, -7.67, 1.92, 37.52], \"ry\": -2.96}, {\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [334.95, 187.99, 374.32, 208.15], \"box_3d\": [1.38, 1.79, 2.94, -18.59, 2.51, 52.56], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 109, - "data_source": "Mono3DRefer", - "scene_name": "000225", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.22, \"box_2d\": [1.07, 175.43, 121.82, 230.51], \"box_3d\": [1.92, 1.58, 3.93, -20.7, 2.04, 27.39], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [584.46, 176.63, 622.38, 214.4], \"box_3d\": [1.62, 1.63, 4.5, -0.32, 1.82, 33.56], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 110, - "data_source": "Mono3DRefer", - "scene_name": "000227", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [783.58, 179.57, 909.44, 248.34], \"box_3d\": [1.48, 1.51, 4.35, 5.94, 1.66, 17.92], \"ry\": 2.25}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [1081.89, 174.98, 1130.16, 204.12], \"box_3d\": [1.46, 1.74, 3.99, 26.31, 1.58, 38.25], \"ry\": 2.26}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [864.51, 181.18, 964.28, 233.34], \"box_3d\": [1.42, 1.68, 4.29, 9.43, 1.7, 22.26], \"ry\": 2.25}, {\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [945.94, 179.75, 1009.13, 217.58], \"box_3d\": [1.36, 1.4, 3.8, 14.4, 1.65, 28.24], \"ry\": 2.26}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [1019.96, 177.36, 1060.88, 205.58], \"box_3d\": [1.44, 1.65, 2.96, 23.12, 1.7, 38.77], \"ry\": 2.18}]\n```" - } - ] - }, - { - "id": 111, - "data_source": "Mono3DRefer", - "scene_name": "000228", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.57, \"box_2d\": [219.58, 178.89, 342.65, 219.2], \"box_3d\": [1.35, 1.6, 3.86, -12.04, 1.59, 26.54], \"ry\": -1.0}, {\"category\": \"Car\", \"angle\": -0.8, \"box_2d\": [444.17, 173.3, 522.3, 205.63], \"box_3d\": [1.49, 1.53, 3.8, -6.17, 1.53, 35.28], \"ry\": -0.97}, {\"category\": \"Car\", \"angle\": -1.01, \"box_2d\": [581.13, 173.89, 633.85, 198.94], \"box_3d\": [1.49, 1.56, 3.65, -0.14, 1.57, 44.92], \"ry\": -1.01}]\n```" - } - ] - }, - { - "id": 112, - "data_source": "Mono3DRefer", - "scene_name": "000230", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [396.86, 183.78, 438.89, 206.49], \"box_3d\": [1.55, 1.88, 4.59, -14.0, 2.37, 52.75], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 113, - "data_source": "Mono3DRefer", - "scene_name": "000236", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.86, 718.86), (cx, cy) = (607.19, 185.22). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [580.36, 176.15, 621.91, 214.02], \"box_3d\": [1.6, 1.76, 3.84, -0.32, 1.21, 32.61], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.06, \"box_2d\": [76.17, 175.86, 288.33, 278.43], \"box_3d\": [1.65, 1.66, 3.96, -7.81, 1.49, 13.61], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.29, \"box_2d\": [323.02, 185.25, 476.65, 282.71], \"box_3d\": [1.48, 1.64, 3.81, -3.62, 1.5, 13.08], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.22, \"box_2d\": [282.95, 183.87, 393.3, 241.13], \"box_3d\": [1.44, 1.71, 3.78, -7.56, 1.42, 20.45], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.43, \"box_2d\": [480.25, 183.11, 543.62, 226.2], \"box_3d\": [1.37, 1.61, 4.21, -3.34, 1.31, 25.49], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -1.38, \"box_2d\": [445.32, 180.31, 489.64, 208.47], \"box_3d\": [1.45, 1.64, 3.86, -7.64, 1.19, 39.34], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -0.91, \"box_2d\": [0.0, 185.85, 139.14, 265.32], \"box_3d\": [1.46, 1.51, 3.83, -12.08, 1.49, 15.38], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.11, \"box_2d\": [183.62, 181.18, 299.1, 234.05], \"box_3d\": [1.51, 1.54, 4.17, -11.52, 1.39, 22.85], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 114, - "data_source": "Mono3DRefer", - "scene_name": "000237", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.91, \"box_2d\": [855.98, 178.03, 953.05, 214.01], \"box_3d\": [1.42, 1.58, 3.48, 12.26, 1.66, 30.18], \"ry\": -2.53}, {\"category\": \"Car\", \"angle\": 2.07, \"box_2d\": [175.09, 192.03, 258.16, 222.09], \"box_3d\": [1.41, 1.85, 4.7, -20.93, 2.49, 38.57], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 115, - "data_source": "Mono3DRefer", - "scene_name": "000238", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.83, \"box_2d\": [0.0, 170.01, 58.09, 216.72], \"box_3d\": [1.52, 1.5, 3.51, -20.63, 1.44, 24.12], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": 1.99, \"box_2d\": [306.59, 174.28, 347.24, 194.2], \"box_3d\": [1.47, 1.82, 3.03, -21.75, 1.6, 55.48], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [574.32, 173.66, 615.03, 214.23], \"box_3d\": [1.62, 1.63, 4.5, -0.68, 1.69, 31.41], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 116, - "data_source": "Mono3DRefer", - "scene_name": "000240", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.58, \"box_2d\": [567.32, 177.55, 609.97, 215.63], \"box_3d\": [1.41, 1.59, 4.47, -0.88, 1.62, 29.36], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 117, - "data_source": "Mono3DRefer", - "scene_name": "000242", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.33, \"box_2d\": [635.14, 163.71, 725.68, 229.58], \"box_3d\": [1.79, 1.72, 4.35, 1.98, 1.57, 21.99], \"ry\": 1.42}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [389.43, 184.88, 455.66, 237.97], \"box_3d\": [1.45, 1.49, 4.07, -5.73, 1.84, 22.28], \"ry\": 1.44}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [432.71, 187.21, 478.54, 220.32], \"box_3d\": [1.27, 1.54, 3.58, -6.51, 1.9, 30.54], \"ry\": 1.46}]\n```" - } - ] - }, - { - "id": 118, - "data_source": "Mono3DRefer", - "scene_name": "000243", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [598.73, 177.45, 678.68, 248.0], \"box_3d\": [1.37, 1.56, 3.54, 0.54, 1.49, 15.96], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.43, \"box_2d\": [498.91, 169.77, 576.76, 230.99], \"box_3d\": [1.61, 1.69, 4.26, -2.06, 1.56, 21.28], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [620.2, 169.23, 672.92, 218.51], \"box_3d\": [1.53, 1.63, 4.5, 1.25, 1.45, 24.87], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": -1.02, \"box_2d\": [1190.53, 133.24, 1241.0, 204.3], \"box_3d\": [1.48, 1.6, 3.84, 14.87, 0.64, 15.68], \"ry\": -0.27}, {\"category\": \"Car\", \"angle\": -0.9, \"box_2d\": [1070.7, 135.74, 1238.26, 198.38], \"box_3d\": [1.55, 1.61, 3.64, 14.31, 0.63, 19.06], \"ry\": -0.26}]\n```" - } - ] - }, - { - "id": 119, - "data_source": "Mono3DRefer", - "scene_name": "000244", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [513.85, 182.74, 563.17, 223.35], \"box_3d\": [1.56, 1.55, 3.64, -2.94, 1.98, 29.97], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [539.66, 172.6, 581.58, 214.11], \"box_3d\": [1.91, 1.66, 4.22, -2.4, 1.92, 35.51], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [561.47, 175.94, 598.84, 209.67], \"box_3d\": [1.83, 1.69, 4.28, -1.69, 2.02, 41.52], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": 2.08, \"box_2d\": [0.0, 183.99, 421.32, 374.0], \"box_3d\": [1.63, 1.6, 3.73, -2.51, 1.71, 4.36], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -2.38, \"box_2d\": [948.87, 191.74, 1241.0, 374.0], \"box_3d\": [1.59, 1.49, 3.34, 2.63, 1.69, 2.33], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.86, \"box_2d\": [730.11, 191.65, 1000.19, 374.0], \"box_3d\": [1.54, 1.6, 3.52, 2.35, 1.77, 7.71], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.35, \"box_2d\": [307.67, 188.31, 487.75, 335.88], \"box_3d\": [1.54, 1.5, 3.48, -2.64, 1.77, 9.56], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [410.78, 188.92, 522.89, 272.41], \"box_3d\": [1.44, 1.63, 3.32, -2.82, 1.8, 14.63], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.82, \"box_2d\": [685.04, 181.86, 806.7, 271.17], \"box_3d\": [1.55, 1.55, 3.27, 2.53, 1.74, 14.37], \"ry\": -1.65}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [656.44, 183.23, 735.82, 245.57], \"box_3d\": [1.48, 1.6, 3.79, 2.18, 1.78, 19.43], \"ry\": -1.59}]\n```" - } - ] - }, - { - "id": 120, - "data_source": "Mono3DRefer", - "scene_name": "000245", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.49, \"box_2d\": [871.24, 138.44, 1241.0, 369.5], \"box_3d\": [1.85, 1.68, 4.34, 4.89, 1.58, 6.76], \"ry\": 3.1}, {\"category\": \"Car\", \"angle\": -2.51, \"box_2d\": [0.0, 199.5, 230.56, 336.43], \"box_3d\": [1.43, 1.56, 3.85, -6.94, 1.8, 8.64], \"ry\": 3.12}, {\"category\": \"Car\", \"angle\": -2.81, \"box_2d\": [174.45, 190.89, 368.89, 263.55], \"box_3d\": [1.42, 1.62, 3.5, -7.3, 1.85, 15.76], \"ry\": 3.04}, {\"category\": \"Car\", \"angle\": -2.94, \"box_2d\": [363.25, 182.66, 466.82, 221.54], \"box_3d\": [1.41, 1.54, 3.6, -7.48, 1.81, 27.77], \"ry\": 3.08}]\n```" - } - ] - }, - { - "id": 121, - "data_source": "Mono3DRefer", - "scene_name": "000247", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [275.5, 189.86, 310.27, 218.5], \"box_3d\": [1.45, 1.62, 3.83, -17.57, 2.42, 40.02], \"ry\": 1.19}]\n```" - } - ] - }, - { - "id": 122, - "data_source": "Mono3DRefer", - "scene_name": "000248", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.86, 718.86), (cx, cy) = (607.19, 185.22). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [577.33, 178.49, 611.19, 209.37], \"box_3d\": [1.6, 1.76, 3.84, -0.76, 1.25, 39.58], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.01, \"box_2d\": [0.0, 182.47, 245.23, 293.47], \"box_3d\": [1.64, 1.71, 4.12, -8.41, 1.61, 12.86], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.25, \"box_2d\": [264.17, 190.75, 441.26, 295.28], \"box_3d\": [1.41, 1.58, 3.99, -4.13, 1.54, 12.21], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.22, \"box_2d\": [264.99, 191.54, 369.63, 239.95], \"box_3d\": [1.22, 1.65, 3.65, -8.42, 1.45, 21.09], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.38, \"box_2d\": [430.72, 185.6, 502.46, 234.14], \"box_3d\": [1.42, 1.61, 3.56, -4.54, 1.45, 23.37], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.33, \"box_2d\": [381.56, 180.8, 439.65, 220.0], \"box_3d\": [1.61, 1.61, 3.68, -8.66, 1.43, 31.76], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.43, \"box_2d\": [473.84, 183.14, 521.92, 216.46], \"box_3d\": [1.43, 1.64, 3.99, -5.08, 1.35, 33.47], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.46, \"box_2d\": [506.71, 185.28, 540.44, 209.67], \"box_3d\": [1.35, 1.57, 3.72, -4.97, 1.37, 42.68], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 123, - "data_source": "Mono3DRefer", - "scene_name": "000249", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.86, 718.86), (cx, cy) = (607.19, 185.22). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [559.9, 176.25, 603.74, 216.23], \"box_3d\": [1.6, 1.76, 3.84, -1.17, 1.23, 31.02], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": -1.22, \"box_2d\": [102.42, 183.7, 425.73, 375.0], \"box_3d\": [1.58, 1.68, 3.95, -3.27, 1.58, 7.68], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -0.82, \"box_2d\": [0.0, 197.12, 93.16, 325.8], \"box_3d\": [1.42, 1.53, 4.2, -7.7, 1.61, 7.57], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.34, \"box_2d\": [357.57, 188.5, 467.94, 263.83], \"box_3d\": [1.39, 1.5, 3.4, -4.11, 1.49, 15.47], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.17, \"box_2d\": [188.31, 186.03, 331.32, 254.48], \"box_3d\": [1.43, 1.62, 4.08, -8.35, 1.48, 17.59], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.06, \"box_2d\": [96.72, 186.51, 228.83, 240.31], \"box_3d\": [1.39, 1.5, 4.04, -13.01, 1.45, 21.26], \"ry\": -1.6}]\n```" - } - ] - }, - { - "id": 124, - "data_source": "Mono3DRefer", - "scene_name": "000250", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [26.43, 196.96, 252.9, 345.96], \"box_3d\": [1.44, 1.58, 3.72, -5.9, 1.8, 9.45], \"ry\": 1.25}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [341.7, 192.77, 383.53, 224.69], \"box_3d\": [1.39, 1.52, 3.56, -11.99, 2.4, 35.04], \"ry\": 1.36}, {\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [231.91, 194.75, 338.29, 272.86], \"box_3d\": [1.46, 1.59, 3.35, -7.14, 1.99, 16.05], \"ry\": 1.32}]\n```" - } - ] - }, - { - "id": 125, - "data_source": "Mono3DRefer", - "scene_name": "000252", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [239.65, 196.51, 447.05, 327.01], \"box_3d\": [1.39, 1.52, 3.56, -3.49, 1.77, 10.0], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [515.45, 183.55, 549.41, 209.28], \"box_3d\": [1.34, 1.64, 4.57, -4.38, 1.97, 41.01], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": 2.98, \"box_2d\": [386.51, 182.63, 485.96, 218.8], \"box_3d\": [1.52, 1.41, 4.22, -7.92, 1.99, 32.8], \"ry\": 2.75}]\n```" - } - ] - }, - { - "id": 126, - "data_source": "Mono3DRefer", - "scene_name": "000253", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [568.83, 176.31, 624.55, 230.24], \"box_3d\": [1.61, 1.66, 3.2, -0.44, 1.74, 23.37], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [330.49, 188.79, 439.69, 251.21], \"box_3d\": [1.49, 1.67, 4.38, -6.19, 1.97, 20.24], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [639.48, 176.79, 659.27, 194.18], \"box_3d\": [1.43, 1.5, 3.38, 3.33, 1.78, 61.64], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 127, - "data_source": "Mono3DRefer", - "scene_name": "000254", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.06, \"box_2d\": [27.82, 201.43, 299.81, 344.53], \"box_3d\": [1.45, 1.52, 3.62, -5.74, 1.9, 9.75], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [380.94, 190.96, 499.06, 277.0], \"box_3d\": [1.46, 1.62, 4.04, -3.39, 1.87, 14.92], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [226.87, 195.63, 385.02, 287.64], \"box_3d\": [1.46, 1.58, 3.72, -5.76, 1.95, 14.05], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 2.47, \"box_2d\": [1109.86, 152.35, 1241.0, 254.2], \"box_3d\": [1.73, 1.55, 4.0, 11.57, 1.38, 13.03], \"ry\": -3.1}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [497.8, 182.98, 547.03, 225.38], \"box_3d\": [1.57, 1.6, 3.99, -3.51, 1.99, 29.17], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.62, \"box_2d\": [548.97, 182.69, 568.89, 200.98], \"box_3d\": [1.52, 1.54, 3.68, -4.44, 2.39, 62.89], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": -0.12, \"box_2d\": [632.82, 178.59, 724.49, 214.71], \"box_3d\": [1.62, 1.65, 4.14, 3.18, 1.92, 34.2], \"ry\": -0.03}, {\"category\": \"Car\", \"angle\": 0.01, \"box_2d\": [612.81, 177.29, 676.27, 203.8], \"box_3d\": [1.69, 1.76, 4.12, 2.27, 2.01, 47.94], \"ry\": 0.05}]\n```" - } - ] - }, - { - "id": 128, - "data_source": "Mono3DRefer", - "scene_name": "000256", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.29, \"box_2d\": [0.0, 185.45, 40.8, 241.07], \"box_3d\": [1.46, 1.4, 3.27, -16.91, 1.81, 18.85], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 2.23, \"box_2d\": [26.81, 171.39, 174.7, 232.27], \"box_3d\": [1.78, 1.6, 4.29, -16.32, 1.76, 23.37], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 2.16, \"box_2d\": [161.03, 181.62, 254.73, 222.65], \"box_3d\": [1.5, 1.47, 3.62, -15.87, 1.86, 28.6], \"ry\": 1.65}]\n```" - } - ] - }, - { - "id": 129, - "data_source": "Mono3DRefer", - "scene_name": "000257", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is ��/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.43, \"box_2d\": [792.7, 174.36, 857.54, 201.38], \"box_3d\": [1.48, 1.36, 3.51, 12.29, 1.59, 41.48], \"ry\": -2.15}]\n```" - } - ] - }, - { - "id": 130, - "data_source": "Mono3DRefer", - "scene_name": "000258", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.49, \"box_2d\": [676.96, 183.63, 707.15, 209.05], \"box_3d\": [1.54, 1.63, 3.67, 5.31, 2.26, 46.59], \"ry\": -1.38}, {\"category\": \"Car\", \"angle\": 2.07, \"box_2d\": [279.38, 194.03, 415.27, 267.25], \"box_3d\": [1.56, 1.59, 3.69, -6.45, 2.13, 18.04], \"ry\": 1.73}, {\"category\": \"Car\", \"angle\": 1.96, \"box_2d\": [375.59, 181.46, 474.75, 249.45], \"box_3d\": [1.93, 1.59, 4.01, -5.74, 2.22, 22.75], \"ry\": 1.72}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [705.24, 181.57, 763.02, 229.01], \"box_3d\": [1.58, 1.67, 3.84, 4.42, 1.91, 26.33], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": -1.53, \"box_2d\": [710.36, 170.09, 750.08, 221.15], \"box_3d\": [2.05, 1.5, 3.83, 5.2, 1.97, 31.07], \"ry\": -1.36}]\n```" - } - ] - }, - { - "id": 131, - "data_source": "Mono3DRefer", - "scene_name": "000259", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.33, \"box_2d\": [354.16, 178.12, 522.14, 291.51], \"box_3d\": [1.47, 1.59, 4.03, -2.39, 1.43, 11.39], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": -1.43, \"box_2d\": [486.95, 174.28, 555.67, 226.08], \"box_3d\": [1.41, 1.54, 3.36, -2.33, 1.21, 21.38], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [597.74, 166.79, 642.45, 207.57], \"box_3d\": [1.52, 1.67, 4.38, 0.76, 0.97, 29.34], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -1.51, \"box_2d\": [558.16, 164.69, 593.61, 196.84], \"box_3d\": [1.56, 1.59, 3.65, -1.27, 0.74, 37.11], \"ry\": -1.55}]\n```" - } - ] - }, - { - "id": 132, - "data_source": "Mono3DRefer", - "scene_name": "000262", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.91, \"box_2d\": [756.75, 181.67, 1038.61, 360.21], \"box_3d\": [1.45, 1.49, 3.79, 2.72, 1.56, 7.87], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [692.19, 178.73, 822.16, 271.94], \"box_3d\": [1.5, 1.53, 4.13, 2.59, 1.63, 13.82], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.42, \"box_2d\": [455.66, 180.62, 533.82, 241.13], \"box_3d\": [1.58, 1.65, 3.91, -3.29, 1.81, 21.0], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [674.1, 178.14, 742.63, 231.97], \"box_3d\": [1.47, 1.57, 4.09, 2.86, 1.64, 21.94], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [651.1, 178.0, 696.17, 215.64], \"box_3d\": [1.45, 1.58, 3.53, 2.55, 1.68, 29.93], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [635.97, 177.32, 669.6, 204.85], \"box_3d\": [1.46, 1.67, 4.4, 2.35, 1.74, 41.01], \"ry\": -1.56}]\n```" - } - ] - }, - { - "id": 133, - "data_source": "Mono3DRefer", - "scene_name": "000264", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.02, \"box_2d\": [818.02, 184.92, 1241.0, 374.0], \"box_3d\": [1.56, 1.48, 3.36, 2.79, 1.67, 5.53], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [653.87, 179.35, 715.48, 225.48], \"box_3d\": [1.38, 1.55, 3.6, 2.33, 1.6, 23.64], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 2.15, \"box_2d\": [71.39, 185.16, 205.32, 232.0], \"box_3d\": [1.35, 1.66, 4.09, -15.23, 1.77, 23.52], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [584.95, 176.28, 621.09, 209.27], \"box_3d\": [1.52, 1.67, 3.61, -0.38, 1.7, 35.35], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.97, \"box_2d\": [290.86, 177.87, 347.79, 206.42], \"box_3d\": [1.44, 1.52, 3.66, -15.52, 1.72, 38.61], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [568.86, 181.02, 626.19, 227.73], \"box_3d\": [1.36, 1.69, 3.38, -0.41, 1.64, 23.09], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.28, \"box_2d\": [269.26, 192.99, 478.9, 331.41], \"box_3d\": [1.4, 1.6, 3.35, -2.86, 1.7, 9.34], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.17, \"box_2d\": [133.71, 193.8, 327.83, 299.58], \"box_3d\": [1.37, 1.52, 3.3, -6.04, 1.72, 11.28], \"ry\": 0.69}, {\"category\": \"Car\", \"angle\": 1.13, \"box_2d\": [217.74, 186.79, 377.32, 265.99], \"box_3d\": [1.4, 1.61, 3.55, -6.49, 1.7, 14.72], \"ry\": 0.72}, {\"category\": \"Car\", \"angle\": 0.99, \"box_2d\": [341.28, 182.24, 455.26, 232.72], \"box_3d\": [1.4, 1.52, 3.76, -6.56, 1.71, 22.04], \"ry\": 0.7}, {\"category\": \"Car\", \"angle\": 0.86, \"box_2d\": [371.8, 177.0, 477.29, 225.87], \"box_3d\": [1.57, 1.64, 3.49, -6.47, 1.74, 24.91], \"ry\": 0.61}, {\"category\": \"Car\", \"angle\": 1.02, \"box_2d\": [421.83, 180.51, 497.71, 213.06], \"box_3d\": [1.38, 1.55, 3.89, -6.92, 1.75, 32.86], \"ry\": 0.81}, {\"category\": \"Car\", \"angle\": -1.45, \"box_2d\": [522.62, 180.83, 566.78, 212.11], \"box_3d\": [1.3, 1.55, 3.51, -2.9, 1.67, 32.28], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": 0.75, \"box_2d\": [457.86, 175.01, 529.62, 202.59], \"box_3d\": [1.57, 1.67, 4.2, -6.99, 1.73, 43.05], \"ry\": 0.59}]\n```" - } - ] - }, - { - "id": 134, - "data_source": "Mono3DRefer", - "scene_name": "000266", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.27, \"box_2d\": [147.94, 195.97, 352.08, 311.04], \"box_3d\": [1.34, 1.69, 3.6, -5.23, 1.74, 10.91], \"ry\": -1.71}, {\"category\": \"Car\", \"angle\": -1.86, \"box_2d\": [646.71, 163.55, 777.49, 258.96], \"box_3d\": [1.69, 1.59, 3.77, 1.9, 1.55, 14.83], \"ry\": -1.73}, {\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [264.17, 184.68, 398.12, 263.55], \"box_3d\": [1.47, 1.74, 4.41, -6.07, 1.76, 16.13], \"ry\": 1.46}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [581.33, 175.14, 640.65, 218.71], \"box_3d\": [1.49, 1.71, 3.58, -0.07, 1.6, 26.82], \"ry\": -1.71}, {\"category\": \"Car\", \"angle\": -1.5, \"box_2d\": [403.87, 183.79, 452.11, 217.83], \"box_3d\": [1.39, 1.81, 3.46, -8.04, 1.89, 31.98], \"ry\": -1.74}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [557.84, 175.64, 594.88, 206.2], \"box_3d\": [1.47, 1.55, 3.37, -1.79, 1.64, 36.88], \"ry\": -1.72}, {\"category\": \"Car\", \"angle\": -1.96, \"box_2d\": [748.25, 180.42, 1034.25, 353.77], \"box_3d\": [1.45, 1.61, 3.34, 2.68, 1.54, 7.82], \"ry\": -1.64}]\n```" - } - ] - }, - { - "id": 135, - "data_source": "Mono3DRefer", - "scene_name": "000268", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [550.79, 169.64, 715.78, 328.61], \"box_3d\": [1.66, 1.73, 3.05, 0.2, 1.65, 9.15], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [256.69, 181.58, 348.02, 228.84], \"box_3d\": [1.58, 1.63, 4.31, -11.24, 1.91, 26.59], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 136, - "data_source": "Mono3DRefer", - "scene_name": "000271", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.65, \"box_2d\": [903.8, 188.46, 976.33, 214.9], \"box_3d\": [1.68, 1.39, 4.26, 22.13, 2.76, 48.48], \"ry\": 3.08}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [557.85, 176.56, 614.45, 232.99], \"box_3d\": [1.62, 1.63, 4.5, -0.78, 1.76, 23.22], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 2.05, \"box_2d\": [238.89, 180.02, 292.21, 204.12], \"box_3d\": [1.46, 1.31, 4.15, -22.0, 1.93, 46.19], \"ry\": 1.61}]\n```" - } - ] - }, - { - "id": 137, - "data_source": "Mono3DRefer", - "scene_name": "000272", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [603.66, 176.98, 642.02, 203.59], \"box_3d\": [1.5, 1.7, 3.77, 0.78, 1.77, 43.0], \"ry\": 1.75}]\n```" - } - ] - }, - { - "id": 138, - "data_source": "Mono3DRefer", - "scene_name": "000275", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [563.62, 170.71, 670.02, 271.31], \"box_3d\": [1.56, 1.65, 3.69, 0.07, 1.55, 13.14], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.37, \"box_2d\": [971.99, 174.01, 1241.0, 374.0], \"box_3d\": [1.45, 1.62, 3.88, 4.57, 1.46, 5.92], \"ry\": -0.74}, {\"category\": \"Car\", \"angle\": -1.1, \"box_2d\": [800.46, 156.91, 1088.76, 308.56], \"box_3d\": [1.54, 1.73, 4.04, 4.38, 1.39, 9.27], \"ry\": -0.68}, {\"category\": \"Car\", \"angle\": -1.08, \"box_2d\": [775.67, 173.75, 964.49, 264.73], \"box_3d\": [1.38, 1.56, 3.8, 4.68, 1.41, 12.83], \"ry\": -0.74}, {\"category\": \"Car\", \"angle\": 2.04, \"box_2d\": [0.0, 167.79, 187.8, 263.06], \"box_3d\": [1.57, 1.71, 4.32, -10.18, 1.52, 14.15], \"ry\": 1.42}]\n```" - } - ] - }, - { - "id": 139, - "data_source": "Mono3DRefer", - "scene_name": "000276", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [441.72, 179.02, 519.66, 239.6], \"box_3d\": [1.52, 1.55, 3.79, -3.57, 1.71, 20.24], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [538.11, 179.68, 568.84, 204.48], \"box_3d\": [1.48, 1.62, 3.21, -3.52, 1.91, 45.03], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [589.8, 176.25, 632.11, 218.39], \"box_3d\": [1.62, 1.63, 4.5, 0.01, 1.79, 30.31], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 140, - "data_source": "Mono3DRefer", - "scene_name": "000280", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.97, \"box_2d\": [265.17, 193.28, 330.23, 223.5], \"box_3d\": [1.48, 1.79, 4.15, -17.12, 2.65, 39.7], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 141, - "data_source": "Mono3DRefer", - "scene_name": "000281", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [593.06, 169.73, 625.83, 201.27], \"box_3d\": [1.66, 1.73, 3.05, -0.06, 1.52, 39.97], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 142, - "data_source": "Mono3DRefer", - "scene_name": "000285", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [595.62, 181.33, 661.35, 243.71], \"box_3d\": [1.56, 1.65, 3.69, 0.41, 1.81, 20.13], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.24, \"box_2d\": [837.5, 168.67, 1076.72, 313.08], \"box_3d\": [1.54, 1.72, 3.88, 4.75, 1.51, 9.7], \"ry\": -0.79}, {\"category\": \"Car\", \"angle\": -1.01, \"box_2d\": [715.76, 178.58, 827.33, 237.3], \"box_3d\": [1.52, 1.49, 3.46, 4.65, 1.7, 20.7], \"ry\": -0.79}, {\"category\": \"Car\", \"angle\": -0.92, \"box_2d\": [677.52, 180.56, 771.35, 220.77], \"box_3d\": [1.39, 1.61, 3.65, 4.34, 1.7, 27.29], \"ry\": -0.77}, {\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [353.01, 183.81, 448.54, 243.22], \"box_3d\": [1.6, 1.6, 3.77, -6.21, 1.94, 21.68], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [484.19, 180.78, 533.53, 213.19], \"box_3d\": [1.61, 1.7, 3.97, -5.33, 2.04, 38.25], \"ry\": 1.67}, {\"category\": \"Car\", \"angle\": 2.07, \"box_2d\": [164.4, 191.99, 311.09, 253.28], \"box_3d\": [1.45, 1.65, 4.63, -10.38, 2.04, 20.48], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.96, \"box_2d\": [280.37, 180.38, 367.24, 233.95], \"box_3d\": [1.81, 1.59, 3.93, -10.48, 2.1, 26.61], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [555.46, 180.3, 589.04, 199.34], \"box_3d\": [1.45, 1.84, 4.59, -3.0, 2.06, 57.9], \"ry\": 1.71}, {\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [388.4, 182.18, 449.36, 214.91], \"box_3d\": [1.56, 1.62, 4.19, -9.79, 2.06, 37.15], \"ry\": 1.68}, {\"category\": \"Car\", \"angle\": 2.0, \"box_2d\": [471.91, 183.21, 521.83, 205.3], \"box_3d\": [1.42, 1.67, 4.53, -7.77, 2.16, 49.85], \"ry\": 1.84}, {\"category\": \"Car\", \"angle\": 2.35, \"box_2d\": [0.0, 188.86, 116.94, 251.89], \"box_3d\": [1.47, 1.48, 3.21, -14.94, 1.91, 18.94], \"ry\": 1.69}, {\"category\": \"Car\", \"angle\": 2.5, \"box_2d\": [11.83, 184.92, 151.34, 225.56], \"box_3d\": [1.47, 1.78, 4.34, -21.25, 1.99, 29.19], \"ry\": 1.88}]\n```" - } - ] - }, - { - "id": 143, - "data_source": "Mono3DRefer", - "scene_name": "000286", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.26, \"box_2d\": [347.45, 185.84, 425.72, 214.38], \"box_3d\": [1.52, 1.51, 3.94, -12.54, 2.28, 40.55], \"ry\": -0.04}, {\"category\": \"Car\", \"angle\": 0.42, \"box_2d\": [243.48, 186.4, 331.56, 216.19], \"box_3d\": [1.58, 1.56, 4.24, -18.08, 2.37, 40.5], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [627.58, 177.72, 652.34, 198.47], \"box_3d\": [1.49, 1.55, 3.93, 2.21, 1.88, 54.51], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [589.39, 176.3, 613.97, 199.68], \"box_3d\": [1.54, 1.63, 3.67, -0.59, 1.8, 50.02], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [405.67, 186.33, 516.16, 266.0], \"box_3d\": [1.45, 1.59, 3.59, -3.07, 1.76, 15.31], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -2.37, \"box_2d\": [955.63, 192.0, 1241.0, 374.0], \"box_3d\": [1.48, 1.47, 3.7, 2.93, 1.59, 2.72], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.94, \"box_2d\": [761.3, 175.99, 1116.88, 374.0], \"box_3d\": [1.64, 1.64, 3.89, 2.72, 1.69, 7.1], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.86, \"box_2d\": [718.78, 171.95, 863.66, 299.07], \"box_3d\": [1.69, 1.39, 2.65, 2.61, 1.7, 11.05], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [664.82, 178.8, 742.05, 237.1], \"box_3d\": [1.5, 1.59, 3.79, 2.54, 1.69, 20.71], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [639.38, 178.03, 675.93, 209.91], \"box_3d\": [1.49, 1.57, 4.07, 2.3, 1.76, 36.03], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [654.65, 176.84, 711.32, 224.76], \"box_3d\": [1.57, 1.59, 3.75, 2.51, 1.74, 25.84], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 144, - "data_source": "Mono3DRefer", - "scene_name": "000291", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [479.73, 180.34, 550.56, 223.09], \"box_3d\": [1.43, 1.68, 4.19, -3.41, 1.71, 26.44], \"ry\": 1.66}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [520.04, 176.85, 562.22, 210.91], \"box_3d\": [1.46, 1.59, 4.27, -3.16, 1.66, 33.43], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [543.28, 174.41, 576.92, 197.75], \"box_3d\": [1.38, 1.62, 4.06, -3.1, 1.5, 45.19], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [618.07, 173.69, 657.64, 213.4], \"box_3d\": [1.57, 1.56, 2.62, 1.12, 1.63, 30.16], \"ry\": -1.54}]\n```" - } - ] - }, - { - "id": 145, - "data_source": "Mono3DRefer", - "scene_name": "000292", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [597.57, 174.78, 664.8, 239.73], \"box_3d\": [1.61, 1.66, 3.2, 0.53, 1.68, 19.65], \"ry\": -1.55}]\n```" - } - ] - }, - { - "id": 146, - "data_source": "Mono3DRefer", - "scene_name": "000293", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.04, \"box_2d\": [0.0, 191.75, 403.27, 374.0], \"box_3d\": [1.5, 1.7, 3.77, -3.47, 1.72, 6.93], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -1.5, \"box_2d\": [494.78, 182.16, 529.62, 214.42], \"box_3d\": [1.57, 1.55, 3.32, -5.05, 2.06, 37.31], \"ry\": -1.64}, {\"category\": \"Car\", \"angle\": -3.05, \"box_2d\": [330.97, 174.77, 444.98, 218.79], \"box_3d\": [1.77, 1.73, 4.53, -9.48, 1.88, 31.02], \"ry\": 2.94}]\n```" - } - ] - }, - { - "id": 147, - "data_source": "Mono3DRefer", - "scene_name": "000295", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [286.19, 186.56, 418.07, 282.76], \"box_3d\": [1.49, 1.67, 4.52, -4.8, 1.78, 13.89], \"ry\": 1.39}]\n```" - } - ] - }, - { - "id": 148, - "data_source": "Mono3DRefer", - "scene_name": "000296", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.4, \"box_2d\": [685.75, 164.92, 739.44, 208.12], \"box_3d\": [1.73, 1.6, 4.11, 4.46, 1.44, 31.14], \"ry\": -1.26}, {\"category\": \"Car\", \"angle\": 2.18, \"box_2d\": [224.94, 186.16, 335.61, 238.74], \"box_3d\": [1.46, 1.37, 3.45, -10.13, 1.9, 22.34], \"ry\": 1.76}, {\"category\": \"Car\", \"angle\": -0.51, \"box_2d\": [0.0, 189.0, 193.29, 336.93], \"box_3d\": [1.53, 1.45, 4.33, -7.13, 1.74, 8.0], \"ry\": -1.22}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [429.91, 181.14, 493.11, 211.5], \"box_3d\": [1.4, 1.68, 4.76, -7.39, 1.83, 36.24], \"ry\": 1.69}]\n```" - } - ] - }, - { - "id": 149, - "data_source": "Mono3DRefer", - "scene_name": "000297", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [412.32, 186.04, 460.72, 217.29], \"box_3d\": [1.56, 1.73, 4.37, -9.44, 2.31, 39.45], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [458.94, 183.36, 488.76, 206.38], \"box_3d\": [1.6, 1.67, 2.65, -9.87, 2.38, 52.36], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 150, - "data_source": "Mono3DRefer", - "scene_name": "000300", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.43, \"box_2d\": [785.01, 175.65, 848.7, 202.34], \"box_3d\": [1.48, 1.36, 3.51, 11.97, 1.67, 42.02], \"ry\": -2.15}]\n```" - } - ] - }, - { - "id": 151, - "data_source": "Mono3DRefer", - "scene_name": "000301", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.86, 718.86), (cx, cy) = (607.19, 185.22). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [553.85, 173.51, 628.19, 241.31], \"box_3d\": [1.6, 1.76, 3.84, -0.48, 1.32, 19.08], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.36, \"box_2d\": [358.8, 191.12, 496.37, 276.59], \"box_3d\": [1.3, 1.61, 4.39, -3.29, 1.45, 13.79], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.41, \"box_2d\": [442.41, 179.69, 525.35, 236.97], \"box_3d\": [1.46, 1.66, 4.29, -3.46, 1.31, 20.58], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.45, \"box_2d\": [485.9, 173.99, 543.29, 220.87], \"box_3d\": [1.65, 1.7, 3.9, -3.5, 1.24, 27.4], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.35, \"box_2d\": [409.78, 174.99, 462.77, 212.3], \"box_3d\": [1.66, 1.62, 3.95, -8.12, 1.2, 34.23], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.49, \"box_2d\": [517.51, 175.34, 558.56, 209.96], \"box_3d\": [1.64, 1.72, 4.24, -3.51, 1.16, 36.56], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 152, - "data_source": "Mono3DRefer", - "scene_name": "000302", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.7, \"box_2d\": [399.86, 175.1, 505.05, 218.48], \"box_3d\": [1.48, 1.56, 3.62, -5.43, 1.25, 26.45], \"ry\": -0.9}, {\"category\": \"Car\", \"angle\": 2.05, \"box_2d\": [328.81, 169.19, 397.46, 202.72], \"box_3d\": [1.68, 1.67, 4.29, -12.65, 1.06, 38.43], \"ry\": 1.73}, {\"category\": \"Car\", \"angle\": -1.42, \"box_2d\": [562.87, 172.16, 598.61, 196.02], \"box_3d\": [1.43, 1.66, 4.15, -1.25, 0.87, 45.6], \"ry\": -1.46}]\n```" - } - ] - }, - { - "id": 153, - "data_source": "Mono3DRefer", - "scene_name": "000306", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [556.85, 173.12, 619.99, 236.08], \"box_3d\": [1.62, 1.63, 4.5, -0.67, 1.66, 21.05], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 2.22, \"box_2d\": [13.95, 169.57, 93.39, 206.59], \"box_3d\": [1.7, 1.64, 2.85, -26.79, 1.57, 34.79], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 154, - "data_source": "Mono3DRefer", - "scene_name": "000307", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [336.48, 186.85, 384.22, 213.59], \"box_3d\": [1.67, 1.87, 3.69, -16.67, 2.63, 48.29], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 155, - "data_source": "Mono3DRefer", - "scene_name": "000310", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [550.26, 175.17, 644.51, 244.53], \"box_3d\": [1.6, 1.7, 3.56, -0.47, 1.68, 18.68], \"ry\": -1.8}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [524.12, 179.11, 600.18, 231.15], \"box_3d\": [1.52, 1.73, 3.89, -1.68, 1.75, 23.47], \"ry\": -1.83}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [275.06, 191.43, 339.86, 242.11], \"box_3d\": [1.44, 1.6, 3.79, -9.83, 2.08, 23.56], \"ry\": 1.27}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [502.96, 182.68, 553.93, 219.83], \"box_3d\": [1.37, 1.51, 3.49, -3.34, 1.77, 28.76], \"ry\": -1.83}, {\"category\": \"Car\", \"angle\": -1.51, \"box_2d\": [790.84, 179.16, 1010.03, 371.7], \"box_3d\": [1.51, 1.69, 4.04, 3.18, 1.59, 7.88], \"ry\": -1.14}, {\"category\": \"Car\", \"angle\": -1.91, \"box_2d\": [613.46, 180.93, 789.58, 289.35], \"box_3d\": [1.46, 1.58, 3.93, 1.26, 1.61, 11.93], \"ry\": -1.81}, {\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [107.38, 194.33, 285.97, 295.88], \"box_3d\": [1.5, 1.68, 4.03, -7.49, 1.95, 13.44], \"ry\": 1.38}]\n```" - } - ] - }, - { - "id": 156, - "data_source": "Mono3DRefer", - "scene_name": "000311", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.31, \"box_2d\": [230.68, 177.46, 487.51, 349.05], \"box_3d\": [1.47, 1.6, 4.41, -2.6, 1.54, 8.47], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.93, \"box_2d\": [763.83, 158.53, 1241.0, 374.0], \"box_3d\": [1.63, 1.69, 3.88, 2.38, 1.58, 5.61], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -1.32, \"box_2d\": [393.96, 184.55, 542.19, 278.98], \"box_3d\": [1.46, 1.68, 4.1, -2.49, 1.7, 13.46], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [662.44, 180.11, 719.57, 223.61], \"box_3d\": [1.49, 1.64, 4.17, 2.92, 1.77, 27.05], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [713.49, 179.11, 879.89, 290.64], \"box_3d\": [1.41, 1.56, 3.97, 2.63, 1.52, 11.23], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [553.74, 176.47, 586.85, 206.69], \"box_3d\": [1.6, 1.75, 3.89, -2.21, 1.82, 40.55], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [648.61, 177.76, 686.28, 208.79], \"box_3d\": [1.59, 1.64, 4.03, 3.06, 1.87, 39.39], \"ry\": -1.59}]\n```" - } - ] - }, - { - "id": 157, - "data_source": "Mono3DRefer", - "scene_name": "000313", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.4, \"box_2d\": [274.89, 186.98, 426.62, 230.51], \"box_3d\": [1.36, 1.68, 4.49, -8.68, 1.86, 24.24], \"ry\": 0.06}]\n```" - } - ] - }, - { - "id": 158, - "data_source": "Mono3DRefer", - "scene_name": "000314", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [0.0, 205.78, 137.85, 333.01], \"box_3d\": [1.44, 1.58, 3.72, -8.16, 2.0, 10.66], \"ry\": 1.08}, {\"category\": \"Car\", \"angle\": 1.6, \"box_2d\": [193.4, 207.86, 231.88, 241.73], \"box_3d\": [1.39, 1.52, 3.56, -19.08, 3.15, 34.67], \"ry\": 1.1}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [99.41, 207.79, 203.65, 284.84], \"box_3d\": [1.46, 1.59, 3.35, -10.63, 2.35, 16.86], \"ry\": 1.15}]\n```" - } - ] - }, - { - "id": 159, - "data_source": "Mono3DRefer", - "scene_name": "000315", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [630.9, 176.65, 662.81, 197.27], \"box_3d\": [1.5, 1.7, 3.77, 2.82, 1.81, 54.91], \"ry\": 1.82}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [722.85, 177.56, 858.36, 295.97], \"box_3d\": [1.57, 1.56, 2.62, 2.57, 1.66, 10.95], \"ry\": -1.49}, {\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [308.02, 190.2, 511.1, 319.33], \"box_3d\": [1.49, 1.53, 3.88, -2.75, 1.78, 10.67], \"ry\": 1.69}]\n```" - } - ] - }, - { - "id": 160, - "data_source": "Mono3DRefer", - "scene_name": "000316", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [588.34, 175.63, 612.67, 198.77], \"box_3d\": [1.54, 1.63, 3.67, -0.67, 1.76, 50.53], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [465.69, 182.44, 538.49, 237.14], \"box_3d\": [1.45, 1.59, 3.59, -3.12, 1.74, 21.22], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -2.04, \"box_2d\": [814.81, 185.31, 1241.0, 374.0], \"box_3d\": [1.48, 1.47, 3.7, 2.93, 1.6, 5.87], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.84, \"box_2d\": [719.32, 174.47, 919.14, 316.36], \"box_3d\": [1.64, 1.64, 3.89, 2.7, 1.69, 10.38], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.82, \"box_2d\": [695.26, 171.46, 800.62, 266.75], \"box_3d\": [1.69, 1.39, 2.65, 2.6, 1.69, 14.3], \"ry\": -1.64}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [657.35, 177.99, 722.28, 227.74], \"box_3d\": [1.5, 1.59, 3.79, 2.53, 1.69, 23.94], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [636.71, 177.31, 669.73, 206.44], \"box_3d\": [1.49, 1.57, 4.07, 2.28, 1.75, 39.25], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [647.98, 176.47, 697.05, 218.73], \"box_3d\": [1.57, 1.59, 3.75, 2.42, 1.74, 29.04], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 0.22, \"box_2d\": [365.66, 182.66, 437.48, 208.97], \"box_3d\": [1.52, 1.51, 3.94, -12.62, 2.14, 43.74], \"ry\": -0.05}, {\"category\": \"Car\", \"angle\": 0.4, \"box_2d\": [268.54, 182.74, 349.53, 210.2], \"box_3d\": [1.58, 1.56, 4.24, -18.19, 2.21, 43.65], \"ry\": 0.0}, {\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [626.13, 176.0, 649.43, 195.55], \"box_3d\": [1.49, 1.55, 3.93, 2.18, 1.77, 57.74], \"ry\": -1.61}]\n```" - } - ] - }, - { - "id": 161, - "data_source": "Mono3DRefer", - "scene_name": "000319", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [599.09, 176.71, 627.07, 204.35], \"box_3d\": [1.65, 1.67, 3.64, 0.16, 1.91, 45.33], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.02, \"box_2d\": [178.46, 193.54, 291.55, 245.16], \"box_3d\": [1.41, 1.55, 4.16, -11.89, 2.12, 23.12], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.98, \"box_2d\": [236.96, 190.66, 348.83, 249.14], \"box_3d\": [1.38, 1.45, 3.63, -8.55, 1.9, 19.67], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [449.64, 182.84, 485.44, 207.63], \"box_3d\": [1.41, 1.53, 3.98, -8.65, 2.03, 43.94], \"ry\": 1.53}]\n```" - } - ] - }, - { - "id": 162, - "data_source": "Mono3DRefer", - "scene_name": "000321", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [528.71, 174.85, 554.02, 195.74], \"box_3d\": [1.39, 1.48, 3.76, -4.78, 1.55, 50.37], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [131.71, 187.57, 430.83, 370.81], \"box_3d\": [1.56, 1.68, 3.91, -3.43, 1.75, 8.31], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [450.5, 185.08, 523.42, 231.98], \"box_3d\": [1.43, 1.75, 4.47, -4.18, 1.88, 24.95], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 163, - "data_source": "Mono3DRefer", - "scene_name": "000323", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.24, \"box_2d\": [3.78, 210.08, 329.81, 358.59], \"box_3d\": [1.47, 1.58, 3.55, -5.49, 1.97, 8.9], \"ry\": -2.78}, {\"category\": \"Car\", \"angle\": -1.46, \"box_2d\": [704.29, 172.82, 803.3, 254.27], \"box_3d\": [1.6, 1.75, 3.89, 3.31, 1.62, 16.27], \"ry\": -1.26}, {\"category\": \"Car\", \"angle\": -1.88, \"box_2d\": [960.23, 157.46, 1142.32, 258.19], \"box_3d\": [1.59, 1.64, 4.03, 7.96, 1.37, 13.52], \"ry\": -1.36}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [932.24, 163.38, 1038.42, 226.34], \"box_3d\": [1.5, 1.8, 4.5, 9.98, 1.3, 19.6], \"ry\": -1.26}, {\"category\": \"Car\", \"angle\": -1.49, \"box_2d\": [768.08, 171.91, 813.8, 210.03], \"box_3d\": [1.46, 1.56, 3.21, 7.39, 1.45, 29.45], \"ry\": -1.25}, {\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [905.56, 154.84, 956.84, 203.11], \"box_3d\": [1.83, 1.75, 4.52, 13.1, 1.17, 29.69], \"ry\": -1.19}]\n```" - } - ] - }, - { - "id": 164, - "data_source": "Mono3DRefer", - "scene_name": "000326", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [687.66, 184.67, 807.99, 262.11], \"box_3d\": [1.38, 1.51, 4.1, 2.72, 1.65, 15.32], \"ry\": -1.64}, {\"category\": \"Car\", \"angle\": -2.23, \"box_2d\": [911.39, 157.06, 1241.0, 374.0], \"box_3d\": [1.6, 1.66, 3.55, 3.65, 1.53, 4.9], \"ry\": -1.62}]\n```" - } - ] - }, - { - "id": 165, - "data_source": "Mono3DRefer", - "scene_name": "000329", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [600.79, 179.94, 628.52, 207.34], \"box_3d\": [1.65, 1.67, 3.64, 0.25, 2.11, 45.71], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [501.21, 184.53, 524.24, 201.5], \"box_3d\": [1.38, 1.45, 3.63, -8.35, 2.4, 62.03], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 166, - "data_source": "Mono3DRefer", - "scene_name": "000334", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.95, \"box_2d\": [689.27, 181.33, 723.69, 201.8], \"box_3d\": [1.48, 1.36, 3.51, 7.24, 2.13, 54.52], \"ry\": -1.82}]\n```" - } - ] - }, - { - "id": 167, - "data_source": "Mono3DRefer", - "scene_name": "000335", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [456.45, 171.7, 543.78, 242.16], \"box_3d\": [1.7, 1.63, 4.08, -2.91, 1.7, 19.62], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": -1.86, \"box_2d\": [752.69, 181.01, 1000.64, 373.47], \"box_3d\": [1.59, 1.59, 2.47, 2.42, 1.68, 7.25], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [706.66, 189.08, 833.31, 287.05], \"box_3d\": [1.37, 1.59, 3.22, 2.49, 1.67, 12.06], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": -1.42, \"box_2d\": [509.97, 181.33, 568.9, 221.43], \"box_3d\": [1.4, 1.67, 3.72, -2.65, 1.74, 27.47], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": 1.51, \"box_2d\": [640.67, 177.82, 670.7, 203.29], \"box_3d\": [1.49, 1.61, 3.74, 2.75, 1.81, 44.43], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -2.65, \"box_2d\": [72.94, 166.25, 360.59, 260.59], \"box_3d\": [1.59, 1.63, 4.43, -7.16, 1.51, 13.33], \"ry\": -3.13}]\n```" - } - ] - }, - { - "id": 168, - "data_source": "Mono3DRefer", - "scene_name": "000339", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [346.34, 185.91, 427.94, 241.3], \"box_3d\": [1.42, 1.51, 3.71, -6.42, 1.82, 20.98], \"ry\": 1.48}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [634.11, 180.06, 712.16, 237.37], \"box_3d\": [1.37, 1.51, 3.49, 1.55, 1.57, 19.19], \"ry\": -1.65}, {\"category\": \"Car\", \"angle\": 2.37, \"box_2d\": [0.0, 203.67, 150.67, 374.0], \"box_3d\": [1.5, 1.68, 4.03, -5.37, 1.79, 5.01], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.88, \"box_2d\": [723.32, 167.34, 974.03, 337.08], \"box_3d\": [1.6, 1.7, 3.56, 2.52, 1.57, 8.66], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.8, \"box_2d\": [661.72, 174.72, 802.4, 268.54], \"box_3d\": [1.52, 1.73, 3.89, 2.12, 1.59, 13.86], \"ry\": -1.66}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [265.91, 190.56, 396.89, 272.86], \"box_3d\": [1.44, 1.6, 3.79, -5.75, 1.85, 15.22], \"ry\": 1.48}]\n```" - } - ] - }, - { - "id": 169, - "data_source": "Mono3DRefer", - "scene_name": "000343", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [795.44, 165.84, 836.84, 206.61], \"box_3d\": [1.82, 1.65, 3.67, 9.69, 1.53, 34.2], \"ry\": -1.35}, {\"category\": \"Car\", \"angle\": -1.83, \"box_2d\": [870.06, 158.97, 1051.89, 278.85], \"box_3d\": [1.72, 1.67, 4.42, 5.84, 1.55, 12.68], \"ry\": -1.4}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [608.99, 177.0, 687.68, 239.1], \"box_3d\": [1.44, 1.56, 3.84, 1.05, 1.56, 18.72], \"ry\": 1.75}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [684.7, 172.14, 719.08, 205.22], \"box_3d\": [1.5, 1.55, 4.21, 4.46, 1.5, 35.16], \"ry\": -1.44}]\n```" - } - ] - }, - { - "id": 170, - "data_source": "Mono3DRefer", - "scene_name": "000345", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.86, 718.86), (cx, cy) = (607.19, 185.22). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [569.9, 176.8, 622.56, 224.82], \"box_3d\": [1.6, 1.76, 3.84, -0.46, 1.31, 26.13], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.27, \"box_2d\": [244.92, 192.9, 443.88, 309.21], \"box_3d\": [1.36, 1.62, 3.99, -3.76, 1.52, 10.91], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.29, \"box_2d\": [361.07, 181.46, 429.1, 221.25], \"box_3d\": [1.4, 1.46, 3.84, -8.09, 1.27, 27.54], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.26, \"box_2d\": [342.45, 176.51, 395.61, 206.98], \"box_3d\": [1.5, 1.57, 3.63, -12.41, 1.06, 37.49], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.53, \"box_2d\": [535.64, 181.22, 564.13, 202.81], \"box_3d\": [1.3, 1.61, 4.39, -3.68, 1.05, 46.02], \"ry\": -1.61}]\n```" - } - ] - }, - { - "id": 171, - "data_source": "Mono3DRefer", - "scene_name": "000349", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [291.36, 186.88, 456.08, 286.55], \"box_3d\": [1.44, 1.61, 3.66, -3.96, 1.71, 12.58], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 2.03, \"box_2d\": [250.72, 188.18, 385.56, 257.82], \"box_3d\": [1.41, 1.53, 3.37, -6.67, 1.79, 16.73], \"ry\": 1.66}, {\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [615.88, 174.36, 669.64, 226.21], \"box_3d\": [1.61, 1.66, 3.2, 1.08, 1.68, 24.19], \"ry\": -1.51}]\n```" - } - ] - }, - { - "id": 172, - "data_source": "Mono3DRefer", - "scene_name": "000352", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.79, \"box_2d\": [0.0, 204.1, 95.82, 332.11], \"box_3d\": [1.5, 1.52, 3.65, -9.2, 1.95, 9.54], \"ry\": 0.04}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [710.76, 170.47, 841.04, 267.61], \"box_3d\": [1.56, 1.57, 3.78, 2.91, 1.55, 13.6], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [299.47, 192.19, 445.76, 284.47], \"box_3d\": [1.47, 1.71, 3.85, -4.47, 1.88, 14.05], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [397.5, 176.83, 489.93, 248.55], \"box_3d\": [1.73, 1.65, 3.7, -4.4, 1.85, 19.41], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [670.6, 174.13, 742.17, 229.83], \"box_3d\": [1.54, 1.66, 3.79, 2.82, 1.6, 22.06], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [455.68, 180.68, 520.43, 233.87], \"box_3d\": [1.68, 1.63, 3.94, -4.18, 1.96, 25.06], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [661.58, 177.82, 724.1, 218.7], \"box_3d\": [1.48, 1.64, 4.15, 3.15, 1.69, 28.47], \"ry\": -1.66}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [626.7, 179.69, 651.2, 198.67], \"box_3d\": [1.45, 1.61, 3.34, 2.26, 2.01, 57.48], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [529.25, 183.74, 555.11, 200.73], \"box_3d\": [1.34, 1.69, 3.6, -5.7, 2.26, 60.12], \"ry\": -1.79}, {\"category\": \"Car\", \"angle\": -1.92, \"box_2d\": [760.92, 175.17, 1106.31, 374.0], \"box_3d\": [1.51, 1.59, 4.28, 2.78, 1.56, 7.43], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 173, - "data_source": "Mono3DRefer", - "scene_name": "000353", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.02, \"box_2d\": [0.0, 177.92, 277.51, 356.22], \"box_3d\": [1.41, 1.72, 4.41, -5.33, 1.49, 7.92], \"ry\": 1.44}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [265.35, 157.96, 398.99, 238.6], \"box_3d\": [1.62, 1.75, 4.3, -6.32, 1.35, 16.81], \"ry\": 1.49}]\n```" - } - ] - }, - { - "id": 174, - "data_source": "Mono3DRefer", - "scene_name": "000356", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [632.32, 179.44, 744.64, 261.52], \"box_3d\": [1.57, 1.72, 3.52, 1.54, 1.72, 15.76], \"ry\": -1.67}, {\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [0.0, 96.93, 369.53, 374.0], \"box_3d\": [2.01, 1.89, 4.91, -3.55, 1.62, 6.43], \"ry\": 1.39}, {\"category\": \"Car\", \"angle\": -1.49, \"box_2d\": [408.14, 180.2, 467.6, 232.49], \"box_3d\": [1.65, 1.67, 3.81, -5.91, 1.91, 24.96], \"ry\": -1.72}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [612.33, 179.35, 678.37, 233.96], \"box_3d\": [1.52, 1.51, 3.1, 0.97, 1.73, 21.96], \"ry\": -1.68}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [588.02, 180.18, 643.47, 225.1], \"box_3d\": [1.53, 1.58, 3.53, 0.11, 1.81, 26.63], \"ry\": -1.7}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [570.29, 179.79, 603.15, 207.92], \"box_3d\": [1.46, 1.6, 3.71, -1.35, 1.85, 39.74], \"ry\": -1.65}]\n```" - } - ] - }, - { - "id": 175, - "data_source": "Mono3DRefer", - "scene_name": "000360", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.43, \"box_2d\": [776.46, 184.15, 929.46, 246.93], \"box_3d\": [1.42, 1.58, 3.48, 6.05, 1.72, 18.41], \"ry\": -2.12}, {\"category\": \"Car\", \"angle\": 1.99, \"box_2d\": [261.47, 187.05, 321.92, 217.2], \"box_3d\": [1.53, 1.81, 3.33, -17.39, 2.33, 39.51], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 176, - "data_source": "Mono3DRefer", - "scene_name": "000363", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [497.17, 176.9, 546.51, 215.54], \"box_3d\": [1.58, 1.6, 4.11, -3.85, 1.78, 31.83], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [546.67, 179.5, 570.87, 201.21], \"box_3d\": [1.52, 1.55, 3.79, -3.76, 2.02, 53.07], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [589.26, 175.64, 633.83, 220.04], \"box_3d\": [1.62, 1.63, 4.5, 0.01, 1.76, 28.89], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 177, - "data_source": "Mono3DRefer", - "scene_name": "000367", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [500.15, 178.76, 534.67, 203.63], \"box_3d\": [1.51, 1.81, 4.07, -5.95, 1.91, 46.56], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 178, - "data_source": "Mono3DRefer", - "scene_name": "000372", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [364.11, 184.83, 411.55, 208.32], \"box_3d\": [1.56, 1.91, 4.44, -15.82, 2.44, 51.55], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 2.19, \"box_2d\": [5.93, 196.11, 172.41, 263.52], \"box_3d\": [1.48, 1.71, 3.54, -13.28, 2.13, 18.61], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 179, - "data_source": "Mono3DRefer", - "scene_name": "000375", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.02, \"box_2d\": [152.36, 186.52, 377.23, 296.17], \"box_3d\": [1.42, 1.78, 3.62, -5.12, 1.53, 11.45], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 180, - "data_source": "Mono3DRefer", - "scene_name": "000379", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.23, \"box_2d\": [0.0, 194.65, 117.25, 259.47], \"box_3d\": [1.63, 1.87, 3.47, -16.34, 2.3, 20.76], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [340.99, 187.85, 386.73, 209.47], \"box_3d\": [1.33, 1.72, 3.63, -16.34, 2.35, 48.0], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 181, - "data_source": "Mono3DRefer", - "scene_name": "000380", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [602.22, 183.66, 675.02, 243.48], \"box_3d\": [1.36, 1.69, 3.38, 0.63, 1.66, 18.55], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.43, \"box_2d\": [1074.45, 174.52, 1235.26, 220.02], \"box_3d\": [1.45, 1.67, 4.32, 18.46, 1.53, 24.59], \"ry\": 3.07}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [364.22, 184.46, 419.74, 217.0], \"box_3d\": [1.59, 1.72, 3.86, -11.45, 2.22, 38.03], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [584.44, 182.86, 628.16, 223.09], \"box_3d\": [1.52, 1.67, 3.61, -0.17, 1.94, 29.53], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 182, - "data_source": "Mono3DRefer", - "scene_name": "000383", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (707.05, 707.05), (cx, cy) = (604.08, 180.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.99, \"box_2d\": [0.0, 193.02, 354.35, 369.0], \"box_3d\": [1.5, 1.78, 3.69, -3.21, 1.64, 4.56], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [452.52, 168.47, 485.07, 192.21], \"box_3d\": [1.45, 1.75, 4.23, -8.77, 0.7, 45.82], \"ry\": 1.46}]\n```" - } - ] - }, - { - "id": 183, - "data_source": "Mono3DRefer", - "scene_name": "000385", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.88, \"box_2d\": [754.96, 190.98, 1024.52, 371.56], \"box_3d\": [1.46, 1.54, 3.62, 2.68, 1.69, 7.91], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [505.9, 180.2, 564.96, 228.71], \"box_3d\": [1.52, 1.57, 4.42, -2.53, 1.78, 25.01], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.43, \"box_2d\": [668.54, 181.0, 743.45, 239.03], \"box_3d\": [1.51, 1.58, 3.92, 2.65, 1.75, 20.96], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.59, \"box_2d\": [552.32, 182.21, 582.49, 209.94], \"box_3d\": [1.46, 1.6, 3.33, -2.38, 1.99, 40.29], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [648.43, 178.32, 683.42, 209.11], \"box_3d\": [1.64, 1.7, 3.54, 3.08, 1.96, 40.62], \"ry\": -1.56}]\n```" - } - ] - }, - { - "id": 184, - "data_source": "Mono3DRefer", - "scene_name": "000386", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.86, 718.86), (cx, cy) = (607.19, 185.22). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.45, \"box_2d\": [498.87, 183.73, 546.62, 217.13], \"box_3d\": [1.37, 1.61, 4.21, -3.79, 1.32, 32.35], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.0, \"box_2d\": [40.27, 185.85, 205.47, 253.22], \"box_3d\": [1.5, 1.55, 3.99, -12.27, 1.54, 18.48], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [569.27, 178.19, 613.88, 218.86], \"box_3d\": [1.6, 1.76, 3.84, -0.71, 1.32, 30.51], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.12, \"box_2d\": [0.0, 180.33, 410.1, 375.0], \"box_3d\": [1.64, 1.72, 4.24, -3.19, 1.61, 6.23], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -0.96, \"box_2d\": [0.0, 190.67, 209.01, 310.58], \"box_3d\": [1.45, 1.61, 3.72, -7.92, 1.57, 10.89], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.25, \"box_2d\": [306.71, 177.56, 400.23, 233.63], \"box_3d\": [1.65, 1.66, 3.96, -8.14, 1.41, 23.27], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.37, \"box_2d\": [421.49, 183.49, 508.95, 243.09], \"box_3d\": [1.48, 1.64, 3.81, -3.9, 1.44, 20.07], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.33, \"box_2d\": [383.23, 182.22, 447.69, 219.8], \"box_3d\": [1.44, 1.71, 3.78, -7.97, 1.33, 30.01], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 185, - "data_source": "Mono3DRefer", - "scene_name": "000387", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [385.14, 187.58, 435.98, 213.25], \"box_3d\": [1.29, 1.57, 4.38, -11.05, 2.14, 40.16], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [466.19, 185.79, 492.19, 203.1], \"box_3d\": [1.4, 1.54, 3.67, -11.23, 2.53, 62.02], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -0.81, \"box_2d\": [633.8, 178.74, 674.57, 196.78], \"box_3d\": [1.53, 1.57, 3.54, 3.91, 2.06, 63.69], \"ry\": -0.75}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [578.22, 183.58, 638.11, 240.81], \"box_3d\": [1.56, 1.65, 3.69, -0.07, 1.9, 21.91], \"ry\": -1.56}]\n```" - } - ] - }, - { - "id": 186, - "data_source": "Mono3DRefer", - "scene_name": "000388", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.51, \"box_2d\": [527.71, 176.48, 549.82, 196.14], \"box_3d\": [1.66, 1.73, 3.05, -6.23, 2.0, 63.22], \"ry\": -1.61}]\n```" - } - ] - }, - { - "id": 187, - "data_source": "Mono3DRefer", - "scene_name": "000390", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.6, \"box_2d\": [478.55, 182.18, 520.33, 225.26], \"box_3d\": [1.68, 1.63, 3.9, -4.67, 2.09, 30.59], \"ry\": 1.45}, {\"category\": \"Car\", \"angle\": 1.45, \"box_2d\": [463.21, 185.28, 499.61, 210.78], \"box_3d\": [1.47, 1.7, 4.39, -8.06, 2.26, 44.83], \"ry\": 1.27}]\n```" - } - ] - }, - { - "id": 188, - "data_source": "Mono3DRefer", - "scene_name": "000391", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.86, 718.86), (cx, cy) = (607.19, 185.22). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.17, \"box_2d\": [210.96, 184.65, 348.23, 254.44], \"box_3d\": [1.48, 1.64, 3.82, -7.88, 1.49, 17.58], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [579.01, 176.54, 617.32, 211.48], \"box_3d\": [1.6, 1.76, 3.84, -0.52, 1.2, 35.2], \"ry\": -1.6}]\n```" - } - ] - }, - { - "id": 189, - "data_source": "Mono3DRefer", - "scene_name": "000397", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.28, \"box_2d\": [276.87, 183.57, 461.62, 243.88], \"box_3d\": [1.52, 1.71, 4.49, -6.52, 1.84, 19.76], \"ry\": -0.03}, {\"category\": \"Car\", \"angle\": 0.22, \"box_2d\": [388.39, 173.86, 510.32, 224.45], \"box_3d\": [1.95, 1.64, 4.57, -6.48, 2.02, 29.27], \"ry\": 0.0}, {\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [646.35, 182.83, 689.23, 218.42], \"box_3d\": [1.54, 1.67, 3.72, 2.61, 2.02, 33.67], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -2.28, \"box_2d\": [888.59, 220.02, 1241.0, 374.0], \"box_3d\": [1.35, 1.58, 4.05, 2.91, 1.7, 3.38], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [659.74, 174.77, 737.6, 237.05], \"box_3d\": [1.76, 1.71, 4.35, 2.65, 1.85, 22.78], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": -2.72, \"box_2d\": [248.74, 184.15, 423.42, 251.5], \"box_3d\": [1.52, 1.58, 3.61, -6.58, 1.82, 17.43], \"ry\": -3.08}]\n```" - } - ] - }, - { - "id": 190, - "data_source": "Mono3DRefer", - "scene_name": "000400", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [685.77, 178.12, 767.02, 235.21], \"box_3d\": [1.5, 1.62, 3.89, 3.27, 1.67, 21.18], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -2.15, \"box_2d\": [885.8, 160.44, 1241.0, 374.0], \"box_3d\": [1.69, 1.58, 3.95, 3.64, 1.65, 5.47], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -2.73, \"box_2d\": [928.61, 177.14, 1016.83, 209.77], \"box_3d\": [1.48, 1.36, 3.51, 17.33, 1.71, 34.63], \"ry\": -2.27}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [642.24, 178.5, 680.14, 208.68], \"box_3d\": [1.38, 1.49, 3.32, 2.41, 1.66, 34.98], \"ry\": -1.6}]\n```" - } - ] - }, - { - "id": 191, - "data_source": "Mono3DRefer", - "scene_name": "000404", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.86, 718.86), (cx, cy) = (607.19, 185.22). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [508.52, 182.69, 530.26, 200.4], \"box_3d\": [1.37, 1.54, 3.56, -7.23, 1.18, 58.51], \"ry\": -1.75}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [541.75, 178.07, 619.77, 243.03], \"box_3d\": [1.6, 1.76, 3.84, -0.88, 1.42, 19.9], \"ry\": -1.71}, {\"category\": \"Car\", \"angle\": -0.93, \"box_2d\": [0.0, 206.32, 92.31, 361.15], \"box_3d\": [1.36, 1.61, 3.76, -6.33, 1.61, 6.04], \"ry\": -1.71}, {\"category\": \"Car\", \"angle\": -1.2, \"box_2d\": [86.85, 193.98, 287.9, 295.95], \"box_3d\": [1.37, 1.59, 3.89, -6.96, 1.57, 12.31], \"ry\": -1.71}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [446.5, 180.26, 502.06, 231.81], \"box_3d\": [1.58, 1.68, 3.95, -4.53, 1.42, 24.3], \"ry\": -1.75}, {\"category\": \"Car\", \"angle\": -1.33, \"box_2d\": [250.73, 185.67, 353.31, 242.13], \"box_3d\": [1.42, 1.69, 4.53, -8.78, 1.45, 20.96], \"ry\": -1.73}, {\"category\": \"Car\", \"angle\": -1.2, \"box_2d\": [142.94, 186.08, 260.39, 238.86], \"box_3d\": [1.42, 1.7, 4.13, -12.31, 1.47, 22.05], \"ry\": -1.7}, {\"category\": \"Car\", \"angle\": -1.4, \"box_2d\": [349.56, 183.91, 401.96, 218.31], \"box_3d\": [1.42, 1.53, 4.2, -10.47, 1.38, 32.59], \"ry\": -1.7}, {\"category\": \"Car\", \"angle\": -1.43, \"box_2d\": [379.97, 183.36, 419.63, 210.15], \"box_3d\": [1.43, 1.62, 4.08, -11.91, 1.34, 41.31], \"ry\": -1.71}]\n```" - } - ] - }, - { - "id": 192, - "data_source": "Mono3DRefer", - "scene_name": "000408", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [468.04, 170.73, 546.7, 233.79], \"box_3d\": [1.7, 1.63, 4.08, -3.01, 1.67, 21.68], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": -1.8, \"box_2d\": [719.27, 175.5, 893.58, 317.32], \"box_3d\": [1.59, 1.59, 2.47, 2.36, 1.64, 9.39], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [692.1, 186.0, 792.66, 266.59], \"box_3d\": [1.37, 1.59, 3.22, 2.45, 1.65, 14.21], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": -1.43, \"box_2d\": [513.36, 178.43, 567.15, 215.41], \"box_3d\": [1.4, 1.67, 3.72, -2.82, 1.64, 29.51], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": 1.52, \"box_2d\": [635.66, 175.74, 663.8, 200.02], \"box_3d\": [1.49, 1.61, 3.74, 2.5, 1.69, 46.52], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -2.71, \"box_2d\": [147.2, 164.76, 390.58, 245.63], \"box_3d\": [1.59, 1.63, 4.43, -7.2, 1.46, 15.42], \"ry\": -3.14}]\n```" - } - ] - }, - { - "id": 193, - "data_source": "Mono3DRefer", - "scene_name": "000423", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [588.54, 175.38, 639.37, 216.74], \"box_3d\": [1.36, 1.69, 3.38, 0.06, 1.48, 25.82], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.22, \"box_2d\": [0.0, 177.11, 267.66, 342.47], \"box_3d\": [1.57, 1.64, 3.49, -5.83, 1.64, 8.56], \"ry\": 0.64}, {\"category\": \"Car\", \"angle\": 1.07, \"box_2d\": [258.28, 182.96, 411.21, 252.17], \"box_3d\": [1.38, 1.55, 3.89, -6.38, 1.63, 16.42], \"ry\": 0.7}, {\"category\": \"Car\", \"angle\": -1.41, \"box_2d\": [455.11, 187.46, 552.0, 256.65], \"box_3d\": [1.3, 1.55, 3.51, -2.27, 1.65, 15.88], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -1.43, \"box_2d\": [498.82, 180.98, 573.83, 235.69], \"box_3d\": [1.43, 1.62, 3.76, -2.09, 1.68, 20.99], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": 0.88, \"box_2d\": [393.2, 171.2, 503.41, 217.02], \"box_3d\": [1.57, 1.67, 4.2, -6.08, 1.55, 26.73], \"ry\": 0.65}, {\"category\": \"Car\", \"angle\": -1.94, \"box_2d\": [777.98, 193.67, 1108.47, 374.0], \"box_3d\": [1.38, 1.55, 3.6, 2.81, 1.63, 7.07], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [653.12, 172.87, 698.79, 209.4], \"box_3d\": [1.45, 1.64, 3.7, 2.72, 1.48, 30.8], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [286.95, 176.44, 352.29, 206.4], \"box_3d\": [1.43, 1.73, 4.09, -14.8, 1.63, 36.92], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [582.32, 173.28, 616.64, 204.6], \"box_3d\": [1.52, 1.67, 3.61, -0.54, 1.56, 37.13], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.17, \"box_2d\": [66.48, 179.17, 194.54, 229.71], \"box_3d\": [1.44, 1.52, 3.66, -14.9, 1.65, 22.6], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 194, - "data_source": "Mono3DRefer", - "scene_name": "000424", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.12, \"box_2d\": [0.0, 212.29, 104.2, 363.9], \"box_3d\": [1.46, 1.77, 4.59, -7.17, 1.97, 7.19], \"ry\": 1.35}, {\"category\": \"Car\", \"angle\": -1.15, \"box_2d\": [1163.91, 101.79, 1241.0, 348.72], \"box_3d\": [2.04, 1.66, 4.62, 7.43, 1.43, 6.44], \"ry\": -0.32}, {\"category\": \"Car\", \"angle\": -0.79, \"box_2d\": [808.1, 168.13, 1113.18, 277.74], \"box_3d\": [1.53, 1.61, 4.73, 5.69, 1.48, 11.77], \"ry\": -0.35}, {\"category\": \"Car\", \"angle\": -0.66, \"box_2d\": [714.47, 179.8, 886.14, 249.89], \"box_3d\": [1.4, 1.56, 3.48, 4.21, 1.57, 16.13], \"ry\": -0.41}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [443.39, 191.54, 525.24, 239.61], \"box_3d\": [1.27, 1.66, 4.06, -4.02, 1.89, 22.33], \"ry\": -1.96}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [346.03, 191.56, 399.53, 239.41], \"box_3d\": [1.48, 1.67, 4.5, -8.58, 2.19, 25.92], \"ry\": -1.9}]\n```" - } - ] - }, - { - "id": 195, - "data_source": "Mono3DRefer", - "scene_name": "000426", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [561.98, 171.7, 598.29, 199.34], \"box_3d\": [1.53, 1.59, 3.64, -1.82, 1.5, 42.27], \"ry\": -1.76}, {\"category\": \"Car\", \"angle\": 1.32, \"box_2d\": [163.01, 175.67, 252.5, 233.64], \"box_3d\": [1.46, 1.42, 3.12, -11.18, 1.56, 19.9], \"ry\": 0.81}, {\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [485.07, 168.49, 508.46, 190.68], \"box_3d\": [1.63, 1.6, 3.73, -8.67, 1.33, 55.32], \"ry\": 1.46}]\n```" - } - ] - }, - { - "id": 196, - "data_source": "Mono3DRefer", - "scene_name": "000431", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [252.46, 192.48, 460.47, 303.07], \"box_3d\": [1.37, 1.79, 4.16, -3.82, 1.73, 11.56], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 0.3, \"box_2d\": [358.54, 180.26, 478.34, 219.94], \"box_3d\": [1.56, 1.71, 4.5, -7.95, 1.89, 30.03], \"ry\": 0.05}]\n```" - } - ] - }, - { - "id": 197, - "data_source": "Mono3DRefer", - "scene_name": "000434", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [523.39, 181.74, 563.85, 210.43], \"box_3d\": [1.45, 1.77, 4.04, -3.58, 1.95, 39.17], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 198, - "data_source": "Mono3DRefer", - "scene_name": "000437", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [578.12, 174.63, 602.09, 193.62], \"box_3d\": [1.26, 1.6, 3.56, -1.41, 1.4, 50.19], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.53, \"box_2d\": [550.59, 172.0, 571.91, 189.09], \"box_3d\": [1.37, 1.63, 3.57, -4.08, 1.33, 60.48], \"ry\": -1.6}]\n```" - } - ] - }, - { - "id": 199, - "data_source": "Mono3DRefer", - "scene_name": "000439", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (707.05, 707.05), (cx, cy) = (604.08, 180.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [653.06, 176.52, 682.68, 201.9], \"box_3d\": [1.44, 1.58, 3.28, 3.74, 1.22, 42.4], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [409.81, 179.85, 461.2, 207.65], \"box_3d\": [1.49, 1.58, 3.84, -9.72, 1.48, 40.84], \"ry\": 1.71}, {\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [721.71, 172.95, 817.57, 256.37], \"box_3d\": [1.54, 1.57, 3.52, 3.31, 1.4, 14.88], \"ry\": -1.45}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [695.16, 175.52, 776.14, 235.1], \"box_3d\": [1.5, 1.63, 3.57, 3.53, 1.38, 19.82], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [677.99, 177.3, 731.55, 221.49], \"box_3d\": [1.47, 1.59, 3.64, 3.55, 1.37, 25.75], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [666.4, 177.18, 714.19, 214.23], \"box_3d\": [1.49, 1.74, 4.09, 3.66, 1.36, 30.98], \"ry\": -1.53}]\n```" - } - ] - }, - { - "id": 200, - "data_source": "Mono3DRefer", - "scene_name": "000442", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -3.13, \"box_2d\": [0.0, 159.03, 494.32, 360.76], \"box_3d\": [1.79, 1.69, 4.72, -3.88, 1.67, 8.06], \"ry\": 2.72}, {\"category\": \"Car\", \"angle\": -1.8, \"box_2d\": [1075.07, 174.07, 1241.0, 335.66], \"box_3d\": [1.57, 1.65, 3.35, 6.74, 1.59, 8.54], \"ry\": -1.15}, {\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [632.66, 179.46, 760.83, 250.58], \"box_3d\": [1.43, 1.7, 3.95, 2.1, 1.6, 16.8], \"ry\": 2.04}, {\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [1006.86, 174.37, 1093.85, 237.63], \"box_3d\": [1.51, 1.6, 3.24, 11.44, 1.56, 18.98], \"ry\": -1.13}, {\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [773.24, 171.55, 840.29, 215.18], \"box_3d\": [1.54, 1.68, 3.79, 7.55, 1.52, 27.6], \"ry\": 2.05}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [807.41, 169.93, 857.61, 205.23], \"box_3d\": [1.49, 1.52, 3.35, 9.98, 1.39, 32.34], \"ry\": 2.06}]\n```" - } - ] - }, - { - "id": 201, - "data_source": "Mono3DRefer", - "scene_name": "000444", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.59, \"box_2d\": [479.11, 183.1, 535.42, 233.56], \"box_3d\": [1.41, 1.59, 4.47, -3.25, 1.75, 22.93], \"ry\": 1.45}]\n```" - } - ] - }, - { - "id": 202, - "data_source": "Mono3DRefer", - "scene_name": "000445", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.93, \"box_2d\": [42.54, 190.42, 309.53, 334.07], \"box_3d\": [1.36, 1.57, 3.73, -5.43, 1.59, 8.76], \"ry\": -2.47}, {\"category\": \"Car\", \"angle\": 1.1, \"box_2d\": [277.41, 173.84, 415.74, 245.98], \"box_3d\": [1.48, 1.57, 3.48, -6.16, 1.53, 16.58], \"ry\": 0.75}, {\"category\": \"Car\", \"angle\": -1.38, \"box_2d\": [423.22, 186.1, 547.54, 261.21], \"box_3d\": [1.25, 1.74, 3.66, -2.36, 1.54, 14.27], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -1.48, \"box_2d\": [482.21, 178.1, 553.59, 237.84], \"box_3d\": [1.46, 1.59, 3.65, -2.46, 1.61, 19.64], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": 0.92, \"box_2d\": [406.03, 178.47, 495.72, 218.4], \"box_3d\": [1.37, 1.52, 3.3, -5.92, 1.6, 26.52], \"ry\": 0.71}, {\"category\": \"Car\", \"angle\": -1.83, \"box_2d\": [736.14, 178.4, 950.04, 311.52], \"box_3d\": [1.46, 1.72, 3.95, 2.88, 1.55, 9.97], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [696.96, 181.57, 821.53, 257.73], \"box_3d\": [1.32, 1.7, 4.04, 2.84, 1.51, 14.77], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [680.79, 168.11, 755.6, 227.63], \"box_3d\": [1.56, 1.48, 3.36, 2.99, 1.46, 20.77], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [642.92, 172.48, 678.59, 199.68], \"box_3d\": [1.38, 1.55, 3.6, 2.66, 1.39, 38.81], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [591.88, 171.81, 629.99, 206.58], \"box_3d\": [1.52, 1.67, 3.61, 0.03, 1.5, 33.62], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 2.08, \"box_2d\": [191.08, 174.86, 278.16, 211.39], \"box_3d\": [1.36, 1.54, 3.57, -15.0, 1.46, 28.95], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": -1.61, \"box_2d\": [589.09, 173.69, 649.29, 222.7], \"box_3d\": [1.36, 1.69, 3.38, 0.17, 1.42, 22.06], \"ry\": -1.6}]\n```" - } - ] - }, - { - "id": 203, - "data_source": "Mono3DRefer", - "scene_name": "000446", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [279.8, 187.2, 457.86, 290.93], \"box_3d\": [1.42, 1.75, 3.45, -3.83, 1.68, 11.93], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 204, - "data_source": "Mono3DRefer", - "scene_name": "000456", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [430.72, 183.25, 527.97, 265.85], \"box_3d\": [1.97, 1.87, 4.53, -3.48, 2.28, 19.82], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [529.14, 190.0, 561.33, 220.42], \"box_3d\": [1.7, 1.53, 3.82, -3.92, 2.78, 43.91], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 205, - "data_source": "Mono3DRefer", - "scene_name": "000458", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [568.92, 187.9, 601.65, 218.78], \"box_3d\": [1.43, 1.55, 3.91, -1.24, 2.22, 36.66], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 2.58, \"box_2d\": [833.36, 179.28, 1241.0, 320.04], \"box_3d\": [1.51, 1.65, 4.21, 5.08, 1.61, 8.89], \"ry\": 3.09}, {\"category\": \"Car\", \"angle\": -0.36, \"box_2d\": [799.64, 182.05, 1110.83, 284.44], \"box_3d\": [1.48, 1.49, 4.16, 5.22, 1.65, 11.39], \"ry\": 0.06}, {\"category\": \"Car\", \"angle\": -0.14, \"box_2d\": [725.97, 183.08, 909.52, 244.22], \"box_3d\": [1.51, 1.69, 4.42, 5.29, 1.8, 19.0], \"ry\": 0.12}, {\"category\": \"Car\", \"angle\": 2.82, \"box_2d\": [778.11, 181.42, 1009.84, 269.98], \"box_3d\": [1.56, 1.5, 3.8, 5.23, 1.74, 13.73], \"ry\": -3.11}]\n```" - } - ] - }, - { - "id": 206, - "data_source": "Mono3DRefer", - "scene_name": "000460", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (707.05, 707.05), (cx, cy) = (604.08, 180.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.49, \"box_2d\": [848.06, 166.46, 1169.62, 266.92], \"box_3d\": [1.31, 1.61, 3.78, 5.7, 1.14, 10.37], \"ry\": 0.01}]\n```" - } - ] - }, - { - "id": 207, - "data_source": "Mono3DRefer", - "scene_name": "000466", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.18, \"box_2d\": [0.0, 202.01, 78.4, 256.99], \"box_3d\": [1.42, 1.55, 3.45, -17.79, 2.35, 21.42], \"ry\": 1.5}, {\"category\": \"Car\", \"angle\": -1.13, \"box_2d\": [168.27, 197.83, 248.95, 232.61], \"box_3d\": [1.42, 1.7, 4.14, -18.84, 2.66, 34.01], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": -1.07, \"box_2d\": [103.56, 195.13, 191.97, 232.11], \"box_3d\": [1.52, 1.71, 4.11, -21.57, 2.62, 33.81], \"ry\": -1.64}]\n```" - } - ] - }, - { - "id": 208, - "data_source": "Mono3DRefer", - "scene_name": "000468", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [545.04, 171.56, 725.68, 345.64], \"box_3d\": [1.66, 1.73, 3.05, 0.19, 1.68, 8.49], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.14, \"box_2d\": [49.02, 188.06, 216.81, 258.63], \"box_3d\": [1.44, 1.59, 3.62, -11.17, 1.83, 17.13], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [357.74, 179.41, 416.33, 212.69], \"box_3d\": [1.6, 1.75, 4.0, -11.43, 1.95, 37.15], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [428.4, 177.97, 465.31, 201.19], \"box_3d\": [1.58, 1.63, 4.31, -11.65, 1.96, 51.67], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 209, - "data_source": "Mono3DRefer", - "scene_name": "000472", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.39, \"box_2d\": [415.93, 191.38, 542.17, 274.1], \"box_3d\": [1.33, 1.71, 4.09, -2.48, 1.74, 14.37], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [668.78, 173.34, 724.4, 214.95], \"box_3d\": [1.49, 1.56, 4.34, 3.28, 1.54, 28.3], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.44, \"box_2d\": [501.66, 182.97, 566.24, 232.28], \"box_3d\": [1.42, 1.53, 4.12, -2.41, 1.77, 23.34], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [705.32, 175.05, 811.89, 257.62], \"box_3d\": [1.59, 1.59, 3.89, 3.11, 1.67, 16.02], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [650.77, 175.51, 678.32, 198.63], \"box_3d\": [1.38, 1.64, 3.51, 3.41, 1.56, 45.16], \"ry\": -1.48}]\n```" - } - ] - }, - { - "id": 210, - "data_source": "Mono3DRefer", - "scene_name": "000473", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [589.61, 176.67, 629.93, 204.43], \"box_3d\": [1.5, 1.7, 3.77, 0.01, 1.74, 41.31], \"ry\": 1.74}]\n```" - } - ] - }, - { - "id": 211, - "data_source": "Mono3DRefer", - "scene_name": "000474", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.42, \"box_2d\": [0.0, 191.88, 126.19, 334.21], \"box_3d\": [1.47, 1.74, 4.26, -8.4, 1.73, 8.36], \"ry\": 3.09}, {\"category\": \"Car\", \"angle\": 0.23, \"box_2d\": [476.61, 182.36, 540.64, 201.35], \"box_3d\": [1.36, 1.68, 4.49, -7.62, 2.1, 54.26], \"ry\": 0.09}]\n```" - } - ] - }, - { - "id": 212, - "data_source": "Mono3DRefer", - "scene_name": "000475", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.36, \"box_2d\": [0.0, 219.52, 124.24, 374.0], \"box_3d\": [1.48, 1.5, 3.58, -5.71, 1.94, 5.48], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.17, \"box_2d\": [778.02, 179.39, 1039.21, 326.67], \"box_3d\": [1.46, 1.61, 3.61, 3.39, 1.55, 9.05], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": 2.02, \"box_2d\": [236.28, 189.45, 408.55, 278.63], \"box_3d\": [1.48, 1.56, 4.09, -5.61, 1.85, 14.55], \"ry\": 1.65}, {\"category\": \"Car\", \"angle\": -1.84, \"box_2d\": [712.17, 178.42, 848.83, 261.79], \"box_3d\": [1.46, 1.67, 3.89, 3.27, 1.59, 14.79], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [440.52, 177.48, 499.12, 218.69], \"box_3d\": [1.56, 1.63, 3.62, -5.68, 1.77, 29.44], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [672.89, 174.42, 744.46, 225.55], \"box_3d\": [1.51, 1.6, 3.54, 3.07, 1.58, 23.33], \"ry\": -1.64}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [663.39, 174.1, 718.81, 215.12], \"box_3d\": [1.53, 1.5, 3.34, 3.16, 1.6, 28.92], \"ry\": -1.68}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [471.97, 178.16, 513.17, 210.58], \"box_3d\": [1.46, 1.43, 3.52, -5.6, 1.72, 34.51], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 213, - "data_source": "Mono3DRefer", - "scene_name": "000476", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [523.34, 179.86, 576.42, 215.46], \"box_3d\": [1.5, 1.7, 3.77, -2.69, 1.83, 32.66], \"ry\": 1.68}]\n```" - } - ] - }, - { - "id": 214, - "data_source": "Mono3DRefer", - "scene_name": "000479", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [589.93, 174.78, 612.54, 196.55], \"box_3d\": [1.66, 1.73, 3.05, -0.7, 1.84, 57.25], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.11, \"box_2d\": [137.13, 189.29, 249.89, 241.12], \"box_3d\": [1.6, 1.77, 3.58, -14.29, 2.2, 24.92], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [364.97, 181.32, 408.12, 204.97], \"box_3d\": [1.52, 1.76, 3.66, -15.14, 2.11, 48.98], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 215, - "data_source": "Mono3DRefer", - "scene_name": "000480", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.06, \"box_2d\": [196.1, 190.3, 289.39, 235.77], \"box_3d\": [1.6, 1.77, 3.58, -14.3, 2.32, 28.23], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [590.44, 177.5, 613.13, 199.35], \"box_3d\": [1.66, 1.73, 3.05, -0.65, 2.05, 57.03], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 216, - "data_source": "Mono3DRefer", - "scene_name": "000481", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.08, \"box_2d\": [148.13, 189.8, 257.18, 236.2], \"box_3d\": [1.56, 1.7, 4.35, -15.47, 2.25, 27.62], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [343.11, 182.31, 397.82, 212.66], \"box_3d\": [1.6, 1.79, 3.91, -13.47, 2.15, 40.7], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 217, - "data_source": "Mono3DRefer", - "scene_name": "000482", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.83, \"box_2d\": [741.97, 183.19, 861.27, 265.09], \"box_3d\": [1.42, 1.45, 3.2, 3.63, 1.64, 14.33], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -2.07, \"box_2d\": [930.97, 181.98, 1104.68, 267.35], \"box_3d\": [1.43, 1.48, 3.25, 7.64, 1.62, 13.92], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 218, - "data_source": "Mono3DRefer", - "scene_name": "000483", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.31, \"box_2d\": [0.0, 188.34, 180.35, 254.08], \"box_3d\": [1.45, 1.5, 4.53, -13.6, 1.89, 18.88], \"ry\": 1.7}, {\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [618.28, 175.23, 687.01, 239.9], \"box_3d\": [1.56, 1.65, 3.69, 1.13, 1.64, 19.37], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": -1.17, \"box_2d\": [819.32, 173.48, 1055.85, 300.28], \"box_3d\": [1.39, 1.61, 3.65, 4.5, 1.41, 9.76], \"ry\": -0.75}, {\"category\": \"Car\", \"angle\": 2.14, \"box_2d\": [23.58, 191.67, 335.55, 316.88], \"box_3d\": [1.45, 1.84, 4.59, -6.2, 1.78, 11.11], \"ry\": 1.64}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [521.48, 171.38, 571.92, 208.69], \"box_3d\": [1.76, 1.58, 3.68, -3.13, 1.71, 36.05], \"ry\": 1.75}, {\"category\": \"Car\", \"angle\": 2.16, \"box_2d\": [186.3, 183.78, 335.75, 248.62], \"box_3d\": [1.56, 1.62, 4.19, -9.46, 1.89, 19.87], \"ry\": 1.72}, {\"category\": \"Car\", \"angle\": 2.0, \"box_2d\": [411.36, 182.32, 488.84, 216.78], \"box_3d\": [1.42, 1.67, 4.53, -7.15, 1.87, 32.6], \"ry\": 1.78}, {\"category\": \"Car\", \"angle\": -1.19, \"box_2d\": [673.08, 170.65, 719.89, 196.26], \"box_3d\": [1.54, 1.63, 3.74, 5.46, 1.43, 45.4], \"ry\": -1.07}]\n```" - } - ] - }, - { - "id": 219, - "data_source": "Mono3DRefer", - "scene_name": "000484", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [598.39, 180.1, 621.78, 202.36], \"box_3d\": [1.54, 1.63, 3.67, -0.03, 2.08, 52.48], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -2.02, \"box_2d\": [799.97, 160.37, 1241.0, 374.0], \"box_3d\": [1.7, 1.62, 3.76, 2.77, 1.66, 5.6], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [684.17, 187.11, 773.82, 255.01], \"box_3d\": [1.4, 1.53, 3.56, 2.68, 1.77, 17.19], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [671.57, 181.3, 743.12, 235.02], \"box_3d\": [1.49, 1.6, 3.81, 2.87, 1.76, 22.25], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [658.55, 180.71, 700.88, 216.04], \"box_3d\": [1.48, 1.47, 3.7, 3.05, 1.84, 32.46], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [646.88, 178.98, 686.81, 212.98], \"box_3d\": [1.64, 1.64, 3.89, 2.84, 1.97, 37.09], \"ry\": -1.59}]\n```" - } - ] - }, - { - "id": 220, - "data_source": "Mono3DRefer", - "scene_name": "000489", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.06, \"box_2d\": [191.64, 193.69, 289.86, 234.81], \"box_3d\": [1.46, 1.86, 4.01, -14.84, 2.35, 29.18], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 221, - "data_source": "Mono3DRefer", - "scene_name": "000491", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [606.74, 175.5, 647.57, 216.5], \"box_3d\": [1.57, 1.56, 2.62, 0.67, 1.7, 29.25], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [360.03, 188.7, 494.13, 267.15], \"box_3d\": [1.43, 1.68, 4.19, -3.86, 1.81, 15.78], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [452.67, 181.72, 523.33, 234.02], \"box_3d\": [1.46, 1.59, 4.27, -3.76, 1.75, 22.62], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [503.1, 178.25, 549.42, 209.29], \"box_3d\": [1.38, 1.62, 4.06, -3.97, 1.65, 34.48], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [536.83, 171.21, 566.47, 193.9], \"box_3d\": [1.49, 1.53, 3.88, -4.02, 1.41, 49.86], \"ry\": 1.62}]\n```" - } - ] - }, - { - "id": 222, - "data_source": "Mono3DRefer", - "scene_name": "000494", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [569.33, 180.88, 613.91, 225.02], \"box_3d\": [1.65, 1.67, 3.64, -0.78, 1.98, 29.11], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 2.2, \"box_2d\": [0.0, 196.14, 197.44, 318.81], \"box_3d\": [1.4, 1.68, 4.45, -8.35, 1.81, 10.84], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [277.8, 187.26, 381.92, 243.5], \"box_3d\": [1.45, 1.62, 3.77, -8.14, 1.9, 21.21], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [372.63, 185.0, 436.72, 225.33], \"box_3d\": [1.53, 1.46, 4.18, -8.55, 2.06, 30.24], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [428.21, 183.03, 466.8, 213.09], \"box_3d\": [1.59, 1.57, 3.39, -9.09, 2.17, 40.46], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [456.59, 185.8, 492.19, 207.04], \"box_3d\": [1.35, 1.63, 4.1, -9.27, 2.26, 49.47], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 223, - "data_source": "Mono3DRefer", - "scene_name": "000497", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.36, \"box_2d\": [967.57, 190.81, 1241.0, 374.0], \"box_3d\": [1.53, 1.65, 4.37, 3.53, 1.65, 3.27], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [167.9, 162.91, 421.8, 327.79], \"box_3d\": [1.74, 1.65, 4.18, -3.96, 1.67, 9.79], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [432.08, 171.21, 549.68, 251.14], \"box_3d\": [1.64, 1.45, 3.48, -2.66, 1.62, 16.62], \"ry\": 1.79}, {\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [782.31, 172.8, 979.41, 307.01], \"box_3d\": [1.61, 1.68, 4.04, 3.69, 1.63, 10.71], \"ry\": -1.49}, {\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [757.95, 174.3, 850.33, 254.04], \"box_3d\": [1.52, 1.58, 4.28, 4.13, 1.58, 16.03], \"ry\": -1.4}, {\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [623.7, 171.8, 703.33, 216.31], \"box_3d\": [1.54, 1.64, 4.08, 2.07, 1.53, 27.2], \"ry\": 2.01}]\n```" - } - ] - }, - { - "id": 224, - "data_source": "Mono3DRefer", - "scene_name": "000505", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (707.05, 707.05), (cx, cy) = (604.08, 180.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.25, \"box_2d\": [715.44, 167.9, 860.58, 213.12], \"box_3d\": [1.41, 1.63, 4.42, 6.03, 1.02, 23.69], \"ry\": -0.0}, {\"category\": \"Car\", \"angle\": -0.83, \"box_2d\": [1057.74, 172.67, 1223.0, 369.0], \"box_3d\": [1.31, 1.61, 3.78, 5.61, 1.25, 5.07], \"ry\": -0.02}]\n```" - } - ] - }, - { - "id": 225, - "data_source": "Mono3DRefer", - "scene_name": "000507", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.11, \"box_2d\": [0.0, 204.34, 451.37, 374.0], \"box_3d\": [1.56, 1.56, 3.6, -2.4, 1.83, 4.69], \"ry\": 1.66}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [251.4, 193.39, 432.89, 340.06], \"box_3d\": [1.7, 1.43, 3.2, -3.67, 2.03, 10.31], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.51, \"box_2d\": [498.13, 182.5, 522.36, 205.1], \"box_3d\": [1.68, 1.63, 3.9, -7.85, 2.45, 56.46], \"ry\": 1.37}]\n```" - } - ] - }, - { - "id": 226, - "data_source": "Mono3DRefer", - "scene_name": "000508", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [592.02, 175.19, 626.32, 208.22], \"box_3d\": [1.66, 1.73, 3.05, -0.07, 1.81, 38.24], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 227, - "data_source": "Mono3DRefer", - "scene_name": "000511", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.18, \"box_2d\": [0.0, 186.41, 80.43, 227.94], \"box_3d\": [1.48, 1.44, 3.59, -22.59, 2.03, 28.06], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 2.11, \"box_2d\": [74.93, 186.07, 158.59, 219.56], \"box_3d\": [1.42, 1.68, 3.38, -22.71, 2.05, 33.31], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": 2.11, \"box_2d\": [141.58, 184.83, 215.08, 212.93], \"box_3d\": [1.38, 1.52, 4.0, -22.93, 2.04, 38.43], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [205.21, 179.74, 256.12, 206.75], \"box_3d\": [1.57, 1.47, 3.35, -23.19, 2.01, 44.16], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.97, \"box_2d\": [242.21, 182.35, 285.78, 204.55], \"box_3d\": [1.43, 1.48, 3.33, -23.43, 2.09, 48.92], \"ry\": 1.52}]\n```" - } - ] - }, - { - "id": 228, - "data_source": "Mono3DRefer", - "scene_name": "000512", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [353.8, 184.03, 410.21, 216.95], \"box_3d\": [1.84, 2.04, 4.33, -13.69, 2.54, 43.52], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 229, - "data_source": "Mono3DRefer", - "scene_name": "000514", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [704.92, 175.38, 733.38, 203.2], \"box_3d\": [1.55, 1.57, 3.64, 6.41, 1.72, 42.44], \"ry\": -1.41}]\n```" - } - ] - }, - { - "id": 230, - "data_source": "Mono3DRefer", - "scene_name": "000523", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.88, \"box_2d\": [308.91, 183.49, 444.19, 227.96], \"box_3d\": [1.47, 1.74, 4.26, -8.25, 1.87, 25.65], \"ry\": 3.09}]\n```" - } - ] - }, - { - "id": 231, - "data_source": "Mono3DRefer", - "scene_name": "000526", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [546.05, 173.72, 575.08, 192.97], \"box_3d\": [1.42, 1.78, 3.62, -3.1, 0.84, 55.55], \"ry\": 1.64}]\n```" - } - ] - }, - { - "id": 232, - "data_source": "Mono3DRefer", - "scene_name": "000528", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.85, \"box_2d\": [921.22, 188.3, 1241.0, 374.0], \"box_3d\": [1.53, 1.62, 3.77, 2.99, 1.63, 3.21], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [478.15, 177.88, 559.52, 233.29], \"box_3d\": [1.76, 1.77, 4.41, -3.1, 1.95, 25.19], \"ry\": 1.7}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [543.72, 185.08, 602.33, 221.16], \"box_3d\": [1.48, 1.65, 4.4, -1.62, 2.05, 32.55], \"ry\": 1.75}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [626.07, 184.14, 658.56, 206.53], \"box_3d\": [1.42, 1.66, 3.64, 2.19, 2.2, 48.61], \"ry\": 1.76}]\n```" - } - ] - }, - { - "id": 233, - "data_source": "Mono3DRefer", - "scene_name": "000529", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.57, \"box_2d\": [0.0, 192.78, 257.63, 288.66], \"box_3d\": [1.49, 1.64, 3.7, -8.43, 1.87, 12.55], \"ry\": 3.13}, {\"category\": \"Car\", \"angle\": -2.68, \"box_2d\": [103.74, 190.33, 334.71, 269.95], \"box_3d\": [1.52, 1.55, 4.02, -8.16, 1.92, 15.21], \"ry\": 3.12}, {\"category\": \"Car\", \"angle\": -2.83, \"box_2d\": [325.05, 173.77, 454.53, 227.77], \"box_3d\": [1.9, 1.78, 4.25, -8.15, 1.96, 26.83], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": 0.16, \"box_2d\": [455.98, 181.54, 519.21, 206.87], \"box_3d\": [1.52, 1.64, 3.7, -7.7, 2.09, 45.42], \"ry\": -0.01}]\n```" - } - ] - }, - { - "id": 234, - "data_source": "Mono3DRefer", - "scene_name": "000530", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (707.05, 707.05), (cx, cy) = (604.08, 180.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [536.71, 161.8, 580.32, 206.77], \"box_3d\": [1.91, 1.65, 4.4, -2.21, 1.11, 32.51], \"ry\": -1.71}]\n```" - } - ] - }, - { - "id": 235, - "data_source": "Mono3DRefer", - "scene_name": "000531", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.01, \"box_2d\": [938.56, 163.43, 1109.63, 242.89], \"box_3d\": [1.64, 1.73, 4.43, 9.59, 1.48, 17.24], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": -2.13, \"box_2d\": [1047.29, 174.4, 1241.0, 243.93], \"box_3d\": [1.38, 1.8, 4.49, 12.24, 1.44, 16.7], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": -0.81, \"box_2d\": [0.0, 192.28, 112.47, 260.5], \"box_3d\": [1.53, 1.61, 4.73, -15.0, 2.06, 17.95], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": -1.18, \"box_2d\": [253.94, 187.21, 409.48, 263.18], \"box_3d\": [1.48, 1.71, 4.65, -6.32, 1.85, 16.91], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": -0.98, \"box_2d\": [111.11, 165.36, 302.2, 263.49], \"box_3d\": [2.04, 1.66, 4.62, -9.46, 1.91, 17.37], \"ry\": -1.47}, {\"category\": \"Car\", \"angle\": -1.43, \"box_2d\": [502.21, 173.05, 570.88, 221.75], \"box_3d\": [1.42, 1.64, 3.95, -2.32, 1.46, 23.27], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": -1.93, \"box_2d\": [851.31, 173.45, 1004.61, 248.3], \"box_3d\": [1.51, 1.72, 4.5, 7.18, 1.55, 16.93], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [675.33, 164.04, 716.83, 202.85], \"box_3d\": [1.7, 1.8, 4.5, 3.98, 1.34, 34.08], \"ry\": -1.47}]\n```" - } - ] - }, - { - "id": 236, - "data_source": "Mono3DRefer", - "scene_name": "000537", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [351.98, 189.16, 399.44, 215.15], \"box_3d\": [1.56, 1.76, 4.18, -15.32, 2.66, 47.31], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 237, - "data_source": "Mono3DRefer", - "scene_name": "000539", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.28, \"box_2d\": [0.0, 192.86, 44.18, 237.7], \"box_3d\": [1.43, 1.81, 4.16, -20.83, 2.12, 23.3], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 2.14, \"box_2d\": [58.33, 187.39, 164.91, 225.94], \"box_3d\": [1.47, 1.67, 4.31, -21.13, 2.12, 30.77], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 2.07, \"box_2d\": [153.65, 185.68, 222.96, 217.59], \"box_3d\": [1.48, 1.44, 3.59, -21.1, 2.14, 36.2], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 2.06, \"box_2d\": [213.83, 183.87, 273.67, 210.15], \"box_3d\": [1.42, 1.68, 3.38, -21.11, 2.08, 41.67], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 2.06, \"box_2d\": [254.58, 184.16, 309.72, 207.04], \"box_3d\": [1.38, 1.52, 4.0, -21.14, 2.13, 46.63], \"ry\": 1.64}, {\"category\": \"Car\", \"angle\": 1.99, \"box_2d\": [299.62, 180.19, 339.87, 202.81], \"box_3d\": [1.57, 1.47, 3.35, -21.05, 2.12, 52.38], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [325.83, 180.98, 360.01, 199.79], \"box_3d\": [1.43, 1.48, 3.33, -21.16, 2.08, 57.21], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 238, - "data_source": "Mono3DRefer", - "scene_name": "000541", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.12, \"box_2d\": [0.0, 208.67, 404.83, 374.0], \"box_3d\": [1.33, 1.71, 4.09, -2.95, 1.69, 5.36], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [660.91, 182.66, 746.6, 246.39], \"box_3d\": [1.49, 1.56, 4.34, 2.35, 1.77, 19.39], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.39, \"box_2d\": [380.4, 190.12, 500.34, 278.66], \"box_3d\": [1.42, 1.53, 4.12, -3.23, 1.8, 14.27], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.92, \"box_2d\": [746.9, 177.32, 1094.39, 374.0], \"box_3d\": [1.59, 1.59, 3.89, 2.5, 1.66, 6.98], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [629.7, 183.36, 664.27, 212.76], \"box_3d\": [1.38, 1.64, 3.51, 1.86, 1.92, 36.28], \"ry\": -1.5}]\n```" - } - ] - }, - { - "id": 239, - "data_source": "Mono3DRefer", - "scene_name": "000543", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [352.2, 188.74, 457.77, 256.25], \"box_3d\": [1.34, 1.64, 4.57, -4.83, 1.76, 17.43], \"ry\": 1.48}, {\"category\": \"Car\", \"angle\": -2.87, \"box_2d\": [0.0, 192.69, 187.99, 319.87], \"box_3d\": [1.52, 1.41, 4.22, -8.1, 1.83, 9.46], \"ry\": 2.72}]\n```" - } - ] - }, - { - "id": 240, - "data_source": "Mono3DRefer", - "scene_name": "000545", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.99, \"box_2d\": [1010.68, 171.8, 1241.0, 251.18], \"box_3d\": [1.54, 1.68, 4.16, 11.0, 1.55, 15.49], \"ry\": -2.68}, {\"category\": \"Car\", \"angle\": -0.8, \"box_2d\": [684.37, 195.55, 778.19, 231.22], \"box_3d\": [1.37, 1.56, 4.19, 5.33, 2.42, 31.65], \"ry\": -0.63}]\n```" - } - ] - }, - { - "id": 241, - "data_source": "Mono3DRefer", - "scene_name": "000546", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.5, \"box_2d\": [1053.94, 193.9, 1218.11, 320.55], \"box_3d\": [1.4, 1.64, 3.73, 7.63, 1.75, 10.43], \"ry\": -0.88}]\n```" - } - ] - }, - { - "id": 242, - "data_source": "Mono3DRefer", - "scene_name": "000547", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [389.08, 188.81, 423.13, 207.42], \"box_3d\": [1.48, 1.79, 4.15, -17.54, 2.89, 62.18], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 243, - "data_source": "Mono3DRefer", - "scene_name": "000552", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.82, \"box_2d\": [720.95, 166.2, 862.05, 268.42], \"box_3d\": [1.64, 1.69, 3.3, 3.16, 1.56, 13.37], \"ry\": -1.6}]\n```" - } - ] - }, - { - "id": 244, - "data_source": "Mono3DRefer", - "scene_name": "000555", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.51, \"box_2d\": [539.85, 174.69, 562.54, 194.92], \"box_3d\": [1.66, 1.73, 3.05, -5.0, 1.84, 61.47], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 2.08, \"box_2d\": [98.97, 197.45, 195.16, 240.56], \"box_3d\": [1.44, 1.55, 3.53, -17.58, 2.43, 27.53], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 2.11, \"box_2d\": [218.58, 191.58, 273.49, 213.87], \"box_3d\": [1.52, 1.76, 4.2, -27.22, 2.97, 54.06], \"ry\": 1.65}]\n```" - } - ] - }, - { - "id": 245, - "data_source": "Mono3DRefer", - "scene_name": "000556", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [522.6, 183.29, 565.12, 209.71], \"box_3d\": [1.37, 1.79, 4.16, -3.67, 1.97, 40.38], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 0.23, \"box_2d\": [420.43, 185.83, 504.81, 214.37], \"box_3d\": [1.56, 1.71, 4.5, -8.52, 2.34, 41.79], \"ry\": 0.03}]\n```" - } - ] - }, - { - "id": 246, - "data_source": "Mono3DRefer", - "scene_name": "000557", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.92, \"box_2d\": [0.0, 200.69, 362.1, 373.0], \"box_3d\": [1.47, 1.59, 4.03, -2.56, 1.62, 3.14], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.36, \"box_2d\": [372.13, 182.28, 505.63, 277.16], \"box_3d\": [1.41, 1.54, 3.36, -2.72, 1.44, 12.55], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [588.01, 170.74, 644.1, 221.89], \"box_3d\": [1.52, 1.67, 4.38, 0.44, 1.2, 23.83], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.47, \"box_2d\": [518.66, 171.36, 572.24, 217.14], \"box_3d\": [1.56, 1.59, 3.65, -2.02, 1.22, 26.59], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [482.96, 174.18, 524.46, 201.4], \"box_3d\": [1.42, 1.56, 4.16, -5.38, 1.04, 40.01], \"ry\": 1.61}]\n```" - } - ] - }, - { - "id": 247, - "data_source": "Mono3DRefer", - "scene_name": "000559", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [488.97, 169.38, 533.65, 209.69], \"box_3d\": [2.11, 2.04, 4.62, -4.97, 1.47, 40.31], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 248, - "data_source": "Mono3DRefer", - "scene_name": "000560", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [576.31, 176.63, 661.89, 259.06], \"box_3d\": [1.66, 1.73, 3.05, 0.14, 1.77, 16.24], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 249, - "data_source": "Mono3DRefer", - "scene_name": "000561", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.01, \"box_2d\": [764.16, 184.92, 1027.68, 332.53], \"box_3d\": [1.42, 1.58, 3.48, 3.19, 1.59, 8.91], \"ry\": -1.67}]\n```" - } - ] - }, - { - "id": 250, - "data_source": "Mono3DRefer", - "scene_name": "000563", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.57, \"box_2d\": [592.59, 187.57, 612.49, 205.87], \"box_3d\": [1.41, 1.59, 4.47, -0.65, 2.68, 60.4], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 251, - "data_source": "Mono3DRefer", - "scene_name": "000570", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.57, \"box_2d\": [597.61, 184.36, 616.69, 201.68], \"box_3d\": [1.41, 1.59, 4.47, -0.27, 2.44, 62.85], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 252, - "data_source": "Mono3DRefer", - "scene_name": "000572", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.86, 718.86), (cx, cy) = (607.19, 185.22). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.48, \"box_2d\": [507.19, 173.28, 542.38, 202.86], \"box_3d\": [1.78, 1.81, 4.32, -5.25, 1.05, 45.78], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [572.73, 173.45, 613.41, 210.54], \"box_3d\": [1.6, 1.76, 3.84, -0.7, 1.08, 33.27], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -0.82, \"box_2d\": [0.0, 196.0, 108.23, 312.26], \"box_3d\": [1.4, 1.66, 4.07, -8.42, 1.58, 8.8], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.17, \"box_2d\": [170.8, 193.97, 398.5, 315.06], \"box_3d\": [1.35, 1.57, 3.72, -4.4, 1.51, 10.34], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.13, \"box_2d\": [187.86, 184.25, 320.69, 248.8], \"box_3d\": [1.44, 1.58, 3.66, -8.87, 1.44, 18.3], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.3, \"box_2d\": [343.09, 180.82, 452.57, 247.16], \"box_3d\": [1.49, 1.64, 3.95, -5.24, 1.39, 18.34], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 253, - "data_source": "Mono3DRefer", - "scene_name": "000574", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.26, \"box_2d\": [439.27, 183.67, 517.5, 206.76], \"box_3d\": [1.36, 1.68, 4.49, -8.17, 2.06, 44.85], \"ry\": 0.08}]\n```" - } - ] - }, - { - "id": 254, - "data_source": "Mono3DRefer", - "scene_name": "000578", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [563.17, 188.0, 699.41, 275.26], \"box_3d\": [1.48, 1.51, 4.35, 0.57, 1.83, 15.0], \"ry\": 1.92}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [775.08, 182.7, 843.45, 236.48], \"box_3d\": [1.47, 1.68, 3.88, 6.16, 1.79, 22.17], \"ry\": 1.93}, {\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [685.42, 187.7, 821.78, 280.68], \"box_3d\": [1.46, 1.74, 3.99, 2.86, 1.78, 13.9], \"ry\": 1.98}, {\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [649.62, 186.08, 748.22, 245.27], \"box_3d\": [1.42, 1.68, 4.29, 2.56, 1.82, 20.18], \"ry\": 1.95}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [729.31, 184.68, 784.82, 223.82], \"box_3d\": [1.36, 1.4, 3.8, 5.67, 1.84, 27.72], \"ry\": 1.96}, {\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [786.59, 180.42, 820.3, 208.08], \"box_3d\": [1.44, 1.65, 2.96, 10.6, 1.87, 39.55], \"ry\": 1.87}]\n```" - } - ] - }, - { - "id": 255, - "data_source": "Mono3DRefer", - "scene_name": "000580", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.3, \"box_2d\": [394.0, 183.47, 491.45, 211.96], \"box_3d\": [1.36, 1.68, 4.49, -8.44, 1.92, 36.47], \"ry\": 0.07}]\n```" - } - ] - }, - { - "id": 256, - "data_source": "Mono3DRefer", - "scene_name": "000581", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [551.47, 204.63, 587.49, 240.3], \"box_3d\": [1.5, 1.62, 4.19, -1.98, 3.15, 35.69], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 257, - "data_source": "Mono3DRefer", - "scene_name": "000583", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [575.79, 175.7, 616.78, 215.19], \"box_3d\": [1.66, 1.73, 3.05, -0.64, 1.81, 32.24], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [324.38, 189.89, 385.19, 221.21], \"box_3d\": [1.47, 1.71, 4.36, -13.28, 2.4, 37.72], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [409.88, 183.96, 445.68, 204.87], \"box_3d\": [1.52, 1.71, 3.89, -14.01, 2.39, 55.59], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 258, - "data_source": "Mono3DRefer", - "scene_name": "000584", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [440.25, 184.17, 503.63, 240.87], \"box_3d\": [1.41, 1.59, 4.47, -3.91, 1.76, 20.75], \"ry\": 1.43}]\n```" - } - ] - }, - { - "id": 259, - "data_source": "Mono3DRefer", - "scene_name": "000587", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [523.46, 168.78, 562.19, 206.68], \"box_3d\": [1.97, 1.87, 4.53, -3.7, 1.79, 40.17], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [555.82, 172.97, 576.55, 192.86], \"box_3d\": [1.7, 1.53, 3.82, -3.89, 1.74, 64.28], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 260, - "data_source": "Mono3DRefer", - "scene_name": "000593", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [568.97, 178.94, 613.85, 208.18], \"box_3d\": [1.44, 1.68, 4.39, -0.94, 1.77, 37.95], \"ry\": 1.7}]\n```" - } - ] - }, - { - "id": 261, - "data_source": "Mono3DRefer", - "scene_name": "000594", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [580.21, 175.92, 637.28, 229.88], \"box_3d\": [1.56, 1.65, 3.69, -0.06, 1.68, 22.91], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [367.47, 184.87, 423.69, 212.29], \"box_3d\": [1.29, 1.57, 4.38, -11.03, 1.93, 37.29], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [459.14, 180.97, 487.82, 198.91], \"box_3d\": [1.4, 1.54, 3.67, -11.13, 2.07, 58.93], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": -0.82, \"box_2d\": [636.22, 171.1, 678.9, 189.77], \"box_3d\": [1.53, 1.57, 3.54, 4.01, 1.39, 60.75], \"ry\": -0.75}]\n```" - } - ] - }, - { - "id": 262, - "data_source": "Mono3DRefer", - "scene_name": "000597", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.0, \"box_2d\": [955.32, 144.92, 1241.0, 363.72], \"box_3d\": [1.5, 1.71, 4.44, 5.21, 1.32, 7.11], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 2.03, \"box_2d\": [97.36, 189.13, 416.38, 374.0], \"box_3d\": [1.54, 1.63, 3.59, -3.56, 1.76, 7.96], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [383.91, 188.06, 507.57, 273.48], \"box_3d\": [1.42, 1.46, 3.63, -3.13, 1.75, 14.23], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [451.86, 182.45, 541.32, 244.31], \"box_3d\": [1.43, 1.66, 4.19, -2.92, 1.7, 19.16], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [690.88, 172.77, 766.05, 241.37], \"box_3d\": [1.58, 1.71, 3.75, 3.06, 1.6, 18.62], \"ry\": -1.38}, {\"category\": \"Car\", \"angle\": -0.47, \"box_2d\": [886.14, 158.67, 1225.88, 263.82], \"box_3d\": [1.49, 1.57, 4.26, 6.71, 1.32, 11.26], \"ry\": 0.06}, {\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [501.53, 175.36, 552.99, 220.31], \"box_3d\": [1.7, 1.5, 3.34, -3.32, 1.82, 29.12], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": 1.29, \"box_2d\": [507.1, 175.74, 565.95, 209.36], \"box_3d\": [1.48, 1.71, 3.93, -3.57, 1.64, 34.14], \"ry\": 1.18}, {\"category\": \"Car\", \"angle\": -0.07, \"box_2d\": [382.77, 173.82, 445.23, 195.13], \"box_3d\": [1.6, 1.75, 4.63, -15.48, 1.7, 56.9], \"ry\": -0.34}, {\"category\": \"Car\", \"angle\": -1.29, \"box_2d\": [708.47, 175.19, 786.58, 226.53], \"box_3d\": [1.51, 1.58, 3.31, 4.41, 1.6, 23.02], \"ry\": -1.1}, {\"category\": \"Car\", \"angle\": -1.06, \"box_2d\": [750.52, 168.44, 831.91, 214.71], \"box_3d\": [1.61, 1.6, 3.11, 6.73, 1.47, 26.78], \"ry\": -0.82}, {\"category\": \"Car\", \"angle\": 2.36, \"box_2d\": [898.04, 161.36, 976.35, 187.78], \"box_3d\": [1.43, 1.7, 3.95, 18.38, 0.81, 40.6], \"ry\": 2.78}]\n```" - } - ] - }, - { - "id": 263, - "data_source": "Mono3DRefer", - "scene_name": "000599", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [604.37, 175.85, 631.18, 201.35], \"box_3d\": [1.54, 1.63, 3.67, 0.47, 1.76, 46.03], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [696.83, 177.84, 797.9, 251.74], \"box_3d\": [1.49, 1.57, 4.07, 3.0, 1.62, 16.7], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.99, \"box_2d\": [804.14, 175.4, 1231.47, 374.0], \"box_3d\": [1.57, 1.59, 3.75, 2.96, 1.62, 6.29], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 0.45, \"box_2d\": [137.06, 195.66, 296.82, 251.05], \"box_3d\": [1.52, 1.51, 3.94, -11.72, 2.24, 21.61], \"ry\": -0.04}, {\"category\": \"Car\", \"angle\": 0.69, \"box_2d\": [0.0, 197.46, 121.38, 254.25], \"box_3d\": [1.58, 1.56, 4.24, -17.36, 2.36, 21.67], \"ry\": 0.02}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [686.83, 177.04, 756.49, 234.29], \"box_3d\": [1.54, 1.51, 3.58, 3.19, 1.68, 21.42], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [658.74, 175.94, 699.96, 208.89], \"box_3d\": [1.49, 1.55, 3.93, 3.28, 1.66, 35.04], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 264, - "data_source": "Mono3DRefer", - "scene_name": "000603", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (707.05, 707.05), (cx, cy) = (604.08, 180.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.96, \"box_2d\": [386.94, 180.98, 444.74, 211.53], \"box_3d\": [1.49, 1.58, 3.84, -9.95, 1.54, 37.46], \"ry\": 1.71}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [736.21, 173.49, 850.12, 269.94], \"box_3d\": [1.54, 1.57, 3.52, 3.31, 1.43, 13.13], \"ry\": -1.44}, {\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [701.98, 174.7, 793.54, 240.71], \"box_3d\": [1.5, 1.63, 3.57, 3.5, 1.37, 18.06], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [682.42, 176.25, 740.62, 223.78], \"box_3d\": [1.47, 1.59, 3.64, 3.52, 1.34, 23.98], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [667.8, 175.85, 718.99, 215.09], \"box_3d\": [1.49, 1.74, 4.09, 3.57, 1.31, 29.29], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [654.6, 176.88, 685.71, 203.37], \"box_3d\": [1.44, 1.58, 3.28, 3.72, 1.25, 40.74], \"ry\": -1.53}]\n```" - } - ] - }, - { - "id": 265, - "data_source": "Mono3DRefer", - "scene_name": "000606", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.26, \"box_2d\": [444.56, 183.98, 520.66, 206.46], \"box_3d\": [1.36, 1.68, 4.49, -8.12, 2.1, 46.04], \"ry\": 0.08}]\n```" - } - ] - }, - { - "id": 266, - "data_source": "Mono3DRefer", - "scene_name": "000607", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [581.72, 178.01, 621.54, 204.59], \"box_3d\": [1.44, 1.68, 4.39, -0.45, 1.75, 41.55], \"ry\": 1.7}]\n```" - } - ] - }, - { - "id": 267, - "data_source": "Mono3DRefer", - "scene_name": "000609", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.97, \"box_2d\": [259.26, 191.38, 333.31, 227.58], \"box_3d\": [1.56, 1.76, 4.18, -15.07, 2.5, 34.82], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [354.19, 182.17, 400.02, 210.15], \"box_3d\": [1.95, 1.83, 4.89, -17.2, 2.65, 53.46], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 268, - "data_source": "Mono3DRefer", - "scene_name": "000610", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [399.15, 179.21, 460.91, 218.14], \"box_3d\": [1.56, 1.73, 4.21, -7.79, 1.86, 31.46], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 269, - "data_source": "Mono3DRefer", - "scene_name": "000611", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [599.75, 177.03, 628.94, 205.13], \"box_3d\": [1.66, 1.73, 3.05, 0.23, 1.94, 44.68], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [309.27, 187.45, 371.76, 220.72], \"box_3d\": [1.44, 1.63, 3.43, -12.7, 2.16, 34.1], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 270, - "data_source": "Mono3DRefer", - "scene_name": "000616", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [329.11, 176.02, 480.45, 292.44], \"box_3d\": [1.7, 1.53, 3.82, -3.42, 1.78, 12.56], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [479.04, 181.77, 539.19, 232.94], \"box_3d\": [1.5, 1.53, 3.38, -3.21, 1.8, 23.2], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [535.86, 174.16, 586.25, 205.62], \"box_3d\": [1.56, 1.56, 3.6, -2.54, 1.65, 37.75], \"ry\": 1.83}]\n```" - } - ] - }, - { - "id": 271, - "data_source": "Mono3DRefer", - "scene_name": "000621", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [195.63, 187.23, 436.96, 361.63], \"box_3d\": [1.7, 1.53, 3.82, -3.47, 1.91, 9.18], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [449.9, 191.29, 524.15, 253.96], \"box_3d\": [1.5, 1.53, 3.38, -3.32, 2.04, 19.78], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [523.19, 185.76, 578.93, 221.55], \"box_3d\": [1.56, 1.56, 3.6, -2.76, 2.19, 34.09], \"ry\": 1.81}]\n```" - } - ] - }, - { - "id": 272, - "data_source": "Mono3DRefer", - "scene_name": "000622", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [655.03, 180.12, 688.7, 206.76], \"box_3d\": [1.69, 1.88, 4.5, 4.19, 2.19, 48.51], \"ry\": 1.74}]\n```" - } - ] - }, - { - "id": 273, - "data_source": "Mono3DRefer", - "scene_name": "000623", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.31, \"box_2d\": [0.0, 216.46, 151.24, 374.0], \"box_3d\": [1.4, 1.59, 3.89, -5.29, 1.81, 5.12], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [666.07, 181.12, 749.19, 241.39], \"box_3d\": [1.41, 1.64, 3.77, 2.43, 1.64, 19.06], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [503.81, 182.17, 540.99, 207.2], \"box_3d\": [1.43, 1.76, 3.97, -5.33, 2.02, 44.11], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [406.0, 187.3, 485.52, 239.32], \"box_3d\": [1.49, 1.61, 3.94, -5.25, 1.98, 23.34], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [442.93, 184.32, 507.01, 225.6], \"box_3d\": [1.48, 1.66, 3.9, -5.29, 1.95, 28.5], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [264.0, 202.62, 416.25, 282.37], \"box_3d\": [1.31, 1.58, 4.23, -5.54, 2.01, 15.26], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 274, - "data_source": "Mono3DRefer", - "scene_name": "000624", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [652.33, 172.82, 725.8, 242.85], \"box_3d\": [1.57, 1.56, 2.62, 1.83, 1.59, 17.66], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [422.71, 183.1, 525.17, 247.51], \"box_3d\": [1.49, 1.53, 3.88, -3.48, 1.78, 18.94], \"ry\": 1.69}]\n```" - } - ] - }, - { - "id": 275, - "data_source": "Mono3DRefer", - "scene_name": "000626", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.44, \"box_2d\": [1031.41, 180.01, 1241.0, 374.0], \"box_3d\": [1.46, 1.41, 3.63, 3.15, 1.48, 2.41], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.84, \"box_2d\": [752.11, 171.69, 911.06, 266.25], \"box_3d\": [1.42, 1.55, 4.36, 3.75, 1.43, 13.13], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.48, \"box_2d\": [558.65, 180.69, 613.39, 218.59], \"box_3d\": [1.33, 1.71, 4.09, -0.88, 1.65, 27.8], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": -1.47, \"box_2d\": [589.02, 177.91, 627.14, 207.54], \"box_3d\": [1.42, 1.53, 4.12, -0.08, 1.7, 37.06], \"ry\": -1.47}]\n```" - } - ] - }, - { - "id": 276, - "data_source": "Mono3DRefer", - "scene_name": "000627", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [409.2, 183.32, 484.47, 243.32], \"box_3d\": [1.47, 1.7, 4.39, -4.51, 1.78, 20.28], \"ry\": 1.43}, {\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [11.5, 172.55, 405.84, 374.0], \"box_3d\": [1.68, 1.63, 3.9, -3.18, 1.71, 6.59], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 1.43, \"box_2d\": [356.48, 185.52, 390.23, 210.49], \"box_3d\": [1.65, 1.75, 3.39, -16.55, 2.54, 50.31], \"ry\": 1.11}]\n```" - } - ] - }, - { - "id": 277, - "data_source": "Mono3DRefer", - "scene_name": "000629", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [357.52, 191.36, 449.22, 267.39], \"box_3d\": [1.41, 1.59, 4.47, -4.63, 1.88, 16.51], \"ry\": 1.39}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [528.63, 183.09, 553.69, 203.09], \"box_3d\": [1.64, 1.69, 3.3, -5.94, 2.53, 61.84], \"ry\": -1.8}]\n```" - } - ] - }, - { - "id": 278, - "data_source": "Mono3DRefer", - "scene_name": "000631", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [653.29, 177.38, 691.06, 207.5], \"box_3d\": [1.42, 1.45, 3.2, 3.02, 1.66, 35.92], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": -1.87, \"box_2d\": [739.84, 187.9, 904.88, 294.55], \"box_3d\": [1.38, 1.49, 3.32, 3.11, 1.64, 11.34], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.83, \"box_2d\": [741.51, 176.79, 793.12, 210.28], \"box_3d\": [1.43, 1.48, 3.25, 7.07, 1.63, 32.84], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": -2.0, \"box_2d\": [825.75, 179.05, 946.91, 237.9], \"box_3d\": [1.44, 1.56, 3.96, 7.42, 1.62, 19.88], \"ry\": -1.65}]\n```" - } - ] - }, - { - "id": 279, - "data_source": "Mono3DRefer", - "scene_name": "000634", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.44, \"box_2d\": [0.0, 202.51, 230.67, 374.0], \"box_3d\": [1.56, 1.68, 3.91, -3.37, 1.74, 2.7], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [407.17, 187.36, 512.93, 250.91], \"box_3d\": [1.43, 1.75, 4.47, -3.86, 1.85, 19.14], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [527.8, 174.51, 556.55, 198.35], \"box_3d\": [1.39, 1.48, 3.76, -4.16, 1.51, 44.36], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 280, - "data_source": "Mono3DRefer", - "scene_name": "000636", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [584.81, 172.59, 604.8, 188.42], \"box_3d\": [1.26, 1.6, 3.56, -1.29, 1.26, 59.84], \"ry\": -1.6}]\n```" - } - ] - }, - { - "id": 281, - "data_source": "Mono3DRefer", - "scene_name": "000637", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (707.05, 707.05), (cx, cy) = (604.08, 180.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.2, \"box_2d\": [124.4, 180.65, 273.81, 242.02], \"box_3d\": [1.49, 1.58, 3.84, -11.05, 1.52, 19.54], \"ry\": 1.69}, {\"category\": \"Car\", \"angle\": -2.27, \"box_2d\": [1032.52, 184.78, 1223.0, 369.0], \"box_3d\": [1.54, 1.57, 3.52, 3.27, 1.58, 2.69], \"ry\": -1.45}, {\"category\": \"Car\", \"angle\": -1.98, \"box_2d\": [802.42, 176.8, 1122.02, 360.3], \"box_3d\": [1.5, 1.63, 3.57, 3.42, 1.47, 7.64], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [729.52, 177.63, 854.26, 267.35], \"box_3d\": [1.47, 1.59, 3.64, 3.4, 1.43, 13.59], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [694.51, 177.52, 783.16, 240.71], \"box_3d\": [1.49, 1.74, 4.09, 3.44, 1.43, 19.02], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [681.79, 175.68, 741.72, 220.92], \"box_3d\": [1.48, 1.71, 4.05, 3.75, 1.32, 25.5], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [663.44, 174.69, 707.8, 210.87], \"box_3d\": [1.44, 1.58, 3.28, 3.38, 1.21, 30.23], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [560.24, 171.85, 583.08, 188.69], \"box_3d\": [1.4, 1.54, 4.5, -2.86, 0.67, 61.98], \"ry\": 1.63}]\n```" - } - ] - }, - { - "id": 282, - "data_source": "Mono3DRefer", - "scene_name": "000639", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.25, \"box_2d\": [734.4, 178.36, 958.26, 323.65], \"box_3d\": [1.53, 1.57, 3.87, 2.82, 1.62, 9.66], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": -1.83, \"box_2d\": [645.97, 179.68, 743.06, 243.31], \"box_3d\": [1.45, 1.61, 3.34, 2.01, 1.64, 18.4], \"ry\": -1.73}, {\"category\": \"Car\", \"angle\": -1.42, \"box_2d\": [363.21, 188.67, 442.6, 241.35], \"box_3d\": [1.34, 1.69, 3.6, -5.98, 1.83, 21.06], \"ry\": -1.7}, {\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [603.7, 169.86, 674.06, 222.18], \"box_3d\": [1.69, 1.59, 3.77, 0.9, 1.62, 25.44], \"ry\": -1.78}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [558.28, 181.17, 601.97, 211.75], \"box_3d\": [1.49, 1.71, 3.58, -1.63, 1.93, 37.49], \"ry\": -1.77}, {\"category\": \"Car\", \"angle\": -2.11, \"box_2d\": [836.46, 120.27, 1241.0, 374.0], \"box_3d\": [1.78, 1.59, 4.33, 2.83, 1.62, 4.41], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 283, - "data_source": "Mono3DRefer", - "scene_name": "000640", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.29, \"box_2d\": [265.37, 182.73, 500.15, 339.26], \"box_3d\": [1.47, 1.59, 4.03, -2.42, 1.5, 8.9], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -1.41, \"box_2d\": [467.44, 175.52, 547.77, 234.95], \"box_3d\": [1.41, 1.54, 3.36, -2.39, 1.27, 18.85], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [594.9, 166.94, 642.7, 210.54], \"box_3d\": [1.52, 1.67, 4.38, 0.64, 1.01, 27.58], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.5, \"box_2d\": [552.99, 164.31, 591.7, 199.03], \"box_3d\": [1.56, 1.59, 3.65, -1.35, 0.78, 34.5], \"ry\": -1.54}]\n```" - } - ] - }, - { - "id": 284, - "data_source": "Mono3DRefer", - "scene_name": "000648", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [494.96, 183.76, 553.21, 224.84], \"box_3d\": [1.46, 1.75, 4.23, -3.34, 1.91, 28.46], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 285, - "data_source": "Mono3DRefer", - "scene_name": "000649", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.24, \"box_2d\": [0.0, 196.31, 179.37, 312.09], \"box_3d\": [1.47, 1.63, 4.11, -9.02, 1.9, 11.56], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 2.03, \"box_2d\": [187.18, 176.46, 326.58, 256.56], \"box_3d\": [1.8, 1.61, 3.83, -8.81, 1.91, 18.29], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 2.03, \"box_2d\": [295.85, 187.72, 388.79, 237.39], \"box_3d\": [1.45, 1.45, 3.4, -8.61, 1.95, 23.36], \"ry\": 1.68}, {\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [350.62, 178.47, 431.74, 227.48], \"box_3d\": [1.75, 1.71, 4.31, -8.47, 1.99, 28.19], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [433.63, 182.48, 484.57, 213.81], \"box_3d\": [1.53, 1.66, 4.01, -7.87, 2.05, 37.81], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [461.89, 181.56, 504.01, 208.04], \"box_3d\": [1.48, 1.6, 3.9, -7.54, 2.01, 42.99], \"ry\": 1.63}]\n```" - } - ] - }, - { - "id": 286, - "data_source": "Mono3DRefer", - "scene_name": "000651", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [583.23, 179.47, 637.6, 223.67], \"box_3d\": [1.36, 1.69, 3.38, -0.03, 1.6, 24.26], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -2.12, \"box_2d\": [114.19, 182.87, 393.81, 308.47], \"box_3d\": [1.42, 1.61, 4.03, -5.07, 1.58, 10.01], \"ry\": -2.58}, {\"category\": \"Car\", \"angle\": -2.24, \"box_2d\": [365.5, 165.62, 509.69, 234.33], \"box_3d\": [1.76, 1.74, 4.16, -4.97, 1.6, 20.4], \"ry\": -2.48}, {\"category\": \"Car\", \"angle\": -2.12, \"box_2d\": [411.46, 163.87, 515.26, 224.57], \"box_3d\": [1.83, 1.64, 3.71, -4.92, 1.59, 23.72], \"ry\": -2.32}, {\"category\": \"Car\", \"angle\": -2.29, \"box_2d\": [467.58, 167.52, 547.57, 207.64], \"box_3d\": [1.8, 1.63, 3.79, -4.91, 1.59, 34.14], \"ry\": -2.43}, {\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [586.25, 174.16, 620.6, 205.5], \"box_3d\": [1.52, 1.67, 3.61, -0.34, 1.61, 37.1], \"ry\": -1.56}]\n```" - } - ] - }, - { - "id": 287, - "data_source": "Mono3DRefer", - "scene_name": "000656", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.49, \"box_2d\": [564.42, 166.78, 593.54, 190.19], \"box_3d\": [1.47, 1.59, 4.03, -1.43, 0.54, 47.61], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [615.84, 163.79, 639.01, 184.33], \"box_3d\": [1.52, 1.67, 4.38, 2.09, 0.2, 56.11], \"ry\": -1.51}]\n```" - } - ] - }, - { - "id": 288, - "data_source": "Mono3DRefer", - "scene_name": "000657", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [586.67, 173.35, 607.94, 189.95], \"box_3d\": [1.26, 1.6, 3.56, -1.0, 1.32, 57.16], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 289, - "data_source": "Mono3DRefer", - "scene_name": "000659", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.1, \"box_2d\": [171.91, 186.92, 262.23, 225.78], \"box_3d\": [1.45, 1.74, 3.42, -16.03, 2.05, 29.55], \"ry\": 1.61}]\n```" - } - ] - }, - { - "id": 290, - "data_source": "Mono3DRefer", - "scene_name": "000661", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [726.07, 183.15, 826.41, 254.12], \"box_3d\": [1.42, 1.45, 3.2, 3.6, 1.66, 16.29], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -2.02, \"box_2d\": [895.18, 181.05, 1037.74, 254.94], \"box_3d\": [1.43, 1.48, 3.25, 7.6, 1.62, 15.79], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 291, - "data_source": "Mono3DRefer", - "scene_name": "000662", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.53, \"box_2d\": [379.9, 194.61, 418.84, 230.55], \"box_3d\": [1.65, 1.75, 3.39, -10.75, 2.79, 36.58], \"ry\": 1.25}, {\"category\": \"Car\", \"angle\": 2.17, \"box_2d\": [0.0, 210.7, 353.24, 374.0], \"box_3d\": [1.47, 1.7, 4.39, -2.79, 1.75, 3.38], \"ry\": 1.52}]\n```" - } - ] - }, - { - "id": 292, - "data_source": "Mono3DRefer", - "scene_name": "000664", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.27, \"box_2d\": [270.69, 185.23, 454.8, 303.7], \"box_3d\": [1.41, 1.54, 3.36, -3.3, 1.48, 10.47], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [533.28, 172.04, 593.61, 227.07], \"box_3d\": [1.52, 1.67, 4.38, -1.15, 1.26, 22.33], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.46, \"box_2d\": [516.03, 169.82, 563.22, 209.42], \"box_3d\": [1.56, 1.59, 3.65, -2.57, 1.1, 30.46], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [408.43, 176.71, 462.6, 210.83], \"box_3d\": [1.39, 1.46, 3.37, -7.17, 1.2, 31.27], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [589.1, 170.17, 620.7, 201.87], \"box_3d\": [1.48, 1.47, 2.23, 0.17, 0.95, 34.94], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 293, - "data_source": "Mono3DRefer", - "scene_name": "000668", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 3.03, \"box_2d\": [452.37, 162.7, 723.79, 270.96], \"box_3d\": [1.79, 1.69, 4.72, -0.48, 1.64, 13.35], \"ry\": 2.99}, {\"category\": \"Car\", \"angle\": -0.05, \"box_2d\": [0.0, 163.89, 29.07, 205.36], \"box_3d\": [1.48, 1.74, 3.92, -23.24, 1.17, 25.56], \"ry\": -0.78}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [1204.53, 175.64, 1241.0, 292.02], \"box_3d\": [1.57, 1.65, 3.35, 9.94, 1.61, 10.71], \"ry\": -0.85}, {\"category\": \"Car\", \"angle\": 1.97, \"box_2d\": [837.18, 176.05, 957.38, 233.84], \"box_3d\": [1.43, 1.7, 3.95, 8.0, 1.54, 20.02], \"ry\": 2.34}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [983.05, 165.46, 1061.76, 207.13], \"box_3d\": [1.54, 1.68, 3.79, 16.38, 1.28, 28.64], \"ry\": 2.38}]\n```" - } - ] - }, - { - "id": 294, - "data_source": "Mono3DRefer", - "scene_name": "000672", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.08, \"box_2d\": [0.0, 190.35, 398.47, 374.0], \"box_3d\": [1.6, 1.68, 3.94, -3.02, 1.76, 5.21], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [524.95, 182.98, 567.7, 215.64], \"box_3d\": [1.42, 1.66, 4.11, -2.99, 1.92, 34.15], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 295, - "data_source": "Mono3DRefer", - "scene_name": "000673", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [243.02, 181.5, 439.76, 304.0], \"box_3d\": [1.59, 1.63, 3.83, -4.04, 1.74, 11.41], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [660.55, 169.66, 694.81, 202.08], \"box_3d\": [1.62, 1.7, 3.29, 3.51, 1.49, 38.17], \"ry\": -1.5}]\n```" - } - ] - }, - { - "id": 296, - "data_source": "Mono3DRefer", - "scene_name": "000674", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.87, \"box_2d\": [705.11, 176.13, 825.04, 258.62], \"box_3d\": [1.53, 1.5, 3.34, 3.11, 1.62, 15.25], \"ry\": -1.67}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [373.04, 187.73, 453.86, 244.25], \"box_3d\": [1.46, 1.43, 3.52, -5.67, 1.92, 21.01], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [693.38, 178.53, 765.98, 232.01], \"box_3d\": [1.39, 1.45, 3.44, 3.31, 1.57, 20.72], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [276.19, 188.51, 412.89, 270.87], \"box_3d\": [1.56, 1.63, 3.62, -5.76, 1.94, 15.98], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -1.89, \"box_2d\": [746.78, 177.72, 961.13, 314.74], \"box_3d\": [1.51, 1.6, 3.54, 3.02, 1.59, 9.83], \"ry\": -1.61}]\n```" - } - ] - }, - { - "id": 297, - "data_source": "Mono3DRefer", - "scene_name": "000678", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.47, \"box_2d\": [349.31, 189.3, 385.09, 217.88], \"box_3d\": [1.65, 1.75, 3.39, -15.11, 2.69, 44.73], \"ry\": 1.14}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [329.85, 190.87, 467.65, 291.03], \"box_3d\": [1.47, 1.7, 4.39, -3.75, 1.84, 13.4], \"ry\": 1.46}]\n```" - } - ] - }, - { - "id": 298, - "data_source": "Mono3DRefer", - "scene_name": "000681", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.28, \"box_2d\": [421.83, 183.09, 507.27, 208.2], \"box_3d\": [1.36, 1.68, 4.49, -8.31, 1.97, 41.27], \"ry\": 0.08}]\n```" - } - ] - }, - { - "id": 299, - "data_source": "Mono3DRefer", - "scene_name": "000682", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [646.17, 188.33, 674.33, 211.09], \"box_3d\": [1.41, 1.58, 4.36, 3.34, 2.49, 48.88], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 300, - "data_source": "Mono3DRefer", - "scene_name": "000688", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.17, \"box_2d\": [52.92, 184.24, 162.46, 228.17], \"box_3d\": [1.58, 1.64, 3.88, -19.73, 2.05, 28.5], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 2.12, \"box_2d\": [143.47, 185.15, 235.33, 218.62], \"box_3d\": [1.43, 1.81, 4.16, -19.65, 2.03, 33.88], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 2.03, \"box_2d\": [234.06, 183.82, 299.94, 211.52], \"box_3d\": [1.47, 1.67, 4.31, -19.61, 2.12, 41.4], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.99, \"box_2d\": [289.52, 181.61, 335.67, 205.61], \"box_3d\": [1.48, 1.44, 3.59, -19.32, 2.06, 46.94], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.97, \"box_2d\": [324.08, 182.9, 366.25, 203.63], \"box_3d\": [1.42, 1.68, 3.38, -19.15, 2.17, 52.23], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [379.56, 179.21, 410.83, 197.87], \"box_3d\": [1.57, 1.47, 3.35, -18.75, 2.14, 63.05], \"ry\": 1.66}]\n```" - } - ] - }, - { - "id": 301, - "data_source": "Mono3DRefer", - "scene_name": "000689", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.29, \"box_2d\": [401.63, 183.37, 495.78, 210.93], \"box_3d\": [1.36, 1.68, 4.49, -8.41, 1.94, 37.67], \"ry\": 0.07}]\n```" - } - ] - }, - { - "id": 302, - "data_source": "Mono3DRefer", - "scene_name": "000697", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.24, \"box_2d\": [16.1, 190.5, 122.12, 230.21], \"box_3d\": [1.62, 1.6, 4.18, -24.52, 2.46, 32.85], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 2.19, \"box_2d\": [120.71, 190.41, 193.29, 219.97], \"box_3d\": [1.48, 1.35, 3.93, -24.92, 2.48, 39.78], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": 2.07, \"box_2d\": [168.72, 186.17, 232.04, 212.7], \"box_3d\": [1.5, 1.62, 4.08, -25.02, 2.34, 44.18], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [223.1, 184.95, 272.06, 210.16], \"box_3d\": [1.65, 1.57, 3.82, -25.19, 2.51, 50.24], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 303, - "data_source": "Mono3DRefer", - "scene_name": "000699", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.17, \"box_2d\": [750.43, 184.37, 940.84, 276.22], \"box_3d\": [1.42, 1.58, 3.48, 4.09, 1.64, 13.24], \"ry\": -1.87}]\n```" - } - ] - }, - { - "id": 304, - "data_source": "Mono3DRefer", - "scene_name": "000700", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [292.06, 183.14, 399.96, 243.71], \"box_3d\": [1.48, 1.64, 3.59, -7.16, 1.78, 19.81], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 305, - "data_source": "Mono3DRefer", - "scene_name": "000704", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.8, \"box_2d\": [0.0, 207.75, 285.16, 374.0], \"box_3d\": [1.42, 1.53, 4.12, -2.99, 1.65, 2.78], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [658.61, 173.96, 712.98, 218.01], \"box_3d\": [1.38, 1.64, 3.51, 2.47, 1.44, 24.53], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": -1.88, \"box_2d\": [750.01, 163.51, 1068.4, 365.01], \"box_3d\": [1.49, 1.56, 4.34, 2.61, 1.45, 7.55], \"ry\": -1.56}]\n```" - } - ] - }, - { - "id": 306, - "data_source": "Mono3DRefer", - "scene_name": "000706", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [692.95, 185.42, 785.1, 256.39], \"box_3d\": [1.51, 1.56, 3.81, 3.0, 1.84, 17.67], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [672.95, 181.89, 745.46, 236.91], \"box_3d\": [1.57, 1.68, 3.83, 3.02, 1.87, 22.9], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.6, \"box_2d\": [525.23, 181.22, 558.8, 214.58], \"box_3d\": [1.52, 1.5, 3.85, -3.3, 1.93, 35.15], \"ry\": 1.51}]\n```" - } - ] - }, - { - "id": 307, - "data_source": "Mono3DRefer", - "scene_name": "000707", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [360.9, 189.01, 501.88, 280.68], \"box_3d\": [1.52, 1.73, 4.28, -3.46, 1.88, 14.64], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [523.92, 182.29, 564.4, 211.63], \"box_3d\": [1.44, 1.7, 3.65, -3.44, 1.95, 37.98], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 308, - "data_source": "Mono3DRefer", - "scene_name": "000708", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [479.01, 178.29, 532.96, 215.21], \"box_3d\": [1.49, 1.63, 3.68, -4.48, 1.74, 31.34], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [694.1, 177.09, 750.76, 219.62], \"box_3d\": [1.49, 1.61, 3.55, 4.16, 1.67, 27.37], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 2.24, \"box_2d\": [0.0, 206.01, 223.92, 374.0], \"box_3d\": [1.43, 1.68, 3.76, -4.82, 1.76, 5.52], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 2.0, \"box_2d\": [133.31, 193.71, 387.66, 317.21], \"box_3d\": [1.38, 1.68, 4.34, -4.91, 1.75, 10.81], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -2.1, \"box_2d\": [852.51, 189.92, 1241.0, 374.0], \"box_3d\": [1.5, 1.64, 4.14, 3.44, 1.67, 5.74], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [451.97, 179.29, 512.11, 221.68], \"box_3d\": [1.45, 1.6, 3.2, -4.69, 1.7, 26.61], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.87, \"box_2d\": [762.14, 182.12, 958.3, 306.43], \"box_3d\": [1.51, 1.66, 3.5, 3.41, 1.66, 10.65], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [726.86, 178.02, 834.57, 253.91], \"box_3d\": [1.51, 1.61, 3.51, 3.67, 1.64, 16.25], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [700.3, 178.28, 779.97, 232.88], \"box_3d\": [1.49, 1.68, 3.8, 3.8, 1.67, 21.84], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 309, - "data_source": "Mono3DRefer", - "scene_name": "000710", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [380.63, 188.4, 463.01, 258.09], \"box_3d\": [1.41, 1.59, 4.47, -4.5, 1.82, 17.6], \"ry\": 1.39}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [532.31, 181.4, 557.03, 200.98], \"box_3d\": [1.64, 1.69, 3.3, -5.73, 2.4, 62.92], \"ry\": -1.8}]\n```" - } - ] - }, - { - "id": 310, - "data_source": "Mono3DRefer", - "scene_name": "000714", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.14, \"box_2d\": [493.65, 190.36, 662.41, 278.85], \"box_3d\": [1.48, 1.71, 4.65, -0.5, 1.89, 15.15], \"ry\": -1.17}, {\"category\": \"Car\", \"angle\": -0.95, \"box_2d\": [364.44, 168.81, 545.33, 272.45], \"box_3d\": [2.04, 1.66, 4.62, -3.4, 1.98, 16.64], \"ry\": -1.15}, {\"category\": \"Car\", \"angle\": -1.46, \"box_2d\": [758.42, 171.96, 835.76, 230.05], \"box_3d\": [1.42, 1.64, 3.95, 5.18, 1.42, 19.8], \"ry\": -1.21}, {\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [968.02, 163.65, 1031.83, 211.79], \"box_3d\": [1.7, 1.8, 4.5, 14.9, 1.4, 27.82], \"ry\": -1.15}, {\"category\": \"Car\", \"angle\": -0.76, \"box_2d\": [182.85, 193.28, 374.91, 264.92], \"box_3d\": [1.53, 1.61, 4.73, -8.47, 2.12, 18.9], \"ry\": -1.18}, {\"category\": \"Car\", \"angle\": -0.68, \"box_2d\": [99.8, 201.34, 248.33, 258.41], \"box_3d\": [1.4, 1.56, 3.48, -12.56, 2.28, 20.95], \"ry\": -1.22}]\n```" - } - ] - }, - { - "id": 311, - "data_source": "Mono3DRefer", - "scene_name": "000715", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [758.36, 173.03, 779.23, 190.1], \"box_3d\": [1.48, 1.36, 3.51, 14.23, 1.52, 64.84], \"ry\": -1.49}]\n```" - } - ] - }, - { - "id": 312, - "data_source": "Mono3DRefer", - "scene_name": "000717", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [672.8, 173.85, 763.09, 256.1], \"box_3d\": [1.57, 1.56, 2.62, 2.15, 1.61, 15.23], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [395.48, 187.82, 522.15, 269.2], \"box_3d\": [1.49, 1.53, 3.88, -3.16, 1.85, 15.63], \"ry\": 1.69}]\n```" - } - ] - }, - { - "id": 313, - "data_source": "Mono3DRefer", - "scene_name": "000728", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [467.5, 184.4, 515.93, 218.42], \"box_3d\": [1.39, 1.52, 3.56, -5.23, 1.92, 32.08], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [290.0, 186.4, 436.7, 286.75], \"box_3d\": [1.46, 1.59, 3.35, -4.16, 1.72, 12.54], \"ry\": 1.5}]\n```" - } - ] - }, - { - "id": 314, - "data_source": "Mono3DRefer", - "scene_name": "000732", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.31, \"box_2d\": [482.87, 167.48, 555.74, 203.65], \"box_3d\": [1.8, 1.63, 3.79, -4.79, 1.56, 37.66], \"ry\": -2.43}, {\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [588.88, 174.51, 622.95, 205.59], \"box_3d\": [1.52, 1.67, 3.61, -0.21, 1.63, 37.39], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [585.72, 178.4, 639.73, 222.31], \"box_3d\": [1.36, 1.69, 3.38, 0.05, 1.57, 24.41], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -2.23, \"box_2d\": [238.82, 183.21, 452.36, 271.77], \"box_3d\": [1.42, 1.61, 4.03, -5.04, 1.64, 13.47], \"ry\": -2.58}, {\"category\": \"Car\", \"angle\": -2.27, \"box_2d\": [401.36, 166.57, 525.71, 224.43], \"box_3d\": [1.76, 1.74, 4.16, -4.93, 1.6, 23.86], \"ry\": -2.47}, {\"category\": \"Car\", \"angle\": -2.16, \"box_2d\": [435.77, 163.18, 528.61, 215.65], \"box_3d\": [1.83, 1.64, 3.71, -4.89, 1.52, 27.13], \"ry\": -2.33}]\n```" - } - ] - }, - { - "id": 315, - "data_source": "Mono3DRefer", - "scene_name": "000734", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [665.44, 175.98, 791.75, 261.89], \"box_3d\": [1.49, 1.64, 4.17, 2.21, 1.57, 14.76], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": -1.11, \"box_2d\": [0.0, 187.93, 365.1, 374.0], \"box_3d\": [1.44, 1.65, 4.25, -3.45, 1.6, 5.77], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": -2.67, \"box_2d\": [160.12, 189.92, 332.37, 252.83], \"box_3d\": [1.47, 1.58, 3.55, -9.15, 1.93, 18.23], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": -1.52, \"box_2d\": [500.49, 173.3, 551.36, 217.64], \"box_3d\": [1.6, 1.75, 3.89, -3.26, 1.64, 28.27], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [632.9, 176.49, 695.02, 222.6], \"box_3d\": [1.59, 1.64, 4.03, 1.92, 1.75, 27.22], \"ry\": -1.67}, {\"category\": \"Car\", \"angle\": -1.45, \"box_2d\": [528.07, 179.56, 562.03, 206.09], \"box_3d\": [1.46, 1.56, 3.21, -3.75, 1.86, 41.81], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [619.47, 170.48, 649.95, 202.16], \"box_3d\": [1.83, 1.75, 4.52, 1.47, 1.72, 44.22], \"ry\": -1.55}]\n```" - } - ] - }, - { - "id": 316, - "data_source": "Mono3DRefer", - "scene_name": "000737", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.65, \"box_2d\": [0.0, 165.23, 270.29, 326.76], \"box_3d\": [1.68, 1.71, 4.07, -6.47, 1.63, 8.46], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": -2.67, \"box_2d\": [0.0, 174.09, 349.34, 289.01], \"box_3d\": [1.53, 1.7, 4.32, -6.41, 1.57, 10.79], \"ry\": 3.09}, {\"category\": \"Car\", \"angle\": 0.36, \"box_2d\": [233.78, 169.37, 448.59, 243.31], \"box_3d\": [1.54, 1.78, 4.2, -6.01, 1.5, 16.36], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": 0.17, \"box_2d\": [378.27, 172.24, 511.34, 217.17], \"box_3d\": [1.5, 1.66, 4.4, -5.84, 1.51, 25.73], \"ry\": -0.05}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [661.27, 176.83, 724.48, 223.2], \"box_3d\": [1.41, 1.62, 4.19, 2.67, 1.56, 24.33], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -0.14, \"box_2d\": [724.12, 175.19, 810.64, 211.99], \"box_3d\": [1.42, 1.59, 3.21, 6.3, 1.54, 29.26], \"ry\": 0.07}]\n```" - } - ] - }, - { - "id": 317, - "data_source": "Mono3DRefer", - "scene_name": "000738", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [644.47, 180.1, 690.95, 216.69], \"box_3d\": [1.64, 1.69, 3.3, 2.67, 2.0, 34.42], \"ry\": -1.65}]\n```" - } - ] - }, - { - "id": 318, - "data_source": "Mono3DRefer", - "scene_name": "000740", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.15, \"box_2d\": [407.36, 179.05, 651.26, 309.32], \"box_3d\": [1.57, 1.65, 4.1, -1.04, 1.68, 10.99], \"ry\": 2.06}, {\"category\": \"Car\", \"angle\": 1.07, \"box_2d\": [784.73, 194.05, 1241.0, 374.0], \"box_3d\": [1.42, 1.6, 3.78, 2.38, 1.59, 4.51], \"ry\": 1.53}]\n```" - } - ] - }, - { - "id": 319, - "data_source": "Mono3DRefer", - "scene_name": "000747", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [610.92, 178.94, 635.52, 201.71], \"box_3d\": [1.54, 1.63, 3.67, 0.95, 1.99, 51.29], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [458.82, 184.03, 510.0, 214.59], \"box_3d\": [1.56, 1.59, 3.69, -6.83, 2.19, 39.43], \"ry\": 1.74}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [485.02, 177.04, 529.24, 210.33], \"box_3d\": [1.93, 1.59, 4.01, -6.26, 2.21, 44.15], \"ry\": 1.72}, {\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [656.55, 176.02, 683.32, 201.11], \"box_3d\": [1.58, 1.67, 3.84, 3.92, 1.81, 47.89], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": 2.78, \"box_2d\": [932.2, 164.14, 1013.15, 191.68], \"box_3d\": [1.39, 1.51, 3.49, 19.07, 0.97, 38.1], \"ry\": -3.04}]\n```" - } - ] - }, - { - "id": 320, - "data_source": "Mono3DRefer", - "scene_name": "000749", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [638.19, 176.1, 666.15, 199.22], \"box_3d\": [1.42, 1.45, 3.2, 2.65, 1.64, 46.29], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [689.1, 183.03, 775.3, 244.77], \"box_3d\": [1.38, 1.49, 3.32, 2.93, 1.65, 18.09], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -2.1, \"box_2d\": [852.71, 187.06, 1241.0, 374.0], \"box_3d\": [1.5, 1.62, 3.89, 3.59, 1.65, 6.27], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [705.49, 175.27, 743.11, 201.08], \"box_3d\": [1.43, 1.48, 3.25, 6.61, 1.59, 42.11], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": -1.89, \"box_2d\": [759.34, 176.75, 831.47, 216.36], \"box_3d\": [1.44, 1.56, 3.96, 7.21, 1.61, 28.54], \"ry\": -1.65}]\n```" - } - ] - }, - { - "id": 321, - "data_source": "Mono3DRefer", - "scene_name": "000753", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [12.54, 191.25, 357.9, 374.0], \"box_3d\": [1.6, 1.73, 3.62, -3.85, 1.82, 7.23], \"ry\": 1.43}, {\"category\": \"Car\", \"angle\": 2.97, \"box_2d\": [607.58, 168.14, 701.4, 202.5], \"box_3d\": [1.73, 1.74, 4.71, 2.29, 1.51, 38.22], \"ry\": 3.03}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [452.54, 178.73, 485.89, 204.42], \"box_3d\": [1.58, 1.76, 4.18, -9.13, 1.98, 46.95], \"ry\": 1.46}]\n```" - } - ] - }, - { - "id": 322, - "data_source": "Mono3DRefer", - "scene_name": "000760", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.56, \"box_2d\": [38.77, 198.38, 422.17, 316.55], \"box_3d\": [1.38, 1.66, 4.38, -5.01, 1.77, 9.73], \"ry\": 0.09}, {\"category\": \"Car\", \"angle\": -0.65, \"box_2d\": [917.51, 170.71, 1241.0, 374.0], \"box_3d\": [1.48, 1.86, 4.18, 4.68, 1.46, 5.23], \"ry\": 0.06}, {\"category\": \"Car\", \"angle\": -0.52, \"box_2d\": [812.4, 151.43, 1241.0, 319.58], \"box_3d\": [1.74, 1.66, 4.37, 4.78, 1.54, 8.52], \"ry\": -0.03}, {\"category\": \"Car\", \"angle\": -2.79, \"box_2d\": [337.68, 182.96, 507.37, 238.86], \"box_3d\": [1.47, 1.67, 4.28, -5.22, 1.78, 20.21], \"ry\": -3.04}, {\"category\": \"Car\", \"angle\": -3.04, \"box_2d\": [404.02, 181.77, 534.37, 223.64], \"box_3d\": [1.38, 1.67, 4.37, -4.94, 1.72, 25.56], \"ry\": 3.05}, {\"category\": \"Car\", \"angle\": -0.32, \"box_2d\": [781.37, 169.96, 956.87, 246.87], \"box_3d\": [1.54, 1.58, 3.2, 5.44, 1.51, 15.54], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": -0.38, \"box_2d\": [790.61, 168.85, 1049.58, 258.33], \"box_3d\": [1.51, 1.72, 4.03, 5.54, 1.47, 13.35], \"ry\": 0.01}]\n```" - } - ] - }, - { - "id": 323, - "data_source": "Mono3DRefer", - "scene_name": "000763", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 3.02, \"box_2d\": [874.58, 161.12, 937.38, 183.85], \"box_3d\": [1.59, 1.64, 4.08, 21.67, 0.78, 53.0], \"ry\": -2.88}, {\"category\": \"Car\", \"angle\": 0.77, \"box_2d\": [362.42, 176.81, 412.89, 197.28], \"box_3d\": [1.49, 1.59, 3.46, -16.62, 1.8, 53.86], \"ry\": 0.47}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [710.47, 173.82, 782.51, 225.46], \"box_3d\": [1.43, 1.59, 4.21, 4.08, 1.49, 22.32], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": 2.96, \"box_2d\": [728.08, 165.87, 785.4, 185.41], \"box_3d\": [1.41, 1.59, 4.01, 11.05, 0.92, 54.59], \"ry\": -3.12}]\n```" - } - ] - }, - { - "id": 324, - "data_source": "Mono3DRefer", - "scene_name": "000776", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [564.61, 188.21, 700.38, 275.32], \"box_3d\": [1.48, 1.51, 4.35, 0.59, 1.83, 15.03], \"ry\": 1.92}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [788.25, 182.23, 856.11, 231.31], \"box_3d\": [1.46, 1.74, 3.99, 7.08, 1.79, 23.97], \"ry\": 1.96}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [652.85, 185.5, 750.87, 244.43], \"box_3d\": [1.42, 1.68, 4.29, 2.65, 1.8, 20.23], \"ry\": 1.95}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [728.41, 185.43, 785.21, 225.57], \"box_3d\": [1.36, 1.4, 3.8, 5.55, 1.86, 27.14], \"ry\": 1.96}, {\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [789.26, 179.99, 822.6, 207.39], \"box_3d\": [1.44, 1.65, 2.96, 10.82, 1.85, 39.85], \"ry\": 1.87}]\n```" - } - ] - }, - { - "id": 325, - "data_source": "Mono3DRefer", - "scene_name": "000780", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [359.33, 174.42, 462.37, 243.47], \"box_3d\": [1.65, 1.61, 3.73, -5.24, 1.72, 19.29], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": -1.98, \"box_2d\": [788.39, 167.01, 1126.78, 374.0], \"box_3d\": [1.57, 1.54, 3.22, 2.85, 1.55, 6.78], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [708.79, 157.07, 863.48, 284.0], \"box_3d\": [1.74, 1.6, 3.86, 2.65, 1.55, 11.89], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [679.6, 157.82, 780.97, 245.86], \"box_3d\": [1.83, 1.69, 4.44, 2.69, 1.54, 17.32], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [671.5, 170.38, 734.94, 226.58], \"box_3d\": [1.61, 1.62, 3.78, 2.82, 1.57, 22.78], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [531.19, 166.63, 559.77, 195.34], \"box_3d\": [1.86, 1.57, 3.83, -4.37, 1.48, 49.07], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [437.17, 176.68, 499.76, 221.06], \"box_3d\": [1.5, 1.46, 3.7, -5.14, 1.65, 26.41], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 2.51, \"box_2d\": [0.0, 212.28, 86.09, 374.0], \"box_3d\": [1.54, 1.7, 4.32, -5.3, 1.85, 3.83], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 2.09, \"box_2d\": [0.0, 192.47, 315.3, 364.81], \"box_3d\": [1.54, 1.54, 3.72, -5.11, 1.81, 8.62], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [247.74, 184.08, 416.03, 283.3], \"box_3d\": [1.57, 1.6, 4.24, -5.13, 1.81, 13.84], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 326, - "data_source": "Mono3DRefer", - "scene_name": "000781", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.39, \"box_2d\": [292.07, 186.29, 435.74, 227.61], \"box_3d\": [1.36, 1.68, 4.49, -8.66, 1.86, 25.47], \"ry\": 0.06}]\n```" - } - ] - }, - { - "id": 327, - "data_source": "Mono3DRefer", - "scene_name": "000783", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [659.21, 167.22, 695.89, 201.77], \"box_3d\": [1.62, 1.7, 3.29, 3.29, 1.38, 35.93], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": 1.99, \"box_2d\": [113.61, 183.49, 397.75, 350.73], \"box_3d\": [1.59, 1.63, 3.83, -4.06, 1.74, 8.92], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 328, - "data_source": "Mono3DRefer", - "scene_name": "000786", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [512.91, 183.16, 557.23, 216.89], \"box_3d\": [1.52, 1.73, 4.28, -3.64, 2.04, 35.42], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [553.18, 180.85, 577.14, 199.17], \"box_3d\": [1.44, 1.7, 3.65, -3.68, 2.11, 59.45], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 329, - "data_source": "Mono3DRefer", - "scene_name": "000788", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.08, \"box_2d\": [0.0, 162.91, 291.69, 363.0], \"box_3d\": [1.86, 1.57, 3.83, -5.45, 1.8, 8.67], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 330, - "data_source": "Mono3DRefer", - "scene_name": "000791", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.15, \"box_2d\": [361.12, 189.23, 520.33, 269.31], \"box_3d\": [1.48, 1.71, 4.65, -3.65, 1.89, 16.34], \"ry\": -1.37}, {\"category\": \"Car\", \"angle\": -0.95, \"box_2d\": [229.66, 167.89, 412.83, 266.77], \"box_3d\": [2.04, 1.66, 4.62, -6.72, 1.96, 17.32], \"ry\": -1.32}, {\"category\": \"Car\", \"angle\": -1.47, \"box_2d\": [613.23, 174.5, 679.88, 226.07], \"box_3d\": [1.42, 1.64, 3.95, 1.16, 1.49, 21.99], \"ry\": -1.42}, {\"category\": \"Car\", \"angle\": -1.96, \"box_2d\": [1012.55, 174.87, 1223.0, 265.94], \"box_3d\": [1.51, 1.72, 4.5, 9.69, 1.57, 14.32], \"ry\": -1.37}, {\"category\": \"Car\", \"angle\": -1.61, \"box_2d\": [796.89, 164.05, 845.19, 205.81], \"box_3d\": [1.7, 1.8, 4.5, 9.18, 1.37, 31.74], \"ry\": -1.33}, {\"category\": \"Car\", \"angle\": -2.03, \"box_2d\": [1125.42, 162.12, 1241.0, 257.11], \"box_3d\": [1.64, 1.73, 4.43, 12.16, 1.46, 14.23], \"ry\": -1.34}, {\"category\": \"Car\", \"angle\": -0.79, \"box_2d\": [27.55, 193.41, 236.41, 265.59], \"box_3d\": [1.53, 1.61, 4.73, -12.15, 2.12, 18.73], \"ry\": -1.36}, {\"category\": \"Car\", \"angle\": -0.7, \"box_2d\": [0.0, 200.25, 89.11, 256.68], \"box_3d\": [1.4, 1.56, 3.48, -16.6, 2.21, 19.86], \"ry\": -1.39}]\n```" - } - ] - }, - { - "id": 331, - "data_source": "Mono3DRefer", - "scene_name": "000793", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.4, \"box_2d\": [462.47, 174.59, 549.96, 238.14], \"box_3d\": [1.47, 1.59, 4.03, -2.39, 1.31, 18.78], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": -1.49, \"box_2d\": [523.4, 175.6, 568.46, 212.46], \"box_3d\": [1.41, 1.54, 3.36, -2.23, 1.18, 29.4], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [601.71, 169.27, 640.76, 204.87], \"box_3d\": [1.52, 1.67, 4.38, 0.95, 1.0, 33.28], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": -1.5, \"box_2d\": [566.8, 167.41, 597.76, 194.69], \"box_3d\": [1.56, 1.59, 3.65, -1.11, 0.75, 43.41], \"ry\": -1.53}]\n```" - } - ] - }, - { - "id": 332, - "data_source": "Mono3DRefer", - "scene_name": "000794", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [520.47, 177.65, 557.26, 197.81], \"box_3d\": [1.45, 1.59, 4.28, -5.34, 1.82, 54.46], \"ry\": 1.76}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [716.6, 176.12, 739.27, 196.19], \"box_3d\": [1.43, 1.59, 4.21, 8.83, 1.7, 54.08], \"ry\": -1.41}]\n```" - } - ] - }, - { - "id": 333, - "data_source": "Mono3DRefer", - "scene_name": "000797", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [297.89, 190.29, 424.13, 288.47], \"box_3d\": [1.86, 1.57, 3.83, -5.44, 2.28, 16.11], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 334, - "data_source": "Mono3DRefer", - "scene_name": "000802", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [420.74, 183.74, 510.25, 243.19], \"box_3d\": [1.58, 1.68, 3.76, -4.23, 1.92, 21.46], \"ry\": 1.64}, {\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [469.41, 181.82, 535.65, 229.55], \"box_3d\": [1.61, 1.74, 4.06, -3.93, 1.96, 26.78], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [693.41, 174.79, 731.81, 208.97], \"box_3d\": [1.77, 1.77, 5.1, 5.64, 1.91, 40.32], \"ry\": -1.49}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [502.1, 185.79, 553.21, 220.56], \"box_3d\": [1.39, 1.55, 3.65, -3.58, 1.98, 31.6], \"ry\": 1.65}, {\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [522.09, 183.29, 567.91, 214.75], \"box_3d\": [1.46, 1.63, 3.92, -3.24, 2.01, 36.3], \"ry\": 1.66}, {\"category\": \"Car\", \"angle\": 0.45, \"box_2d\": [131.2, 185.36, 276.88, 230.0], \"box_3d\": [1.5, 1.87, 4.23, -14.73, 1.99, 26.3], \"ry\": -0.06}, {\"category\": \"Car\", \"angle\": -2.07, \"box_2d\": [875.97, 205.29, 1241.0, 374.0], \"box_3d\": [1.38, 1.73, 3.97, 3.62, 1.72, 5.75], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": -1.84, \"box_2d\": [773.57, 187.49, 959.71, 305.98], \"box_3d\": [1.5, 1.62, 4.11, 3.78, 1.76, 11.49], \"ry\": -1.53}]\n```" - } - ] - }, - { - "id": 335, - "data_source": "Mono3DRefer", - "scene_name": "000807", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [345.19, 185.83, 402.41, 214.58], \"box_3d\": [1.55, 1.88, 4.59, -13.84, 2.34, 42.48], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 336, - "data_source": "Mono3DRefer", - "scene_name": "000808", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [585.86, 182.94, 627.43, 221.2], \"box_3d\": [1.52, 1.67, 3.61, -0.13, 1.97, 30.96], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [601.88, 184.97, 665.33, 237.34], \"box_3d\": [1.36, 1.69, 3.38, 0.59, 1.74, 21.03], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.59, \"box_2d\": [967.93, 175.21, 1087.4, 209.95], \"box_3d\": [1.45, 1.67, 4.32, 18.35, 1.58, 31.86], \"ry\": 3.1}, {\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [402.14, 186.23, 446.06, 213.49], \"box_3d\": [1.59, 1.72, 3.86, -11.63, 2.45, 45.28], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 337, - "data_source": "Mono3DRefer", - "scene_name": "000810", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [668.22, 193.07, 764.47, 262.68], \"box_3d\": [1.41, 1.58, 4.36, 2.43, 1.95, 17.76], \"ry\": -1.6}]\n```" - } - ] - }, - { - "id": 338, - "data_source": "Mono3DRefer", - "scene_name": "000813", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [501.87, 178.19, 542.47, 200.03], \"box_3d\": [1.45, 1.59, 4.28, -6.11, 1.83, 50.43], \"ry\": 1.76}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [715.81, 176.69, 739.49, 197.66], \"box_3d\": [1.43, 1.59, 4.21, 8.42, 1.72, 51.84], \"ry\": -1.41}]\n```" - } - ] - }, - { - "id": 339, - "data_source": "Mono3DRefer", - "scene_name": "000816", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [537.6, 170.87, 608.26, 229.25], \"box_3d\": [1.49, 1.68, 4.08, -0.74, 1.22, 20.54], \"ry\": 1.61}]\n```" - } - ] - }, - { - "id": 340, - "data_source": "Mono3DRefer", - "scene_name": "000817", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.15, \"box_2d\": [0.0, 188.31, 337.38, 374.0], \"box_3d\": [1.62, 1.75, 4.3, -3.98, 1.78, 5.89], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 2.16, \"box_2d\": [0.0, 189.55, 239.41, 329.16], \"box_3d\": [1.47, 1.73, 4.48, -7.25, 1.75, 10.16], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": -1.19, \"box_2d\": [339.47, 190.35, 430.51, 241.23], \"box_3d\": [1.39, 1.59, 3.17, -6.84, 1.96, 22.1], \"ry\": -1.49}, {\"category\": \"Car\", \"angle\": -1.07, \"box_2d\": [263.11, 190.53, 379.52, 243.46], \"box_3d\": [1.46, 1.55, 4.28, -9.08, 2.06, 22.99], \"ry\": -1.44}]\n```" - } - ] - }, - { - "id": 341, - "data_source": "Mono3DRefer", - "scene_name": "000818", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [543.69, 164.21, 572.9, 193.67], \"box_3d\": [1.97, 1.87, 4.53, -3.63, 1.42, 51.02], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 2.03, \"box_2d\": [0.0, 204.69, 402.33, 374.0], \"box_3d\": [1.42, 1.66, 4.11, -3.05, 1.75, 5.62], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 342, - "data_source": "Mono3DRefer", - "scene_name": "000821", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.19, \"box_2d\": [424.61, 186.64, 805.46, 324.08], \"box_3d\": [1.45, 1.68, 4.53, -0.04, 1.66, 9.24], \"ry\": -0.2}]\n```" - } - ] - }, - { - "id": 343, - "data_source": "Mono3DRefer", - "scene_name": "000822", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.46, \"box_2d\": [545.87, 172.35, 584.73, 202.04], \"box_3d\": [1.47, 1.59, 4.03, -1.85, 1.01, 37.93], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [612.22, 168.98, 639.37, 192.76], \"box_3d\": [1.52, 1.67, 4.38, 1.71, 0.71, 48.75], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": -1.53, \"box_2d\": [589.63, 168.17, 610.78, 187.69], \"box_3d\": [1.56, 1.59, 3.65, -0.04, 0.49, 59.95], \"ry\": -1.53}]\n```" - } - ] - }, - { - "id": 344, - "data_source": "Mono3DRefer", - "scene_name": "000829", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [442.72, 183.1, 527.89, 243.64], \"box_3d\": [1.44, 1.62, 3.67, -3.28, 1.73, 19.36], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [505.35, 179.44, 559.04, 221.44], \"box_3d\": [1.51, 1.62, 3.68, -3.0, 1.78, 28.13], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 345, - "data_source": "Mono3DRefer", - "scene_name": "000830", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.99, \"box_2d\": [362.18, 180.43, 437.46, 215.63], \"box_3d\": [1.45, 1.59, 4.28, -9.31, 1.8, 32.2], \"ry\": 1.71}, {\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [703.86, 175.82, 734.24, 202.01], \"box_3d\": [1.43, 1.59, 4.21, 6.27, 1.62, 41.93], \"ry\": -1.45}]\n```" - } - ] - }, - { - "id": 346, - "data_source": "Mono3DRefer", - "scene_name": "000832", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [676.67, 174.6, 707.67, 194.83], \"box_3d\": [1.45, 1.61, 4.2, 6.13, 1.61, 54.37], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": -2.46, \"box_2d\": [800.46, 174.47, 866.86, 201.82], \"box_3d\": [1.48, 1.36, 3.51, 12.63, 1.6, 41.0], \"ry\": -2.16}]\n```" - } - ] - }, - { - "id": 347, - "data_source": "Mono3DRefer", - "scene_name": "000834", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.62, \"box_2d\": [547.4, 177.04, 572.4, 199.87], \"box_3d\": [1.7, 1.72, 4.11, -3.9, 2.05, 56.37], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 348, - "data_source": "Mono3DRefer", - "scene_name": "000835", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [543.59, 181.61, 571.93, 204.06], \"box_3d\": [1.52, 1.73, 4.28, -3.73, 2.16, 51.84], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 349, - "data_source": "Mono3DRefer", - "scene_name": "000836", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [612.33, 174.27, 666.38, 226.41], \"box_3d\": [1.61, 1.66, 3.2, 0.99, 1.68, 24.06], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [476.52, 180.84, 542.39, 225.69], \"box_3d\": [1.44, 1.61, 3.66, -3.48, 1.73, 25.25], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [426.55, 182.07, 489.85, 219.62], \"box_3d\": [1.41, 1.53, 3.37, -6.09, 1.79, 29.12], \"ry\": 1.67}]\n```" - } - ] - }, - { - "id": 350, - "data_source": "Mono3DRefer", - "scene_name": "000839", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [435.69, 182.58, 507.07, 237.77], \"box_3d\": [1.7, 1.72, 4.11, -4.68, 2.05, 24.69], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 351, - "data_source": "Mono3DRefer", - "scene_name": "000842", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [536.71, 170.99, 568.58, 202.82], \"box_3d\": [1.97, 1.87, 4.53, -3.74, 1.89, 47.4], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 352, - "data_source": "Mono3DRefer", - "scene_name": "000845", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.41, \"box_2d\": [196.85, 184.35, 370.3, 243.7], \"box_3d\": [1.46, 1.63, 3.85, -8.6, 1.79, 19.12], \"ry\": -0.0}, {\"category\": \"Car\", \"angle\": -2.74, \"box_2d\": [253.12, 173.9, 407.24, 233.22], \"box_3d\": [1.71, 1.69, 4.06, -8.54, 1.77, 22.12], \"ry\": -3.1}, {\"category\": \"Car\", \"angle\": -2.84, \"box_2d\": [308.45, 181.37, 428.83, 227.56], \"box_3d\": [1.51, 1.61, 3.65, -8.37, 1.83, 25.1], \"ry\": 3.13}, {\"category\": \"Car\", \"angle\": -2.84, \"box_2d\": [335.89, 180.89, 440.39, 218.3], \"box_3d\": [1.45, 1.54, 3.79, -9.06, 1.8, 29.51], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": -2.94, \"box_2d\": [365.72, 180.29, 449.34, 219.29], \"box_3d\": [1.66, 1.62, 3.34, -9.06, 2.02, 32.36], \"ry\": 3.07}, {\"category\": \"Car\", \"angle\": -2.97, \"box_2d\": [391.35, 183.43, 477.67, 213.53], \"box_3d\": [1.41, 1.59, 3.94, -8.68, 1.95, 35.78], \"ry\": 3.08}, {\"category\": \"Car\", \"angle\": -2.91, \"box_2d\": [410.7, 173.35, 485.32, 206.85], \"box_3d\": [1.69, 1.77, 3.51, -8.52, 1.74, 38.03], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": -2.96, \"box_2d\": [427.75, 176.73, 501.76, 206.49], \"box_3d\": [1.61, 1.66, 3.86, -8.21, 1.85, 40.87], \"ry\": 3.12}, {\"category\": \"Car\", \"angle\": 2.77, \"box_2d\": [442.74, 177.03, 511.51, 203.73], \"box_3d\": [1.59, 1.71, 3.9, -8.4, 1.88, 45.53], \"ry\": 2.58}, {\"category\": \"Car\", \"angle\": -2.39, \"box_2d\": [0.0, 195.91, 81.32, 330.88], \"box_3d\": [1.44, 1.52, 3.5, -8.43, 1.74, 8.26], \"ry\": 3.12}, {\"category\": \"Car\", \"angle\": -2.6, \"box_2d\": [34.88, 186.35, 269.23, 278.14], \"box_3d\": [1.59, 1.52, 3.47, -8.61, 1.87, 13.68], \"ry\": 3.13}, {\"category\": \"Car\", \"angle\": -2.66, \"box_2d\": [140.66, 179.77, 331.03, 253.29], \"box_3d\": [1.58, 1.55, 3.59, -8.59, 1.76, 16.66], \"ry\": -3.13}]\n```" - } - ] - }, - { - "id": 353, - "data_source": "Mono3DRefer", - "scene_name": "000846", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [735.32, 179.57, 851.17, 254.87], \"box_3d\": [1.43, 1.59, 4.21, 3.84, 1.59, 15.96], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": 2.7, \"box_2d\": [651.6, 172.62, 718.25, 195.4], \"box_3d\": [1.41, 1.59, 4.01, 4.86, 1.42, 47.02], \"ry\": 2.8}, {\"category\": \"Car\", \"angle\": 3.06, \"box_2d\": [828.02, 166.76, 895.02, 191.69], \"box_3d\": [1.59, 1.64, 4.08, 16.69, 1.23, 48.12], \"ry\": -2.89}, {\"category\": \"Car\", \"angle\": 2.88, \"box_2d\": [979.42, 166.53, 1068.34, 190.59], \"box_3d\": [1.46, 1.64, 4.63, 26.26, 1.1, 45.97], \"ry\": -2.88}, {\"category\": \"Car\", \"angle\": 0.59, \"box_2d\": [365.97, 179.39, 427.65, 203.44], \"box_3d\": [1.49, 1.59, 3.46, -13.6, 1.92, 45.99], \"ry\": 0.3}]\n```" - } - ] - }, - { - "id": 354, - "data_source": "Mono3DRefer", - "scene_name": "000848", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 3.0, \"box_2d\": [0.0, 168.21, 274.15, 255.34], \"box_3d\": [1.49, 1.67, 4.5, -9.39, 1.42, 14.4], \"ry\": 2.43}, {\"category\": \"Car\", \"angle\": -1.1, \"box_2d\": [1023.12, 152.52, 1237.84, 256.67], \"box_3d\": [1.65, 1.67, 3.64, 9.36, 1.33, 13.0], \"ry\": -0.48}, {\"category\": \"Car\", \"angle\": 2.21, \"box_2d\": [1161.66, 156.97, 1241.0, 207.71], \"box_3d\": [1.59, 1.55, 3.93, 20.59, 1.09, 23.9], \"ry\": 2.91}]\n```" - } - ] - }, - { - "id": 355, - "data_source": "Mono3DRefer", - "scene_name": "000849", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.22, \"box_2d\": [788.53, 189.93, 1137.59, 374.0], \"box_3d\": [1.49, 1.61, 3.74, 2.91, 1.68, 6.97], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [589.72, 173.63, 630.7, 206.64], \"box_3d\": [1.46, 1.63, 4.47, 0.04, 1.53, 34.58], \"ry\": 1.64}, {\"category\": \"Car\", \"angle\": -2.92, \"box_2d\": [373.7, 173.33, 526.7, 225.22], \"box_3d\": [1.47, 1.68, 4.26, -4.8, 1.52, 21.89], \"ry\": -3.14}]\n```" - } - ] - }, - { - "id": 356, - "data_source": "Mono3DRefer", - "scene_name": "000851", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [533.07, 181.1, 581.41, 220.64], \"box_3d\": [1.53, 1.59, 3.64, -2.3, 1.89, 30.21], \"ry\": -1.76}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [433.9, 176.63, 468.02, 205.31], \"box_3d\": [1.63, 1.6, 3.73, -9.52, 1.87, 43.29], \"ry\": 1.46}]\n```" - } - ] - }, - { - "id": 357, - "data_source": "Mono3DRefer", - "scene_name": "000854", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [236.74, 186.46, 321.18, 226.99], \"box_3d\": [1.48, 1.71, 3.54, -13.22, 2.05, 28.95], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [412.68, 178.85, 448.08, 197.34], \"box_3d\": [1.56, 1.91, 4.44, -15.85, 2.1, 63.82], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 358, - "data_source": "Mono3DRefer", - "scene_name": "000856", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.91, \"box_2d\": [756.9, 202.27, 1114.22, 374.0], \"box_3d\": [1.48, 1.6, 3.79, 2.49, 1.82, 6.65], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [459.12, 189.11, 552.13, 264.02], \"box_3d\": [1.56, 1.55, 3.64, -2.44, 1.98, 17.36], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [514.91, 174.08, 584.22, 240.85], \"box_3d\": [1.91, 1.66, 4.22, -1.85, 1.97, 22.86], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [557.8, 178.84, 612.8, 228.56], \"box_3d\": [1.83, 1.69, 4.28, -0.94, 2.08, 28.82], \"ry\": 1.65}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [621.61, 185.69, 663.66, 212.3], \"box_3d\": [1.51, 1.63, 4.03, 2.03, 2.33, 44.23], \"ry\": 1.86}]\n```" - } - ] - }, - { - "id": 359, - "data_source": "Mono3DRefer", - "scene_name": "000858", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.11, \"box_2d\": [49.58, 201.59, 198.11, 265.08], \"box_3d\": [1.35, 1.5, 3.64, -12.39, 2.15, 18.6], \"ry\": 1.53}]\n```" - } - ] - }, - { - "id": 360, - "data_source": "Mono3DRefer", - "scene_name": "000859", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [791.11, 177.42, 961.57, 298.1], \"box_3d\": [1.57, 1.66, 4.02, 3.96, 1.66, 11.51], \"ry\": -1.45}, {\"category\": \"Car\", \"angle\": 1.42, \"box_2d\": [737.92, 178.5, 832.02, 249.27], \"box_3d\": [1.53, 1.62, 3.77, 4.1, 1.68, 17.59], \"ry\": 1.65}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [585.08, 174.08, 634.32, 207.87], \"box_3d\": [1.76, 1.77, 4.41, 0.02, 1.85, 39.89], \"ry\": 1.8}, {\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [622.53, 178.6, 661.32, 202.34], \"box_3d\": [1.48, 1.65, 4.4, 2.12, 1.86, 47.33], \"ry\": 1.82}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [682.01, 176.83, 708.49, 193.76], \"box_3d\": [1.42, 1.66, 3.64, 7.42, 1.78, 62.68], \"ry\": 1.87}]\n```" - } - ] - }, - { - "id": 361, - "data_source": "Mono3DRefer", - "scene_name": "000865", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [203.04, 201.88, 468.2, 374.0], \"box_3d\": [1.6, 1.68, 3.94, -3.04, 2.02, 8.8], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [535.61, 197.0, 573.54, 227.99], \"box_3d\": [1.42, 1.66, 4.11, -2.88, 2.74, 37.75], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 362, - "data_source": "Mono3DRefer", - "scene_name": "000870", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [601.64, 181.69, 678.8, 244.68], \"box_3d\": [1.36, 1.69, 3.38, 0.62, 1.59, 17.59], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [337.87, 181.1, 401.28, 216.61], \"box_3d\": [1.59, 1.72, 3.86, -11.49, 2.0, 34.65], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -2.31, \"box_2d\": [475.68, 178.03, 550.03, 211.62], \"box_3d\": [1.59, 1.63, 3.64, -4.91, 1.87, 36.05], \"ry\": -2.45}, {\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [583.31, 181.52, 627.65, 222.14], \"box_3d\": [1.52, 1.67, 3.61, -0.17, 1.88, 29.14], \"ry\": -1.55}]\n```" - } - ] - }, - { - "id": 363, - "data_source": "Mono3DRefer", - "scene_name": "000875", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.57, \"box_2d\": [0.0, 179.69, 205.93, 293.37], \"box_3d\": [1.49, 1.75, 3.73, -8.27, 1.61, 10.46], \"ry\": 3.06}, {\"category\": \"Car\", \"angle\": -2.68, \"box_2d\": [16.18, 180.31, 269.69, 267.44], \"box_3d\": [1.49, 1.64, 3.69, -8.73, 1.65, 13.68], \"ry\": 3.04}, {\"category\": \"Car\", \"angle\": 0.27, \"box_2d\": [114.26, 167.93, 315.1, 251.0], \"box_3d\": [1.77, 1.6, 3.79, -9.06, 1.68, 16.74], \"ry\": -0.22}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [511.86, 177.33, 567.47, 218.88], \"box_3d\": [1.66, 1.66, 3.98, -3.14, 1.87, 31.17], \"ry\": -1.86}]\n```" - } - ] - }, - { - "id": 364, - "data_source": "Mono3DRefer", - "scene_name": "000877", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [369.82, 187.81, 431.48, 225.1], \"box_3d\": [1.56, 1.73, 4.37, -9.7, 2.29, 33.64], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [434.33, 183.67, 469.24, 209.82], \"box_3d\": [1.6, 1.67, 2.65, -10.14, 2.31, 46.27], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 365, - "data_source": "Mono3DRefer", - "scene_name": "000878", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [376.73, 183.14, 429.04, 211.56], \"box_3d\": [1.62, 1.86, 4.31, -12.61, 2.26, 44.11], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 366, - "data_source": "Mono3DRefer", - "scene_name": "000879", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [369.4, 185.46, 420.32, 210.43], \"box_3d\": [1.5, 1.95, 4.6, -13.99, 2.35, 47.1], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 367, - "data_source": "Mono3DRefer", - "scene_name": "000881", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [578.34, 182.19, 605.49, 209.17], \"box_3d\": [1.65, 1.67, 3.64, -1.19, 2.27, 46.64], \"ry\": -1.59}]\n```" - } - ] - }, - { - "id": 368, - "data_source": "Mono3DRefer", - "scene_name": "000884", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [454.15, 182.71, 548.63, 254.64], \"box_3d\": [1.55, 1.68, 4.42, -2.62, 1.81, 18.07], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [669.56, 182.91, 731.03, 232.74], \"box_3d\": [1.6, 1.53, 4.18, 3.09, 1.97, 25.65], \"ry\": -1.6}]\n```" - } - ] - }, - { - "id": 369, - "data_source": "Mono3DRefer", - "scene_name": "000885", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [430.33, 182.04, 532.74, 261.28], \"box_3d\": [1.52, 1.57, 4.42, -2.78, 1.74, 16.27], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.35, \"box_2d\": [692.66, 184.52, 844.71, 294.35], \"box_3d\": [1.51, 1.58, 3.92, 2.42, 1.72, 12.14], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.38, \"box_2d\": [688.22, 180.36, 775.62, 250.85], \"box_3d\": [1.55, 1.47, 3.36, 2.85, 1.74, 17.72], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [524.85, 182.32, 565.09, 218.24], \"box_3d\": [1.46, 1.6, 3.33, -2.83, 1.88, 31.5], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [646.98, 177.32, 692.12, 216.89], \"box_3d\": [1.64, 1.7, 3.54, 2.55, 1.86, 32.0], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 370, - "data_source": "Mono3DRefer", - "scene_name": "000887", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.11, \"box_2d\": [2.75, 196.29, 83.42, 230.2], \"box_3d\": [1.43, 1.48, 3.33, -26.68, 2.58, 34.03], \"ry\": 1.45}, {\"category\": \"Car\", \"angle\": 2.19, \"box_2d\": [83.34, 190.17, 163.91, 222.12], \"box_3d\": [1.62, 1.6, 4.18, -27.08, 2.62, 40.3], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 2.09, \"box_2d\": [148.84, 191.01, 205.29, 215.97], \"box_3d\": [1.48, 1.35, 3.93, -28.07, 2.7, 46.86], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.99, \"box_2d\": [184.01, 187.29, 234.58, 210.12], \"box_3d\": [1.5, 1.62, 4.08, -28.38, 2.55, 51.19], \"ry\": 1.48}, {\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [223.46, 189.41, 263.85, 211.75], \"box_3d\": [1.65, 1.57, 3.82, -29.02, 2.99, 57.22], \"ry\": 1.47}]\n```" - } - ] - }, - { - "id": 371, - "data_source": "Mono3DRefer", - "scene_name": "000890", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.85, \"box_2d\": [730.12, 186.87, 877.86, 284.24], \"box_3d\": [1.38, 1.49, 3.32, 3.08, 1.64, 12.22], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.83, \"box_2d\": [735.88, 176.5, 785.41, 208.7], \"box_3d\": [1.43, 1.48, 3.25, 7.03, 1.62, 34.1], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": -1.98, \"box_2d\": [813.98, 178.61, 925.48, 233.92], \"box_3d\": [1.44, 1.56, 3.96, 7.38, 1.62, 21.02], \"ry\": -1.65}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [650.72, 177.24, 686.94, 206.17], \"box_3d\": [1.42, 1.45, 3.2, 2.98, 1.66, 37.33], \"ry\": -1.62}]\n```" - } - ] - }, - { - "id": 372, - "data_source": "Mono3DRefer", - "scene_name": "000893", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.1, \"box_2d\": [113.18, 192.89, 246.14, 248.37], \"box_3d\": [1.57, 1.9, 4.16, -14.02, 2.28, 23.76], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 373, - "data_source": "Mono3DRefer", - "scene_name": "000895", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.59, \"box_2d\": [494.56, 181.99, 548.2, 229.92], \"box_3d\": [1.41, 1.59, 4.47, -2.94, 1.73, 23.92], \"ry\": 1.47}]\n```" - } - ] - }, - { - "id": 374, - "data_source": "Mono3DRefer", - "scene_name": "000901", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.58, \"box_2d\": [589.84, 190.33, 626.7, 224.66], \"box_3d\": [1.41, 1.59, 4.47, -0.11, 2.27, 33.61], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 375, - "data_source": "Mono3DRefer", - "scene_name": "000905", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.57, \"box_2d\": [411.68, 193.66, 453.77, 233.86], \"box_3d\": [1.65, 1.75, 3.39, -8.09, 2.63, 32.82], \"ry\": 1.33}]\n```" - } - ] - }, - { - "id": 376, - "data_source": "Mono3DRefer", - "scene_name": "000920", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [204.69, 192.16, 430.14, 333.28], \"box_3d\": [1.45, 1.64, 4.21, -3.77, 1.76, 10.0], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 1.41, \"box_2d\": [647.31, 180.76, 726.02, 243.96], \"box_3d\": [1.62, 1.58, 3.89, 2.05, 1.85, 20.62], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [469.15, 184.41, 530.51, 227.54], \"box_3d\": [1.45, 1.67, 4.43, -4.09, 1.91, 27.19], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [630.54, 183.61, 687.77, 225.83], \"box_3d\": [1.47, 1.59, 3.62, 1.77, 1.89, 27.45], \"ry\": -1.67}]\n```" - } - ] - }, - { - "id": 377, - "data_source": "Mono3DRefer", - "scene_name": "000921", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [540.42, 180.79, 568.75, 207.31], \"box_3d\": [1.7, 1.53, 3.82, -3.73, 2.25, 48.76], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [469.26, 175.15, 541.73, 239.43], \"box_3d\": [1.97, 1.87, 4.53, -3.49, 2.08, 24.62], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 378, - "data_source": "Mono3DRefer", - "scene_name": "000930", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.96, \"box_2d\": [74.42, 159.2, 440.97, 374.0], \"box_3d\": [1.7, 1.63, 4.08, -3.01, 1.64, 7.12], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.35, \"box_2d\": [402.09, 186.56, 523.35, 266.07], \"box_3d\": [1.4, 1.67, 3.72, -2.96, 1.71, 15.01], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": 1.47, \"box_2d\": [636.4, 180.08, 681.86, 216.2], \"box_3d\": [1.49, 1.61, 3.74, 2.08, 1.82, 31.91], \"ry\": 1.53}]\n```" - } - ] - }, - { - "id": 379, - "data_source": "Mono3DRefer", - "scene_name": "000933", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [481.18, 184.27, 541.92, 227.32], \"box_3d\": [1.44, 1.72, 4.14, -3.63, 1.89, 26.95], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 380, - "data_source": "Mono3DRefer", - "scene_name": "000941", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.1, \"box_2d\": [0.0, 187.55, 425.15, 374.0], \"box_3d\": [1.77, 1.75, 4.67, -3.17, 1.93, 6.27], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [432.73, 183.27, 536.13, 260.78], \"box_3d\": [1.6, 1.68, 3.94, -2.88, 1.86, 17.12], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [558.58, 183.21, 588.78, 207.04], \"box_3d\": [1.42, 1.66, 4.11, -2.3, 2.1, 46.1], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 381, - "data_source": "Mono3DRefer", - "scene_name": "000945", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [363.97, 182.25, 518.82, 301.31], \"box_3d\": [1.63, 1.6, 3.73, -2.62, 1.8, 11.88], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -2.04, \"box_2d\": [875.51, 219.55, 1241.0, 374.0], \"box_3d\": [1.45, 1.5, 3.26, 2.57, 1.79, 3.78], \"ry\": -1.47}, {\"category\": \"Car\", \"angle\": -1.84, \"box_2d\": [721.52, 186.08, 904.77, 327.88], \"box_3d\": [1.59, 1.49, 3.34, 2.56, 1.8, 9.99], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [670.11, 189.77, 777.05, 274.71], \"box_3d\": [1.54, 1.6, 3.52, 2.25, 1.93, 15.36], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.42, \"box_2d\": [449.02, 187.12, 535.12, 261.03], \"box_3d\": [1.54, 1.5, 3.48, -2.76, 1.91, 17.24], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [479.73, 186.94, 547.47, 238.96], \"box_3d\": [1.44, 1.63, 3.32, -2.95, 1.9, 22.34], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [654.31, 184.86, 726.06, 240.73], \"box_3d\": [1.55, 1.55, 3.27, 2.35, 1.93, 22.14], \"ry\": -1.66}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [636.74, 186.69, 691.2, 230.76], \"box_3d\": [1.48, 1.6, 3.79, 1.91, 2.03, 27.0], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [524.87, 185.58, 563.02, 217.95], \"box_3d\": [1.56, 1.55, 3.64, -3.41, 2.24, 37.46], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [547.18, 177.05, 580.74, 210.91], \"box_3d\": [1.91, 1.66, 4.22, -2.72, 2.17, 43.08], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [560.37, 179.67, 591.66, 208.01], \"box_3d\": [1.83, 1.69, 4.28, -2.28, 2.3, 49.02], \"ry\": 1.62}]\n```" - } - ] - }, - { - "id": 382, - "data_source": "Mono3DRefer", - "scene_name": "000950", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.53, \"box_2d\": [540.2, 186.06, 568.58, 214.59], \"box_3d\": [1.68, 1.63, 3.9, -3.58, 2.54, 45.71], \"ry\": 1.46}]\n```" - } - ] - }, - { - "id": 383, - "data_source": "Mono3DRefer", - "scene_name": "000951", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.19, \"box_2d\": [159.13, 186.76, 316.12, 260.55], \"box_3d\": [1.48, 1.71, 4.65, -8.75, 1.84, 17.37], \"ry\": -1.65}, {\"category\": \"Car\", \"angle\": -1.0, \"box_2d\": [0.12, 165.41, 200.99, 264.21], \"box_3d\": [2.04, 1.66, 4.62, -11.97, 1.92, 17.34], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.44, \"box_2d\": [404.21, 171.04, 470.78, 217.94], \"box_3d\": [1.42, 1.64, 3.95, -5.72, 1.4, 24.15], \"ry\": -1.67}, {\"category\": \"Car\", \"angle\": -1.9, \"box_2d\": [729.04, 171.75, 852.09, 236.91], \"box_3d\": [1.51, 1.72, 4.5, 4.58, 1.52, 19.2], \"ry\": -1.67}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [572.27, 161.95, 611.42, 198.88], \"box_3d\": [1.7, 1.8, 4.5, -0.95, 1.22, 35.74], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.96, \"box_2d\": [805.75, 162.56, 936.29, 230.6], \"box_3d\": [1.64, 1.73, 4.43, 6.94, 1.42, 19.82], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": -2.08, \"box_2d\": [892.49, 171.92, 1049.68, 229.86], \"box_3d\": [1.38, 1.8, 4.49, 9.6, 1.39, 19.74], \"ry\": -1.64}]\n```" - } - ] - }, - { - "id": 384, - "data_source": "Mono3DRefer", - "scene_name": "000952", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.28, \"box_2d\": [1052.37, 179.61, 1209.11, 263.49], \"box_3d\": [1.47, 1.6, 3.66, 10.56, 1.62, 14.61], \"ry\": -0.66}, {\"category\": \"Car\", \"angle\": -0.86, \"box_2d\": [577.72, 168.44, 893.86, 359.62], \"box_3d\": [1.6, 1.57, 3.23, 1.38, 1.58, 7.68], \"ry\": -0.69}, {\"category\": \"Car\", \"angle\": 2.02, \"box_2d\": [839.56, 167.77, 1067.47, 299.57], \"box_3d\": [1.57, 1.5, 3.68, 4.97, 1.52, 10.33], \"ry\": 2.46}]\n```" - } - ] - }, - { - "id": 385, - "data_source": "Mono3DRefer", - "scene_name": "000954", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.91, \"box_2d\": [762.07, 159.9, 989.43, 312.81], \"box_3d\": [1.64, 1.69, 3.3, 3.19, 1.53, 9.49], \"ry\": -1.59}]\n```" - } - ] - }, - { - "id": 386, - "data_source": "Mono3DRefer", - "scene_name": "000955", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.36, \"box_2d\": [235.76, 183.17, 390.22, 241.04], \"box_3d\": [1.52, 1.64, 3.7, -8.36, 1.84, 20.43], \"ry\": -0.03}, {\"category\": \"Car\", \"angle\": -2.73, \"box_2d\": [307.68, 183.82, 428.31, 228.58], \"box_3d\": [1.5, 1.62, 3.69, -8.54, 1.91, 25.49], \"ry\": -3.05}, {\"category\": \"Car\", \"angle\": 1.44, \"box_2d\": [885.78, 162.78, 921.38, 185.62], \"box_3d\": [1.51, 1.7, 4.45, 20.46, 0.86, 50.46], \"ry\": 1.82}]\n```" - } - ] - }, - { - "id": 387, - "data_source": "Mono3DRefer", - "scene_name": "000960", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [551.18, 178.6, 576.02, 197.25], \"box_3d\": [1.45, 1.77, 4.04, -3.77, 1.94, 58.91], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 388, - "data_source": "Mono3DRefer", - "scene_name": "000962", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [636.56, 175.95, 663.41, 198.34], \"box_3d\": [1.42, 1.45, 3.2, 2.6, 1.64, 47.77], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [684.67, 182.62, 764.91, 240.58], \"box_3d\": [1.38, 1.49, 3.32, 2.91, 1.65, 19.14], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -2.05, \"box_2d\": [829.38, 185.73, 1238.19, 374.0], \"box_3d\": [1.5, 1.62, 3.89, 3.58, 1.65, 7.08], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [701.33, 175.12, 737.13, 200.1], \"box_3d\": [1.43, 1.48, 3.25, 6.52, 1.59, 43.48], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": -1.88, \"box_2d\": [751.8, 176.57, 819.57, 214.29], \"box_3d\": [1.44, 1.56, 3.96, 7.15, 1.61, 29.87], \"ry\": -1.65}]\n```" - } - ] - }, - { - "id": 389, - "data_source": "Mono3DRefer", - "scene_name": "000968", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [592.53, 176.62, 628.22, 200.99], \"box_3d\": [1.44, 1.68, 4.39, 0.05, 1.69, 45.14], \"ry\": 1.7}]\n```" - } - ] - }, - { - "id": 390, - "data_source": "Mono3DRefer", - "scene_name": "000973", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.94, \"box_2d\": [746.37, 192.28, 1016.45, 356.87], \"box_3d\": [1.52, 1.54, 4.5, 3.06, 1.82, 9.35], \"ry\": -1.63}]\n```" - } - ] - }, - { - "id": 391, - "data_source": "Mono3DRefer", - "scene_name": "000974", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [600.07, 176.31, 629.48, 204.63], \"box_3d\": [1.66, 1.73, 3.05, 0.26, 1.89, 44.36], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [356.61, 184.82, 406.28, 212.61], \"box_3d\": [1.44, 1.63, 3.43, -12.7, 2.13, 40.17], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 392, - "data_source": "Mono3DRefer", - "scene_name": "000976", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [400.47, 183.04, 525.97, 278.57], \"box_3d\": [1.46, 1.63, 4.47, -2.61, 1.67, 13.57], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [502.14, 169.74, 576.7, 224.18], \"box_3d\": [1.7, 1.82, 4.45, -2.36, 1.63, 24.91], \"ry\": 1.65}, {\"category\": \"Car\", \"angle\": -1.05, \"box_2d\": [637.33, 171.85, 686.64, 201.83], \"box_3d\": [1.57, 1.56, 2.62, 2.83, 1.53, 39.36], \"ry\": -0.97}]\n```" - } - ] - }, - { - "id": 393, - "data_source": "Mono3DRefer", - "scene_name": "000977", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.2, \"box_2d\": [0.0, 202.07, 257.92, 374.0], \"box_3d\": [1.58, 1.76, 4.18, -4.4, 1.86, 5.18], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": -0.65, \"box_2d\": [1072.59, 111.3, 1241.0, 225.67], \"box_3d\": [1.9, 1.78, 4.72, 11.22, 0.87, 12.9], \"ry\": 0.05}, {\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [379.22, 189.75, 492.79, 260.4], \"box_3d\": [1.42, 1.67, 3.8, -4.02, 1.86, 17.11], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [209.32, 188.44, 436.01, 317.25], \"box_3d\": [1.55, 1.71, 4.5, -4.21, 1.83, 11.32], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.37, \"box_2d\": [733.72, 170.63, 819.1, 224.14], \"box_3d\": [1.5, 1.71, 4.44, 5.05, 1.47, 22.63], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [453.38, 182.82, 528.33, 235.57], \"box_3d\": [1.54, 1.63, 3.59, -3.8, 1.88, 23.29], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [504.67, 184.58, 552.34, 222.41], \"box_3d\": [1.42, 1.46, 3.63, -3.32, 1.92, 29.57], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [517.73, 184.08, 560.68, 216.71], \"box_3d\": [1.43, 1.66, 4.19, -3.37, 1.99, 34.66], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.48, \"box_2d\": [645.72, 178.52, 689.2, 214.05], \"box_3d\": [1.58, 1.71, 3.75, 2.74, 1.86, 34.17], \"ry\": -1.4}, {\"category\": \"Car\", \"angle\": -0.15, \"box_2d\": [725.27, 170.78, 849.39, 213.34], \"box_3d\": [1.49, 1.57, 4.26, 6.46, 1.45, 26.76], \"ry\": 0.08}, {\"category\": \"Car\", \"angle\": -0.99, \"box_2d\": [687.44, 176.24, 740.44, 205.07], \"box_3d\": [1.61, 1.6, 3.11, 6.08, 1.82, 42.18], \"ry\": -0.84}]\n```" - } - ] - }, - { - "id": 394, - "data_source": "Mono3DRefer", - "scene_name": "000978", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [439.76, 194.95, 517.78, 270.88], \"box_3d\": [1.65, 1.75, 3.39, -3.28, 2.25, 18.27], \"ry\": 1.43}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [476.47, 194.86, 535.3, 245.55], \"box_3d\": [1.57, 1.54, 4.01, -3.67, 2.4, 25.72], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 395, - "data_source": "Mono3DRefer", - "scene_name": "000983", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [715.8, 190.09, 860.92, 299.6], \"box_3d\": [1.37, 1.59, 3.22, 2.51, 1.66, 10.99], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": -1.4, \"box_2d\": [507.7, 181.68, 570.26, 223.46], \"box_3d\": [1.4, 1.67, 3.72, -2.56, 1.74, 26.46], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": 1.51, \"box_2d\": [642.75, 178.07, 673.96, 204.16], \"box_3d\": [1.49, 1.61, 3.74, 2.84, 1.81, 43.42], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -2.62, \"box_2d\": [26.26, 168.08, 342.31, 270.94], \"box_3d\": [1.59, 1.63, 4.43, -7.13, 1.55, 12.29], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [448.21, 171.65, 541.42, 246.52], \"box_3d\": [1.7, 1.63, 4.08, -2.88, 1.7, 18.58], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": -1.91, \"box_2d\": [776.6, 184.16, 1088.58, 374.0], \"box_3d\": [1.59, 1.59, 2.47, 2.45, 1.7, 6.17], \"ry\": -1.55}]\n```" - } - ] - }, - { - "id": 396, - "data_source": "Mono3DRefer", - "scene_name": "000989", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [424.49, 177.87, 501.36, 222.53], \"box_3d\": [1.42, 1.78, 3.62, -4.73, 1.31, 24.93], \"ry\": 1.62}]\n```" - } - ] - }, - { - "id": 397, - "data_source": "Mono3DRefer", - "scene_name": "000993", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.32, \"box_2d\": [757.58, 174.01, 829.51, 202.22], \"box_3d\": [1.62, 1.72, 4.35, 11.09, 1.72, 43.98], \"ry\": -2.07}]\n```" - } - ] - }, - { - "id": 398, - "data_source": "Mono3DRefer", - "scene_name": "001000", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.69, \"box_2d\": [798.93, 169.95, 1060.55, 250.42], \"box_3d\": [1.47, 1.77, 4.49, 6.27, 1.45, 14.55], \"ry\": 3.09}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [561.02, 184.64, 587.94, 208.4], \"box_3d\": [1.52, 1.62, 4.5, -2.5, 2.35, 49.56], \"ry\": -1.67}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [468.53, 187.25, 501.75, 210.97], \"box_3d\": [1.42, 1.66, 4.5, -8.15, 2.39, 47.26], \"ry\": 1.51}]\n```" - } - ] - }, - { - "id": 399, - "data_source": "Mono3DRefer", - "scene_name": "001001", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.56, \"box_2d\": [400.58, 191.83, 441.15, 230.23], \"box_3d\": [1.65, 1.75, 3.39, -8.99, 2.58, 34.09], \"ry\": 1.3}]\n```" - } - ] - }, - { - "id": 400, - "data_source": "Mono3DRefer", - "scene_name": "001002", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [607.09, 174.99, 647.58, 215.65], \"box_3d\": [1.57, 1.56, 2.62, 0.68, 1.68, 29.48], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [391.06, 187.26, 506.26, 256.04], \"box_3d\": [1.43, 1.68, 4.19, -3.82, 1.81, 17.63], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [467.49, 180.33, 530.42, 228.14], \"box_3d\": [1.46, 1.59, 4.27, -3.71, 1.72, 24.45], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [511.35, 178.6, 554.75, 207.95], \"box_3d\": [1.38, 1.62, 4.06, -3.85, 1.68, 36.35], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [541.3, 172.58, 569.78, 194.45], \"box_3d\": [1.49, 1.53, 3.88, -3.88, 1.5, 51.64], \"ry\": 1.63}]\n```" - } - ] - }, - { - "id": 401, - "data_source": "Mono3DRefer", - "scene_name": "001003", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.75, \"box_2d\": [0.0, 206.78, 324.17, 374.0], \"box_3d\": [1.38, 1.66, 4.38, -5.02, 1.72, 6.21], \"ry\": 0.09}, {\"category\": \"Car\", \"angle\": -0.73, \"box_2d\": [926.19, 141.69, 1241.0, 374.0], \"box_3d\": [1.74, 1.66, 4.37, 4.71, 1.56, 5.01], \"ry\": -0.0}, {\"category\": \"Car\", \"angle\": -2.74, \"box_2d\": [273.5, 182.7, 480.59, 250.25], \"box_3d\": [1.47, 1.67, 4.28, -5.39, 1.72, 16.83], \"ry\": -3.05}, {\"category\": \"Car\", \"angle\": -3.01, \"box_2d\": [362.85, 181.79, 516.48, 230.7], \"box_3d\": [1.38, 1.67, 4.37, -5.13, 1.67, 22.04], \"ry\": 3.05}, {\"category\": \"Car\", \"angle\": -0.4, \"box_2d\": [823.68, 170.37, 1060.18, 271.26], \"box_3d\": [1.54, 1.58, 3.2, 5.36, 1.53, 12.03], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": -0.5, \"box_2d\": [852.36, 170.37, 1231.56, 296.98], \"box_3d\": [1.51, 1.72, 4.03, 5.52, 1.51, 9.68], \"ry\": 0.01}]\n```" - } - ] - }, - { - "id": 402, - "data_source": "Mono3DRefer", - "scene_name": "001005", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [709.89, 170.92, 835.41, 263.18], \"box_3d\": [1.64, 1.69, 3.3, 3.12, 1.63, 14.64], \"ry\": -1.6}]\n```" - } - ] - }, - { - "id": 403, - "data_source": "Mono3DRefer", - "scene_name": "001008", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [610.08, 162.36, 649.92, 195.49], \"box_3d\": [1.49, 1.68, 4.08, 1.43, 0.62, 34.68], \"ry\": 1.66}]\n```" - } - ] - }, - { - "id": 404, - "data_source": "Mono3DRefer", - "scene_name": "001009", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.45, \"box_2d\": [525.22, 169.11, 574.76, 207.01], \"box_3d\": [1.47, 1.59, 4.03, -2.1, 0.99, 30.15], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": -1.49, \"box_2d\": [555.11, 170.81, 587.12, 196.89], \"box_3d\": [1.41, 1.54, 3.36, -1.68, 0.83, 40.86], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [609.38, 165.49, 640.33, 193.27], \"box_3d\": [1.52, 1.67, 4.38, 1.43, 0.64, 42.05], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [398.33, 172.98, 477.3, 222.57], \"box_3d\": [1.44, 1.55, 3.93, -5.15, 1.19, 23.02], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -1.52, \"box_2d\": [581.42, 164.08, 605.69, 186.52], \"box_3d\": [1.56, 1.59, 3.65, -0.52, 0.34, 52.4], \"ry\": -1.54}]\n```" - } - ] - }, - { - "id": 405, - "data_source": "Mono3DRefer", - "scene_name": "001012", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.04, \"box_2d\": [0.0, 184.27, 361.81, 374.0], \"box_3d\": [1.52, 1.61, 3.59, -3.47, 1.63, 5.96], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.17, \"box_2d\": [755.48, 188.1, 1191.01, 374.0], \"box_3d\": [1.51, 1.65, 4.28, 2.62, 1.68, 6.55], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [376.92, 178.87, 498.99, 260.59], \"box_3d\": [1.48, 1.64, 3.95, -3.49, 1.61, 15.16], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [682.87, 187.19, 791.28, 261.44], \"box_3d\": [1.35, 1.64, 3.84, 2.56, 1.69, 15.57], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.47, \"box_2d\": [654.95, 182.52, 711.84, 226.32], \"box_3d\": [1.42, 1.59, 4.19, 2.52, 1.78, 25.93], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [481.55, 179.45, 536.88, 221.67], \"box_3d\": [1.42, 1.64, 3.63, -3.65, 1.67, 26.36], \"ry\": 1.53}]\n```" - } - ] - }, - { - "id": 406, - "data_source": "Mono3DRefer", - "scene_name": "001018", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.84, \"box_2d\": [716.81, 184.62, 907.47, 300.32], \"box_3d\": [1.4, 1.61, 4.37, 2.81, 1.6, 11.2], \"ry\": -1.6}]\n```" - } - ] - }, - { - "id": 407, - "data_source": "Mono3DRefer", - "scene_name": "001020", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.17, \"box_2d\": [3.77, 189.5, 446.2, 373.0], \"box_3d\": [1.47, 1.59, 4.03, -2.61, 1.57, 6.08], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.38, \"box_2d\": [421.39, 178.87, 521.95, 251.31], \"box_3d\": [1.41, 1.54, 3.36, -2.76, 1.36, 15.77], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [580.76, 170.42, 636.21, 221.0], \"box_3d\": [1.52, 1.67, 4.38, 0.23, 1.19, 24.07], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.49, \"box_2d\": [534.79, 169.21, 578.81, 207.77], \"box_3d\": [1.56, 1.59, 3.65, -1.89, 1.06, 31.24], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [495.91, 173.14, 531.34, 194.5], \"box_3d\": [1.34, 1.76, 3.98, -5.78, 0.81, 47.8], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 408, - "data_source": "Mono3DRefer", - "scene_name": "001021", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [384.01, 183.15, 468.79, 246.74], \"box_3d\": [1.41, 1.57, 3.53, -4.56, 1.69, 18.18], \"ry\": 1.47}, {\"category\": \"Car\", \"angle\": 2.59, \"box_2d\": [503.59, 176.14, 582.13, 203.3], \"box_3d\": [1.47, 1.72, 4.22, -3.87, 1.68, 41.58], \"ry\": 2.5}, {\"category\": \"Car\", \"angle\": 1.11, \"box_2d\": [343.9, 182.9, 434.63, 222.63], \"box_3d\": [1.42, 1.71, 4.14, -8.74, 1.84, 28.25], \"ry\": 0.81}, {\"category\": \"Car\", \"angle\": 2.68, \"box_2d\": [620.38, 173.53, 701.9, 198.66], \"box_3d\": [1.48, 1.87, 4.74, 3.18, 1.55, 45.14], \"ry\": 2.75}]\n```" - } - ] - }, - { - "id": 409, - "data_source": "Mono3DRefer", - "scene_name": "001024", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.19, \"box_2d\": [0.0, 193.19, 36.95, 231.64], \"box_3d\": [1.42, 1.68, 3.38, -24.01, 2.24, 27.55], \"ry\": 1.48}, {\"category\": \"Car\", \"angle\": 2.19, \"box_2d\": [19.57, 191.48, 118.1, 225.78], \"box_3d\": [1.38, 1.52, 4.0, -24.38, 2.26, 32.63], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 2.08, \"box_2d\": [108.31, 186.17, 174.1, 218.0], \"box_3d\": [1.57, 1.47, 3.35, -24.86, 2.3, 38.33], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [161.19, 188.01, 214.59, 213.72], \"box_3d\": [1.43, 1.48, 3.33, -25.25, 2.36, 43.22], \"ry\": 1.48}, {\"category\": \"Car\", \"angle\": 2.0, \"box_2d\": [205.4, 186.39, 258.46, 211.98], \"box_3d\": [1.62, 1.6, 4.18, -25.72, 2.57, 49.18], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": 2.02, \"box_2d\": [249.91, 183.27, 291.74, 203.43], \"box_3d\": [1.48, 1.35, 3.93, -26.3, 2.31, 56.03], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 410, - "data_source": "Mono3DRefer", - "scene_name": "001025", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [640.98, 186.87, 666.75, 207.47], \"box_3d\": [1.41, 1.58, 4.36, 3.19, 2.47, 53.45], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [190.27, 172.08, 431.56, 317.88], \"box_3d\": [1.5, 1.62, 4.19, -3.69, 1.52, 9.63], \"ry\": 1.53}]\n```" - } - ] - }, - { - "id": 411, - "data_source": "Mono3DRefer", - "scene_name": "001026", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [507.51, 179.15, 576.33, 227.78], \"box_3d\": [1.45, 1.62, 3.83, -2.18, 1.67, 23.62], \"ry\": 1.64}, {\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [585.47, 174.72, 618.38, 202.34], \"box_3d\": [1.43, 1.66, 3.67, -0.43, 1.56, 39.73], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 412, - "data_source": "Mono3DRefer", - "scene_name": "001028", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (707.05, 707.05), (cx, cy) = (604.08, 180.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.05, \"box_2d\": [843.99, 192.18, 1223.0, 369.0], \"box_3d\": [1.4, 1.5, 3.72, 3.39, 1.55, 6.08], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.88, \"box_2d\": [769.07, 179.09, 959.56, 291.29], \"box_3d\": [1.52, 1.64, 4.41, 4.08, 1.52, 12.02], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 2.05, \"box_2d\": [292.4, 170.21, 381.49, 218.81], \"box_3d\": [1.85, 1.68, 4.11, -10.96, 1.46, 29.2], \"ry\": 1.7}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [516.58, 178.02, 547.04, 196.29], \"box_3d\": [1.49, 1.58, 3.84, -6.24, 1.3, 60.89], \"ry\": 1.75}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [699.81, 177.29, 751.02, 226.74], \"box_3d\": [1.54, 1.57, 3.52, 4.05, 1.45, 24.03], \"ry\": -1.41}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [693.18, 180.39, 742.55, 220.34], \"box_3d\": [1.5, 1.63, 3.57, 4.55, 1.52, 29.0], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": -1.61, \"box_2d\": [684.2, 180.39, 719.96, 212.63], \"box_3d\": [1.47, 1.59, 3.64, 4.75, 1.49, 34.97], \"ry\": -1.47}, {\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [669.55, 180.1, 692.98, 201.07], \"box_3d\": [1.44, 1.58, 3.28, 5.55, 1.44, 51.56], \"ry\": -1.49}]\n```" - } - ] - }, - { - "id": 413, - "data_source": "Mono3DRefer", - "scene_name": "001030", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [342.69, 183.4, 455.01, 246.28], \"box_3d\": [1.51, 1.81, 4.07, -5.68, 1.82, 19.81], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 414, - "data_source": "Mono3DRefer", - "scene_name": "001032", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [642.26, 171.96, 681.05, 205.41], \"box_3d\": [1.69, 1.58, 3.95, 2.7, 1.68, 38.83], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [666.55, 178.57, 716.49, 215.93], \"box_3d\": [1.41, 1.57, 3.23, 3.2, 1.65, 29.15], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.86, \"box_2d\": [760.2, 176.26, 833.85, 216.83], \"box_3d\": [1.45, 1.61, 4.2, 7.17, 1.6, 28.19], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -2.45, \"box_2d\": [801.12, 175.06, 867.43, 202.44], \"box_3d\": [1.48, 1.36, 3.51, 12.66, 1.63, 40.98], \"ry\": -2.15}]\n```" - } - ] - }, - { - "id": 415, - "data_source": "Mono3DRefer", - "scene_name": "001035", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.01, \"box_2d\": [828.78, 178.35, 1208.26, 374.0], \"box_3d\": [1.6, 1.65, 3.77, 3.56, 1.67, 7.23], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [444.5, 161.47, 548.47, 258.65], \"box_3d\": [1.86, 1.63, 4.2, -2.41, 1.67, 16.01], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 2.12, \"box_2d\": [0.0, 209.15, 372.37, 374.0], \"box_3d\": [1.45, 1.46, 3.41, -2.96, 1.77, 4.79], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [714.94, 183.7, 803.06, 240.78], \"box_3d\": [1.37, 1.61, 4.26, 3.94, 1.69, 19.9], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": 1.34, \"box_2d\": [757.48, 181.76, 896.53, 273.6], \"box_3d\": [1.46, 1.44, 4.53, 3.92, 1.65, 13.91], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [558.23, 173.89, 590.94, 204.8], \"box_3d\": [1.51, 1.55, 3.92, -1.83, 1.59, 37.64], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [494.18, 178.27, 563.23, 233.47], \"box_3d\": [1.51, 1.55, 3.71, -2.4, 1.69, 21.76], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [707.39, 179.35, 768.09, 226.25], \"box_3d\": [1.5, 1.53, 3.37, 4.32, 1.74, 25.03], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [691.03, 178.26, 738.99, 214.08], \"box_3d\": [1.47, 1.56, 3.16, 4.52, 1.72, 31.61], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [679.53, 176.83, 720.43, 208.06], \"box_3d\": [1.48, 1.56, 3.62, 4.44, 1.69, 36.28], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 416, - "data_source": "Mono3DRefer", - "scene_name": "001040", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.18, \"box_2d\": [0.0, 197.78, 215.83, 305.14], \"box_3d\": [1.53, 1.66, 4.01, -9.08, 2.03, 12.99], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 2.04, \"box_2d\": [174.04, 191.53, 317.79, 260.89], \"box_3d\": [1.48, 1.6, 3.9, -9.02, 1.99, 18.17], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [412.01, 180.61, 466.15, 213.51], \"box_3d\": [1.61, 1.67, 4.56, -8.94, 2.03, 37.98], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 417, - "data_source": "Mono3DRefer", - "scene_name": "001041", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [92.56, 197.39, 440.65, 374.0], \"box_3d\": [1.42, 1.66, 4.11, -3.05, 1.73, 7.35], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [547.36, 161.37, 575.39, 189.75], \"box_3d\": [1.97, 1.87, 4.53, -3.54, 1.2, 52.85], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 418, - "data_source": "Mono3DRefer", - "scene_name": "001042", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [288.42, 168.87, 462.44, 274.57], \"box_3d\": [1.76, 1.77, 4.41, -4.42, 1.72, 14.26], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [443.04, 183.24, 538.53, 238.91], \"box_3d\": [1.48, 1.65, 4.4, -3.49, 1.81, 21.71], \"ry\": 1.69}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [573.12, 180.58, 614.56, 209.28], \"box_3d\": [1.42, 1.66, 3.64, -0.83, 1.83, 37.89], \"ry\": 1.69}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [655.23, 178.01, 686.05, 195.74], \"box_3d\": [1.44, 1.64, 3.78, 5.14, 1.89, 60.94], \"ry\": 1.92}]\n```" - } - ] - }, - { - "id": 419, - "data_source": "Mono3DRefer", - "scene_name": "001045", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.52, \"box_2d\": [560.18, 179.24, 617.82, 225.04], \"box_3d\": [1.36, 1.69, 3.38, -0.65, 1.58, 23.45], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -2.37, \"box_2d\": [90.7, 174.19, 370.46, 294.19], \"box_3d\": [1.66, 1.63, 3.83, -5.95, 1.71, 11.31], \"ry\": -2.85}, {\"category\": \"Car\", \"angle\": -2.29, \"box_2d\": [352.37, 178.09, 478.66, 228.2], \"box_3d\": [1.45, 1.61, 3.96, -6.2, 1.64, 22.71], \"ry\": -2.56}, {\"category\": \"Car\", \"angle\": -2.27, \"box_2d\": [393.5, 179.62, 497.27, 220.82], \"box_3d\": [1.36, 1.57, 3.73, -5.95, 1.63, 25.77], \"ry\": -2.5}, {\"category\": \"Car\", \"angle\": -1.48, \"box_2d\": [515.95, 179.71, 564.03, 210.72], \"box_3d\": [1.25, 1.74, 3.66, -3.01, 1.56, 31.32], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [580.59, 174.91, 619.62, 210.53], \"box_3d\": [1.52, 1.67, 3.61, -0.47, 1.63, 32.87], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 420, - "data_source": "Mono3DRefer", - "scene_name": "001047", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.97, \"box_2d\": [690.66, 173.77, 735.18, 196.56], \"box_3d\": [1.62, 1.72, 4.35, 7.63, 1.72, 54.05], \"ry\": -1.83}]\n```" - } - ] - }, - { - "id": 421, - "data_source": "Mono3DRefer", - "scene_name": "001051", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.62, \"box_2d\": [536.66, 199.09, 576.82, 236.7], \"box_3d\": [1.5, 1.62, 4.19, -2.43, 2.78, 33.33], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 422, - "data_source": "Mono3DRefer", - "scene_name": "001052", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.52, \"box_2d\": [555.03, 183.98, 611.86, 229.69], \"box_3d\": [1.36, 1.69, 3.38, -0.85, 1.75, 23.77], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -2.53, \"box_2d\": [271.15, 176.47, 449.53, 250.64], \"box_3d\": [1.66, 1.63, 3.83, -6.06, 1.77, 17.46], \"ry\": -2.86}, {\"category\": \"Car\", \"angle\": -2.35, \"box_2d\": [402.11, 178.71, 502.87, 217.41], \"box_3d\": [1.45, 1.61, 3.96, -6.37, 1.71, 28.89], \"ry\": -2.57}, {\"category\": \"Car\", \"angle\": -2.31, \"box_2d\": [430.44, 181.84, 515.37, 214.78], \"box_3d\": [1.36, 1.57, 3.73, -6.12, 1.78, 31.87], \"ry\": -2.5}, {\"category\": \"Car\", \"angle\": -1.5, \"box_2d\": [528.28, 181.17, 567.29, 206.89], \"box_3d\": [1.25, 1.74, 3.66, -3.21, 1.69, 37.49], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [578.26, 179.04, 617.44, 214.79], \"box_3d\": [1.52, 1.67, 3.61, -0.56, 1.81, 32.76], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 423, - "data_source": "Mono3DRefer", - "scene_name": "001053", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [401.3, 185.01, 440.42, 204.85], \"box_3d\": [1.46, 1.86, 4.01, -14.8, 2.43, 56.58], \"ry\": 1.61}]\n```" - } - ] - }, - { - "id": 424, - "data_source": "Mono3DRefer", - "scene_name": "001054", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.86, 718.86), (cx, cy) = (607.19, 185.22). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [518.39, 182.61, 540.51, 200.41], \"box_3d\": [1.37, 1.54, 3.56, -6.38, 1.17, 58.22], \"ry\": -1.74}, {\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [557.67, 178.77, 643.58, 248.25], \"box_3d\": [1.6, 1.76, 3.84, -0.33, 1.45, 18.74], \"ry\": -1.71}, {\"category\": \"Car\", \"angle\": -1.07, \"box_2d\": [0.0, 200.33, 204.23, 364.62], \"box_3d\": [1.36, 1.61, 3.76, -6.29, 1.59, 8.17], \"ry\": -1.71}, {\"category\": \"Car\", \"angle\": -1.26, \"box_2d\": [177.77, 191.2, 330.99, 274.65], \"box_3d\": [1.37, 1.59, 3.89, -6.93, 1.53, 14.45], \"ry\": -1.7}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [462.22, 180.88, 514.78, 229.8], \"box_3d\": [1.58, 1.68, 3.95, -4.26, 1.43, 25.49], \"ry\": -1.74}, {\"category\": \"Car\", \"angle\": -1.37, \"box_2d\": [290.34, 184.7, 377.93, 235.37], \"box_3d\": [1.42, 1.69, 4.53, -8.64, 1.42, 22.99], \"ry\": -1.72}, {\"category\": \"Car\", \"angle\": -1.25, \"box_2d\": [186.78, 185.04, 287.71, 233.76], \"box_3d\": [1.42, 1.7, 4.13, -12.06, 1.43, 23.62], \"ry\": -1.72}, {\"category\": \"Car\", \"angle\": -1.41, \"box_2d\": [370.12, 183.27, 417.41, 215.28], \"box_3d\": [1.42, 1.53, 4.2, -10.3, 1.34, 34.77], \"ry\": -1.7}, {\"category\": \"Car\", \"angle\": -1.45, \"box_2d\": [396.72, 183.18, 432.9, 208.54], \"box_3d\": [1.43, 1.62, 4.08, -11.64, 1.32, 43.49], \"ry\": -1.71}]\n```" - } - ] - }, - { - "id": 425, - "data_source": "Mono3DRefer", - "scene_name": "001055", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.0, \"box_2d\": [735.24, 163.66, 999.61, 327.36], \"box_3d\": [1.51, 1.53, 3.38, 2.7, 1.45, 8.49], \"ry\": -1.7}, {\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [599.45, 166.52, 671.97, 213.45], \"box_3d\": [1.39, 1.54, 3.42, 0.72, 1.22, 23.3], \"ry\": -1.78}, {\"category\": \"Car\", \"angle\": 2.51, \"box_2d\": [541.77, 159.23, 614.13, 189.88], \"box_3d\": [1.56, 1.39, 3.73, -1.7, 0.87, 38.41], \"ry\": 2.47}, {\"category\": \"Car\", \"angle\": -0.86, \"box_2d\": [663.51, 158.74, 741.24, 192.66], \"box_3d\": [1.56, 1.67, 3.78, 4.49, 0.92, 35.01], \"ry\": -0.74}, {\"category\": \"Car\", \"angle\": -0.96, \"box_2d\": [723.59, 158.16, 793.2, 188.28], \"box_3d\": [1.5, 1.53, 4.04, 7.81, 0.78, 37.92], \"ry\": -0.76}, {\"category\": \"Car\", \"angle\": -0.97, \"box_2d\": [786.44, 159.5, 838.94, 184.85], \"box_3d\": [1.38, 1.28, 3.16, 11.45, 0.66, 40.78], \"ry\": -0.7}, {\"category\": \"Car\", \"angle\": -1.08, \"box_2d\": [823.11, 156.24, 885.88, 185.7], \"box_3d\": [1.7, 1.7, 4.46, 14.91, 0.75, 43.94], \"ry\": -0.76}]\n```" - } - ] - }, - { - "id": 426, - "data_source": "Mono3DRefer", - "scene_name": "001056", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [632.12, 175.57, 656.04, 195.95], \"box_3d\": [1.42, 1.45, 3.2, 2.42, 1.63, 52.3], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [671.51, 180.94, 737.14, 229.57], \"box_3d\": [1.38, 1.49, 3.32, 2.81, 1.64, 22.37], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.94, \"box_2d\": [776.62, 182.77, 1023.94, 326.48], \"box_3d\": [1.5, 1.62, 3.89, 3.51, 1.65, 9.66], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.85, \"box_2d\": [734.15, 176.1, 791.67, 209.15], \"box_3d\": [1.44, 1.56, 3.96, 7.06, 1.61, 33.81], \"ry\": -1.65}]\n```" - } - ] - }, - { - "id": 427, - "data_source": "Mono3DRefer", - "scene_name": "001059", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.08, \"box_2d\": [40.81, 193.18, 113.18, 224.54], \"box_3d\": [1.43, 1.48, 3.33, -26.77, 2.49, 36.31], \"ry\": 1.45}, {\"category\": \"Car\", \"angle\": 2.16, \"box_2d\": [102.06, 187.62, 176.88, 217.77], \"box_3d\": [1.62, 1.6, 4.18, -27.47, 2.51, 42.23], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 2.07, \"box_2d\": [169.45, 188.93, 221.66, 212.52], \"box_3d\": [1.48, 1.35, 3.93, -28.17, 2.61, 49.11], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 428, - "data_source": "Mono3DRefer", - "scene_name": "001061", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.5, \"box_2d\": [742.25, 181.65, 775.99, 206.43], \"box_3d\": [1.48, 1.8, 4.31, 9.56, 2.07, 46.14], \"ry\": -1.3}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [655.01, 180.18, 688.71, 206.81], \"box_3d\": [1.69, 1.88, 4.5, 4.19, 2.19, 48.52], \"ry\": 1.74}]\n```" - } - ] - }, - { - "id": 429, - "data_source": "Mono3DRefer", - "scene_name": "001063", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.07, \"box_2d\": [136.2, 186.67, 270.96, 244.93], \"box_3d\": [1.62, 1.86, 4.31, -12.73, 2.08, 22.88], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.99, \"box_2d\": [296.43, 182.68, 351.66, 209.12], \"box_3d\": [1.67, 1.82, 4.42, -19.29, 2.36, 48.81], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [407.23, 180.3, 440.69, 200.21], \"box_3d\": [1.63, 1.87, 3.47, -15.86, 2.28, 61.59], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 430, - "data_source": "Mono3DRefer", - "scene_name": "001064", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.27, \"box_2d\": [746.23, 174.82, 800.49, 199.56], \"box_3d\": [1.48, 1.36, 3.51, 10.17, 1.63, 45.23], \"ry\": -2.05}]\n```" - } - ] - }, - { - "id": 431, - "data_source": "Mono3DRefer", - "scene_name": "001072", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [656.07, 177.53, 695.54, 208.95], \"box_3d\": [1.42, 1.45, 3.2, 3.07, 1.66, 34.5], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.89, \"box_2d\": [750.57, 189.07, 936.56, 306.8], \"box_3d\": [1.38, 1.49, 3.32, 3.13, 1.64, 10.46], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.83, \"box_2d\": [748.14, 177.09, 802.09, 211.97], \"box_3d\": [1.43, 1.48, 3.25, 7.14, 1.63, 31.6], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -2.02, \"box_2d\": [837.99, 179.42, 970.43, 242.09], \"box_3d\": [1.44, 1.56, 3.96, 7.45, 1.62, 18.79], \"ry\": -1.65}]\n```" - } - ] - }, - { - "id": 432, - "data_source": "Mono3DRefer", - "scene_name": "001073", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.95, \"box_2d\": [424.06, 178.06, 498.55, 206.9], \"box_3d\": [1.49, 1.64, 3.7, -8.05, 1.8, 39.14], \"ry\": 3.13}]\n```" - } - ] - }, - { - "id": 433, - "data_source": "Mono3DRefer", - "scene_name": "001074", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.87, \"box_2d\": [745.63, 161.59, 934.82, 292.77], \"box_3d\": [1.64, 1.69, 3.3, 3.18, 1.52, 10.78], \"ry\": -1.59}]\n```" - } - ] - }, - { - "id": 434, - "data_source": "Mono3DRefer", - "scene_name": "001081", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [668.53, 174.31, 696.2, 192.73], \"box_3d\": [1.45, 1.61, 4.2, 5.92, 1.6, 59.52], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": -2.45, \"box_2d\": [799.54, 175.15, 865.73, 202.49], \"box_3d\": [1.48, 1.36, 3.51, 12.59, 1.63, 41.05], \"ry\": -2.16}]\n```" - } - ] - }, - { - "id": 435, - "data_source": "Mono3DRefer", - "scene_name": "001086", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [537.06, 185.61, 571.09, 214.71], \"box_3d\": [1.55, 1.68, 4.42, -3.22, 2.31, 41.86], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 0.88, \"box_2d\": [902.8, 189.59, 1241.0, 374.0], \"box_3d\": [1.55, 1.47, 3.36, 2.78, 1.66, 3.33], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [444.61, 192.53, 530.53, 263.48], \"box_3d\": [1.46, 1.6, 3.33, -2.87, 1.97, 17.27], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [670.34, 179.86, 761.93, 254.1], \"box_3d\": [1.64, 1.7, 3.54, 2.48, 1.82, 17.89], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [660.0, 188.58, 723.99, 238.4], \"box_3d\": [1.37, 1.51, 3.39, 2.42, 1.88, 22.31], \"ry\": -1.59}]\n```" - } - ] - }, - { - "id": 436, - "data_source": "Mono3DRefer", - "scene_name": "001087", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [582.56, 180.78, 614.75, 208.09], \"box_3d\": [1.43, 1.55, 3.91, -0.7, 1.88, 40.17], \"ry\": -1.64}, {\"category\": \"Car\", \"angle\": 0.96, \"box_2d\": [1005.15, 189.01, 1241.0, 338.95], \"box_3d\": [1.45, 1.51, 3.7, 6.7, 1.68, 9.09], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 2.87, \"box_2d\": [701.04, 177.18, 853.58, 230.51], \"box_3d\": [1.51, 1.65, 4.21, 4.97, 1.67, 21.93], \"ry\": 3.09}, {\"category\": \"Car\", \"angle\": -0.2, \"box_2d\": [696.84, 177.67, 829.09, 223.96], \"box_3d\": [1.48, 1.49, 4.16, 5.09, 1.67, 24.51], \"ry\": 0.0}, {\"category\": \"Car\", \"angle\": -0.03, \"box_2d\": [674.65, 179.84, 778.42, 215.47], \"box_3d\": [1.51, 1.69, 4.42, 5.09, 1.84, 32.05], \"ry\": 0.12}]\n```" - } - ] - }, - { - "id": 437, - "data_source": "Mono3DRefer", - "scene_name": "001107", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.75, \"box_2d\": [0.0, 221.17, 281.48, 374.0], \"box_3d\": [1.4, 1.67, 3.72, -2.99, 1.7, 2.68], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": 1.44, \"box_2d\": [647.48, 177.36, 726.98, 239.27], \"box_3d\": [1.49, 1.61, 3.74, 1.94, 1.62, 19.4], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": -2.95, \"box_2d\": [417.0, 178.0, 514.79, 211.01], \"box_3d\": [1.47, 1.68, 4.26, -6.76, 1.74, 33.95], \"ry\": -3.14}, {\"category\": \"Car\", \"angle\": -0.47, \"box_2d\": [918.26, 159.56, 1211.09, 264.51], \"box_3d\": [1.84, 1.6, 4.46, 8.4, 1.64, 13.7], \"ry\": 0.07}]\n```" - } - ] - }, - { - "id": 438, - "data_source": "Mono3DRefer", - "scene_name": "001112", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.38, \"box_2d\": [210.92, 184.56, 376.98, 246.45], \"box_3d\": [1.52, 1.64, 3.7, -8.36, 1.86, 19.2], \"ry\": -0.03}, {\"category\": \"Car\", \"angle\": -2.71, \"box_2d\": [292.45, 184.87, 419.77, 231.99], \"box_3d\": [1.5, 1.62, 3.69, -8.53, 1.92, 24.27], \"ry\": -3.05}, {\"category\": \"Car\", \"angle\": 1.42, \"box_2d\": [894.3, 162.15, 932.76, 185.75], \"box_3d\": [1.51, 1.7, 4.45, 20.49, 0.85, 48.89], \"ry\": 1.81}]\n```" - } - ] - }, - { - "id": 439, - "data_source": "Mono3DRefer", - "scene_name": "001119", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.13, \"box_2d\": [247.64, 202.13, 366.05, 265.63], \"box_3d\": [1.39, 1.59, 3.17, -7.74, 2.2, 18.61], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": -1.01, \"box_2d\": [152.18, 201.88, 304.6, 268.48], \"box_3d\": [1.46, 1.55, 4.28, -10.1, 2.31, 19.41], \"ry\": -1.49}]\n```" - } - ] - }, - { - "id": 440, - "data_source": "Mono3DRefer", - "scene_name": "001121", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.46, \"box_2d\": [351.45, 188.33, 386.78, 215.92], \"box_3d\": [1.65, 1.75, 3.39, -15.45, 2.66, 46.13], \"ry\": 1.13}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [356.52, 189.69, 472.74, 276.51], \"box_3d\": [1.47, 1.7, 4.39, -3.94, 1.85, 15.06], \"ry\": 1.46}]\n```" - } - ] - }, - { - "id": 441, - "data_source": "Mono3DRefer", - "scene_name": "001122", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.96, \"box_2d\": [325.84, 178.85, 371.06, 202.22], \"box_3d\": [1.58, 1.64, 3.88, -18.52, 2.02, 51.19], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [354.4, 180.4, 396.96, 199.57], \"box_3d\": [1.43, 1.81, 4.16, -18.34, 2.03, 56.6], \"ry\": 1.62}]\n```" - } - ] - }, - { - "id": 442, - "data_source": "Mono3DRefer", - "scene_name": "001125", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.48, \"box_2d\": [947.8, 162.36, 1241.0, 336.23], \"box_3d\": [1.76, 1.71, 4.4, 6.45, 1.66, 8.29], \"ry\": 3.12}, {\"category\": \"Car\", \"angle\": -2.4, \"box_2d\": [0.0, 188.69, 137.68, 349.48], \"box_3d\": [1.4, 1.46, 3.23, -6.79, 1.58, 7.09], \"ry\": 3.14}, {\"category\": \"Car\", \"angle\": -2.72, \"box_2d\": [144.52, 180.03, 400.86, 260.91], \"box_3d\": [1.46, 1.61, 4.31, -6.57, 1.63, 14.25], \"ry\": -3.14}, {\"category\": \"Car\", \"angle\": -2.96, \"box_2d\": [288.24, 169.08, 471.42, 242.12], \"box_3d\": [1.8, 1.77, 4.41, -6.06, 1.74, 19.31], \"ry\": 3.02}, {\"category\": \"Car\", \"angle\": -2.92, \"box_2d\": [423.46, 173.62, 499.82, 208.12], \"box_3d\": [1.58, 1.54, 3.3, -7.08, 1.64, 34.47], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": 0.18, \"box_2d\": [441.13, 172.79, 526.21, 207.02], \"box_3d\": [1.68, 1.71, 4.07, -6.49, 1.71, 37.19], \"ry\": 0.01}]\n```" - } - ] - }, - { - "id": 443, - "data_source": "Mono3DRefer", - "scene_name": "001126", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [442.17, 174.06, 494.78, 227.32], \"box_3d\": [1.68, 1.63, 3.9, -4.88, 1.75, 25.04], \"ry\": 1.42}, {\"category\": \"Car\", \"angle\": 1.46, \"box_2d\": [447.24, 180.03, 487.94, 208.86], \"box_3d\": [1.47, 1.7, 4.39, -7.87, 1.88, 39.46], \"ry\": 1.27}]\n```" - } - ] - }, - { - "id": 444, - "data_source": "Mono3DRefer", - "scene_name": "001128", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [638.63, 178.15, 673.6, 206.26], \"box_3d\": [1.38, 1.49, 3.32, 2.32, 1.67, 37.43], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [676.8, 177.34, 747.4, 228.05], \"box_3d\": [1.5, 1.62, 3.89, 3.21, 1.67, 23.59], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -2.05, \"box_2d\": [834.68, 162.34, 1241.0, 374.0], \"box_3d\": [1.69, 1.58, 3.95, 3.61, 1.65, 7.05], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -2.71, \"box_2d\": [911.0, 177.09, 996.49, 209.04], \"box_3d\": [1.48, 1.36, 3.51, 16.76, 1.71, 35.34], \"ry\": -2.27}]\n```" - } - ] - }, - { - "id": 445, - "data_source": "Mono3DRefer", - "scene_name": "001134", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [421.55, 190.86, 543.65, 265.57], \"box_3d\": [1.42, 1.66, 3.64, -2.75, 1.86, 16.15], \"ry\": 1.7}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [617.33, 180.64, 665.18, 208.59], \"box_3d\": [1.44, 1.64, 3.78, 1.74, 1.88, 39.58], \"ry\": 1.89}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [769.8, 177.81, 796.17, 199.46], \"box_3d\": [1.45, 1.71, 3.93, 12.13, 1.81, 50.62], \"ry\": -1.32}]\n```" - } - ] - }, - { - "id": 446, - "data_source": "Mono3DRefer", - "scene_name": "001135", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [688.83, 190.67, 812.75, 275.74], \"box_3d\": [1.41, 1.58, 4.36, 2.66, 1.81, 14.81], \"ry\": -1.6}]\n```" - } - ] - }, - { - "id": 447, - "data_source": "Mono3DRefer", - "scene_name": "001137", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [817.46, 171.98, 888.89, 232.04], \"box_3d\": [1.56, 1.65, 3.69, 6.84, 1.56, 20.71], \"ry\": -1.32}, {\"category\": \"Car\", \"angle\": 2.25, \"box_2d\": [120.62, 186.63, 420.87, 306.75], \"box_3d\": [1.42, 1.67, 4.53, -4.93, 1.67, 11.24], \"ry\": 1.85}]\n```" - } - ] - }, - { - "id": 448, - "data_source": "Mono3DRefer", - "scene_name": "001147", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [364.27, 187.73, 505.2, 294.06], \"box_3d\": [1.46, 1.6, 3.33, -2.78, 1.73, 11.93], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [698.25, 175.68, 847.78, 289.35], \"box_3d\": [1.64, 1.7, 3.54, 2.55, 1.71, 12.3], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [677.84, 182.79, 768.35, 249.36], \"box_3d\": [1.37, 1.51, 3.39, 2.49, 1.61, 16.82], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [528.88, 177.16, 568.53, 210.21], \"box_3d\": [1.55, 1.68, 4.42, -3.06, 1.79, 36.44], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 449, - "data_source": "Mono3DRefer", - "scene_name": "001150", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [251.16, 183.07, 401.23, 270.98], \"box_3d\": [1.45, 1.6, 3.2, -5.3, 1.66, 13.76], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -2.31, \"box_2d\": [962.46, 190.79, 1241.0, 374.0], \"box_3d\": [1.51, 1.61, 3.51, 3.4, 1.63, 3.57], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.97, \"box_2d\": [767.57, 180.19, 1035.7, 330.57], \"box_3d\": [1.49, 1.68, 3.8, 3.32, 1.6, 9.18], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [347.16, 179.67, 455.37, 245.41], \"box_3d\": [1.49, 1.63, 3.68, -5.23, 1.68, 18.41], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [729.47, 177.45, 854.94, 261.92], \"box_3d\": [1.49, 1.61, 3.55, 3.51, 1.6, 14.67], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 450, - "data_source": "Mono3DRefer", - "scene_name": "001152", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [460.16, 183.49, 498.25, 210.07], \"box_3d\": [1.51, 1.51, 4.15, -7.93, 2.17, 43.87], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 451, - "data_source": "Mono3DRefer", - "scene_name": "001154", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.06, \"box_2d\": [189.45, 194.3, 279.52, 229.9], \"box_3d\": [1.39, 1.82, 4.11, -16.66, 2.4, 32.19], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [363.95, 184.94, 408.43, 209.97], \"box_3d\": [1.45, 1.64, 3.61, -13.88, 2.22, 44.83], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 452, - "data_source": "Mono3DRefer", - "scene_name": "001159", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [627.62, 174.34, 653.79, 195.78], \"box_3d\": [1.5, 1.62, 3.89, 2.21, 1.64, 53.1], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [666.21, 171.23, 725.87, 220.32], \"box_3d\": [1.69, 1.58, 3.95, 3.12, 1.66, 27.09], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [712.38, 181.73, 806.6, 246.92], \"box_3d\": [1.41, 1.57, 3.23, 3.47, 1.63, 17.45], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.99, \"box_2d\": [854.3, 179.29, 1003.62, 249.14], \"box_3d\": [1.45, 1.61, 4.2, 7.37, 1.61, 17.25], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -2.47, \"box_2d\": [807.36, 175.17, 874.98, 202.85], \"box_3d\": [1.48, 1.36, 3.51, 12.92, 1.63, 40.55], \"ry\": -2.17}]\n```" - } - ] - }, - { - "id": 453, - "data_source": "Mono3DRefer", - "scene_name": "001163", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.17, \"box_2d\": [0.0, 186.86, 396.73, 374.0], \"box_3d\": [1.45, 1.67, 4.43, -2.92, 1.57, 4.4], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": -2.15, \"box_2d\": [849.87, 210.66, 1241.0, 374.0], \"box_3d\": [1.47, 1.59, 3.62, 2.83, 1.78, 4.28], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [257.95, 174.6, 467.69, 308.08], \"box_3d\": [1.56, 1.68, 3.69, -3.33, 1.61, 10.35], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.31, \"box_2d\": [726.68, 194.68, 953.71, 341.56], \"box_3d\": [1.44, 1.64, 4.5, 2.75, 1.79, 9.86], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [448.79, 172.79, 525.44, 234.0], \"box_3d\": [1.46, 1.51, 3.24, -3.19, 1.48, 19.01], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.45, \"box_2d\": [675.51, 179.09, 740.3, 225.31], \"box_3d\": [1.38, 1.59, 4.19, 3.1, 1.59, 23.83], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [496.51, 175.41, 553.08, 216.73], \"box_3d\": [1.38, 1.69, 3.74, -3.06, 1.49, 26.22], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [651.36, 174.08, 680.72, 200.21], \"box_3d\": [1.51, 1.56, 3.81, 3.35, 1.61, 44.0], \"ry\": -1.55}]\n```" - } - ] - }, - { - "id": 454, - "data_source": "Mono3DRefer", - "scene_name": "001167", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.97, \"box_2d\": [286.97, 181.64, 356.33, 215.7], \"box_3d\": [1.58, 1.82, 3.87, -14.29, 2.03, 35.91], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 455, - "data_source": "Mono3DRefer", - "scene_name": "001170", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.39, \"box_2d\": [530.92, 164.0, 607.15, 221.47], \"box_3d\": [1.77, 1.75, 4.67, -1.32, 1.53, 24.69], \"ry\": -1.44}, {\"category\": \"Car\", \"angle\": -2.01, \"box_2d\": [68.64, 216.47, 186.07, 292.35], \"box_3d\": [1.72, 1.39, 3.07, -12.72, 2.93, 18.91], \"ry\": -2.6}, {\"category\": \"Car\", \"angle\": -0.65, \"box_2d\": [1057.51, 144.42, 1241.0, 260.09], \"box_3d\": [1.48, 1.63, 3.98, 8.72, 1.12, 10.06], \"ry\": 0.05}, {\"category\": \"Car\", \"angle\": -0.45, \"box_2d\": [883.53, 146.24, 1176.18, 233.16], \"box_3d\": [1.47, 1.52, 4.43, 7.51, 1.04, 13.28], \"ry\": 0.06}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [591.91, 161.74, 633.34, 196.46], \"box_3d\": [1.6, 1.68, 3.94, 0.15, 1.11, 35.48], \"ry\": 1.66}, {\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [639.3, 159.22, 660.87, 176.18], \"box_3d\": [1.42, 1.66, 4.11, 3.58, 0.28, 64.15], \"ry\": 1.68}]\n```" - } - ] - }, - { - "id": 456, - "data_source": "Mono3DRefer", - "scene_name": "001172", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [688.7, 177.89, 741.62, 229.33], \"box_3d\": [1.65, 1.67, 3.64, 3.64, 1.84, 25.06], \"ry\": -1.41}, {\"category\": \"Car\", \"angle\": 2.35, \"box_2d\": [0.0, 183.0, 237.63, 328.11], \"box_3d\": [1.47, 1.57, 4.24, -6.86, 1.62, 9.1], \"ry\": 1.71}, {\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [422.24, 181.85, 517.79, 229.18], \"box_3d\": [1.4, 1.68, 4.45, -4.54, 1.71, 23.91], \"ry\": 1.73}, {\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [520.91, 181.34, 575.0, 215.41], \"box_3d\": [1.45, 1.62, 3.77, -2.81, 1.86, 33.09], \"ry\": 1.72}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [559.4, 179.49, 598.95, 207.75], \"box_3d\": [1.53, 1.46, 4.18, -1.75, 1.92, 41.48], \"ry\": 1.73}, {\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [586.81, 179.17, 617.43, 202.78], \"box_3d\": [1.59, 1.57, 3.39, -0.54, 2.04, 50.54], \"ry\": 1.73}]\n```" - } - ] - }, - { - "id": 457, - "data_source": "Mono3DRefer", - "scene_name": "001181", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (707.05, 707.05), (cx, cy) = (604.08, 180.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [878.21, 187.77, 1046.06, 296.43], \"box_3d\": [1.4, 1.5, 3.72, 5.54, 1.56, 11.5], \"ry\": -1.37}, {\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [850.53, 178.02, 960.06, 250.62], \"box_3d\": [1.52, 1.64, 4.41, 7.16, 1.48, 17.34], \"ry\": -1.36}, {\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [554.79, 169.57, 604.09, 200.54], \"box_3d\": [1.85, 1.68, 4.11, -1.54, 1.19, 44.51], \"ry\": 1.91}, {\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [814.28, 175.57, 857.46, 215.94], \"box_3d\": [1.54, 1.57, 3.52, 9.51, 1.35, 29.04], \"ry\": -1.23}]\n```" - } - ] - }, - { - "id": 458, - "data_source": "Mono3DRefer", - "scene_name": "001182", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.03, \"box_2d\": [837.61, 172.88, 1241.0, 374.0], \"box_3d\": [1.46, 1.41, 3.63, 3.18, 1.48, 6.13], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [729.53, 172.17, 838.72, 242.25], \"box_3d\": [1.42, 1.55, 4.36, 3.88, 1.44, 16.95], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.48, \"box_2d\": [569.62, 181.28, 617.14, 214.34], \"box_3d\": [1.33, 1.71, 4.09, -0.7, 1.72, 31.65], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": -1.46, \"box_2d\": [594.87, 178.04, 630.43, 204.81], \"box_3d\": [1.42, 1.53, 4.12, 0.17, 1.73, 40.79], \"ry\": -1.45}]\n```" - } - ] - }, - { - "id": 459, - "data_source": "Mono3DRefer", - "scene_name": "001185", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (707.05, 707.05), (cx, cy) = (604.08, 180.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [1033.65, 191.58, 1183.93, 292.52], \"box_3d\": [1.4, 1.5, 3.72, 8.61, 1.64, 12.43], \"ry\": -1.13}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [1027.66, 184.43, 1143.76, 259.3], \"box_3d\": [1.52, 1.64, 4.41, 11.58, 1.66, 17.36], \"ry\": -1.14}, {\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [757.36, 175.72, 800.92, 203.38], \"box_3d\": [1.85, 1.68, 4.11, 12.4, 1.54, 50.11], \"ry\": 2.16}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [998.75, 181.33, 1047.09, 223.41], \"box_3d\": [1.54, 1.57, 3.52, 16.79, 1.6, 28.48], \"ry\": -1.05}]\n```" - } - ] - }, - { - "id": 460, - "data_source": "Mono3DRefer", - "scene_name": "001190", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.99, \"box_2d\": [783.71, 202.46, 1241.0, 374.0], \"box_3d\": [1.53, 1.59, 3.64, 2.51, 1.82, 5.42], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [484.75, 180.78, 565.81, 247.79], \"box_3d\": [1.63, 1.6, 3.73, -2.24, 1.85, 19.57], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [734.93, 195.8, 865.22, 309.76], \"box_3d\": [1.45, 1.5, 3.26, 2.79, 1.85, 11.34], \"ry\": -1.46}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [694.19, 186.78, 779.63, 261.05], \"box_3d\": [1.59, 1.49, 3.34, 2.95, 1.95, 17.56], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [666.01, 185.66, 732.21, 239.34], \"box_3d\": [1.54, 1.6, 3.52, 2.72, 1.97, 22.99], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [647.68, 184.95, 688.85, 218.59], \"box_3d\": [1.48, 1.6, 3.79, 2.71, 2.08, 34.57], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [574.33, 176.12, 602.26, 204.66], \"box_3d\": [1.91, 1.66, 4.22, -1.5, 2.16, 50.71], \"ry\": 1.61}]\n```" - } - ] - }, - { - "id": 461, - "data_source": "Mono3DRefer", - "scene_name": "001192", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.8, \"box_2d\": [722.77, 184.99, 871.99, 290.89], \"box_3d\": [1.51, 1.56, 3.81, 3.0, 1.74, 12.46], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [688.02, 180.19, 788.51, 252.97], \"box_3d\": [1.57, 1.68, 3.83, 2.98, 1.76, 17.71], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [509.48, 179.58, 550.05, 219.25], \"box_3d\": [1.52, 1.5, 3.85, -3.3, 1.81, 29.83], \"ry\": 1.5}]\n```" - } - ] - }, - { - "id": 462, - "data_source": "Mono3DRefer", - "scene_name": "001194", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [497.85, 179.81, 525.14, 199.42], \"box_3d\": [1.51, 1.51, 4.15, -7.91, 2.08, 58.06], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 463, - "data_source": "Mono3DRefer", - "scene_name": "001197", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [507.05, 177.97, 545.05, 210.62], \"box_3d\": [1.7, 1.72, 4.11, -4.63, 2.0, 40.04], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 464, - "data_source": "Mono3DRefer", - "scene_name": "001201", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.23, \"box_2d\": [0.0, 184.07, 85.04, 234.86], \"box_3d\": [1.58, 1.64, 3.88, -20.09, 1.98, 24.52], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 2.17, \"box_2d\": [67.31, 185.31, 179.24, 223.52], \"box_3d\": [1.43, 1.81, 4.16, -20.09, 1.97, 29.96], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 2.06, \"box_2d\": [180.29, 180.89, 257.37, 211.35], \"box_3d\": [1.47, 1.67, 4.31, -20.22, 1.9, 37.44], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [247.13, 180.1, 299.93, 206.28], \"box_3d\": [1.48, 1.44, 3.59, -20.01, 1.92, 42.98], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.99, \"box_2d\": [288.3, 180.41, 335.64, 202.72], \"box_3d\": [1.42, 1.68, 3.38, -19.91, 1.94, 48.27], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.97, \"box_2d\": [354.14, 177.17, 388.21, 197.06], \"box_3d\": [1.57, 1.47, 3.35, -19.61, 1.94, 59.29], \"ry\": 1.65}]\n```" - } - ] - }, - { - "id": 465, - "data_source": "Mono3DRefer", - "scene_name": "001206", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [551.11, 185.85, 736.26, 361.07], \"box_3d\": [1.49, 1.6, 3.83, 0.24, 1.66, 8.22], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.36, \"box_2d\": [394.24, 179.82, 507.47, 251.17], \"box_3d\": [1.46, 1.73, 4.19, -3.64, 1.64, 17.06], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [697.37, 176.79, 763.13, 219.89], \"box_3d\": [1.42, 1.66, 4.18, 4.24, 1.59, 26.23], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 466, - "data_source": "Mono3DRefer", - "scene_name": "001215", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.21, \"box_2d\": [120.46, 189.03, 484.96, 309.67], \"box_3d\": [1.45, 1.68, 4.53, -4.15, 1.72, 10.33], \"ry\": -0.17}]\n```" - } - ] - }, - { - "id": 467, - "data_source": "Mono3DRefer", - "scene_name": "001217", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.23, \"box_2d\": [227.13, 196.68, 469.41, 343.38], \"box_3d\": [1.4, 1.67, 3.72, -3.09, 1.76, 9.23], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": 1.47, \"box_2d\": [631.88, 185.03, 687.15, 230.4], \"box_3d\": [1.49, 1.61, 3.74, 1.69, 1.95, 26.16], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 468, - "data_source": "Mono3DRefer", - "scene_name": "001218", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.05, \"box_2d\": [200.75, 188.84, 302.11, 233.27], \"box_3d\": [1.51, 1.86, 3.91, -13.56, 2.15, 27.48], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [288.62, 185.97, 349.5, 213.64], \"box_3d\": [1.51, 1.88, 4.32, -17.25, 2.32, 42.92], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 469, - "data_source": "Mono3DRefer", - "scene_name": "001223", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.77, \"box_2d\": [734.8, 159.89, 940.52, 232.68], \"box_3d\": [1.73, 1.74, 4.71, 5.71, 1.44, 18.54], \"ry\": 3.07}, {\"category\": \"Car\", \"angle\": -0.64, \"box_2d\": [1049.81, 146.91, 1166.89, 182.26], \"box_3d\": [1.51, 1.64, 4.17, 22.32, 0.4, 32.44], \"ry\": -0.04}, {\"category\": \"Car\", \"angle\": -0.61, \"box_2d\": [1020.25, 148.81, 1132.58, 181.9], \"box_3d\": [1.52, 1.6, 4.42, 22.42, 0.41, 34.81], \"ry\": -0.04}, {\"category\": \"Car\", \"angle\": -0.53, \"box_2d\": [994.99, 150.16, 1106.41, 183.23], \"box_3d\": [1.62, 1.77, 4.64, 22.7, 0.51, 37.31], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [451.16, 182.92, 507.9, 227.58], \"box_3d\": [1.58, 1.76, 4.18, -5.03, 1.98, 28.1], \"ry\": 1.49}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [537.85, 177.02, 579.79, 209.73], \"box_3d\": [1.65, 1.88, 4.45, -2.73, 1.89, 38.98], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -0.26, \"box_2d\": [764.36, 159.01, 866.82, 198.0], \"box_3d\": [1.9, 1.78, 4.72, 10.44, 1.24, 36.95], \"ry\": 0.02}, {\"category\": \"Car\", \"angle\": -2.0, \"box_2d\": [891.41, 151.05, 942.47, 180.42], \"box_3d\": [1.8, 1.65, 3.55, 19.63, 0.48, 46.32], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -2.02, \"box_2d\": [925.4, 154.65, 981.16, 181.36], \"box_3d\": [1.64, 1.7, 3.91, 22.14, 0.54, 46.73], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [474.71, 184.05, 524.55, 219.46], \"box_3d\": [1.55, 1.71, 4.5, -5.27, 2.11, 34.73], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.44, \"box_2d\": [648.8, 174.17, 684.76, 199.03], \"box_3d\": [1.5, 1.71, 4.44, 3.56, 1.61, 46.21], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": -1.48, \"box_2d\": [605.08, 178.94, 630.87, 199.58], \"box_3d\": [1.58, 1.71, 3.75, 0.65, 2.08, 57.65], \"ry\": -1.47}]\n```" - } - ] - }, - { - "id": 470, - "data_source": "Mono3DRefer", - "scene_name": "001224", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [404.04, 205.12, 446.44, 243.93], \"box_3d\": [1.37, 1.61, 4.26, -7.95, 2.82, 30.71], \"ry\": -1.85}, {\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [278.72, 202.89, 308.89, 230.62], \"box_3d\": [1.51, 1.55, 3.92, -19.66, 3.45, 44.9], \"ry\": 1.2}, {\"category\": \"Car\", \"angle\": -2.23, \"box_2d\": [680.72, 205.73, 1234.85, 374.0], \"box_3d\": [1.47, 1.58, 3.55, 1.93, 1.77, 5.23], \"ry\": -1.9}, {\"category\": \"Car\", \"angle\": -1.94, \"box_2d\": [551.88, 205.52, 798.39, 355.38], \"box_3d\": [1.37, 1.7, 4.38, 0.52, 1.9, 9.76], \"ry\": -1.89}, {\"category\": \"Car\", \"angle\": -1.83, \"box_2d\": [498.65, 205.72, 629.97, 299.42], \"box_3d\": [1.4, 1.56, 3.91, -1.08, 2.12, 14.07], \"ry\": -1.9}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [440.64, 197.77, 527.46, 269.95], \"box_3d\": [1.6, 1.65, 3.77, -3.46, 2.31, 19.04], \"ry\": -1.91}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [214.19, 198.94, 281.25, 262.38], \"box_3d\": [1.86, 1.63, 4.2, -12.34, 2.81, 24.7], \"ry\": 1.21}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [113.13, 213.46, 232.24, 308.16], \"box_3d\": [1.45, 1.46, 3.41, -8.5, 2.33, 14.19], \"ry\": 1.19}]\n```" - } - ] - }, - { - "id": 471, - "data_source": "Mono3DRefer", - "scene_name": "001225", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.56, \"box_2d\": [473.76, 175.83, 577.58, 212.43], \"box_3d\": [1.47, 1.72, 4.22, -3.65, 1.62, 31.33], \"ry\": 2.44}, {\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [171.25, 192.03, 385.9, 332.28], \"box_3d\": [1.42, 1.71, 4.14, -4.29, 1.73, 9.93], \"ry\": 1.39}, {\"category\": \"Car\", \"angle\": 2.63, \"box_2d\": [622.77, 173.56, 727.68, 206.46], \"box_3d\": [1.48, 1.87, 4.74, 3.13, 1.54, 34.91], \"ry\": 2.71}, {\"category\": \"Car\", \"angle\": 0.36, \"box_2d\": [267.87, 176.14, 410.65, 227.04], \"box_3d\": [1.63, 1.69, 4.23, -9.2, 1.77, 24.62], \"ry\": 0.01}]\n```" - } - ] - }, - { - "id": 472, - "data_source": "Mono3DRefer", - "scene_name": "001229", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [652.52, 171.63, 699.41, 211.55], \"box_3d\": [1.69, 1.58, 3.95, 2.91, 1.67, 32.85], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [685.19, 179.97, 751.16, 227.76], \"box_3d\": [1.41, 1.57, 3.23, 3.36, 1.65, 23.14], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.91, \"box_2d\": [798.67, 177.4, 898.9, 229.01], \"box_3d\": [1.45, 1.61, 4.2, 7.3, 1.61, 22.61], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -2.46, \"box_2d\": [801.89, 175.07, 868.48, 202.5], \"box_3d\": [1.48, 1.36, 3.51, 12.69, 1.63, 40.91], \"ry\": -2.16}]\n```" - } - ] - }, - { - "id": 473, - "data_source": "Mono3DRefer", - "scene_name": "001235", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [659.43, 186.18, 696.95, 215.03], \"box_3d\": [1.41, 1.58, 4.36, 3.57, 2.15, 38.7], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 474, - "data_source": "Mono3DRefer", - "scene_name": "001238", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.22, \"box_2d\": [0.0, 202.34, 208.45, 374.0], \"box_3d\": [1.37, 1.78, 4.65, -5.49, 1.7, 6.06], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 2.12, \"box_2d\": [1136.42, 165.83, 1241.0, 220.99], \"box_3d\": [1.48, 1.51, 4.35, 17.48, 1.29, 20.81], \"ry\": 2.81}]\n```" - } - ] - }, - { - "id": 475, - "data_source": "Mono3DRefer", - "scene_name": "001241", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.53, \"box_2d\": [560.83, 171.57, 652.63, 210.88], \"box_3d\": [1.56, 1.39, 3.73, -0.14, 1.52, 30.38], \"ry\": 2.53}, {\"category\": \"Car\", \"angle\": -0.86, \"box_2d\": [711.78, 170.15, 815.86, 215.6], \"box_3d\": [1.56, 1.67, 3.78, 5.66, 1.47, 26.54], \"ry\": -0.65}, {\"category\": \"Car\", \"angle\": -0.99, \"box_2d\": [790.61, 168.48, 881.12, 207.86], \"box_3d\": [1.5, 1.53, 4.04, 9.21, 1.34, 29.38], \"ry\": -0.69}, {\"category\": \"Car\", \"angle\": -1.04, \"box_2d\": [869.02, 169.04, 934.95, 201.4], \"box_3d\": [1.38, 1.28, 3.16, 13.01, 1.22, 32.17], \"ry\": -0.66}, {\"category\": \"Car\", \"angle\": -1.13, \"box_2d\": [911.99, 161.66, 990.44, 198.87], \"box_3d\": [1.7, 1.7, 4.46, 16.62, 1.2, 35.11], \"ry\": -0.69}, {\"category\": \"Car\", \"angle\": -1.84, \"box_2d\": [633.97, 181.71, 752.79, 258.78], \"box_3d\": [1.39, 1.54, 3.42, 1.56, 1.58, 14.96], \"ry\": -1.74}]\n```" - } - ] - }, - { - "id": 476, - "data_source": "Mono3DRefer", - "scene_name": "001242", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [712.96, 131.23, 747.56, 158.35], \"box_3d\": [1.5, 1.62, 4.19, 7.67, -0.99, 45.89], \"ry\": 1.87}]\n```" - } - ] - }, - { - "id": 477, - "data_source": "Mono3DRefer", - "scene_name": "001245", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.04, \"box_2d\": [230.07, 193.21, 303.66, 224.15], \"box_3d\": [1.33, 1.72, 3.63, -16.44, 2.35, 34.68], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [368.22, 183.61, 413.89, 207.5], \"box_3d\": [1.56, 1.7, 4.35, -15.27, 2.33, 50.48], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [443.08, 181.59, 472.55, 200.29], \"box_3d\": [1.6, 1.79, 3.91, -13.61, 2.4, 64.66], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 478, - "data_source": "Mono3DRefer", - "scene_name": "001257", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [445.77, 182.94, 488.7, 211.91], \"box_3d\": [1.51, 1.51, 4.15, -7.95, 2.09, 40.34], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [500.16, 178.53, 526.03, 196.19], \"box_3d\": [1.49, 1.71, 3.94, -8.49, 2.0, 63.4], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 479, - "data_source": "Mono3DRefer", - "scene_name": "001258", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [544.05, 178.05, 569.95, 202.19], \"box_3d\": [1.7, 1.72, 4.11, -3.9, 2.1, 53.41], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 480, - "data_source": "Mono3DRefer", - "scene_name": "001269", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [582.37, 175.44, 600.77, 190.02], \"box_3d\": [1.26, 1.6, 3.56, -1.68, 1.51, 64.82], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 2.06, \"box_2d\": [112.68, 186.67, 310.31, 282.87], \"box_3d\": [1.48, 1.64, 3.59, -7.13, 1.76, 13.27], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 481, - "data_source": "Mono3DRefer", - "scene_name": "001273", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [491.98, 175.69, 542.02, 211.93], \"box_3d\": [1.39, 1.52, 3.56, -3.82, 1.53, 29.81], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [243.79, 185.14, 441.69, 312.26], \"box_3d\": [1.46, 1.59, 3.35, -3.6, 1.65, 10.19], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 482, - "data_source": "Mono3DRefer", - "scene_name": "001282", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [476.0, 184.59, 542.85, 235.0], \"box_3d\": [1.45, 1.56, 3.47, -3.16, 1.84, 22.97], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 483, - "data_source": "Mono3DRefer", - "scene_name": "001283", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [487.18, 185.34, 543.72, 228.7], \"box_3d\": [1.45, 1.53, 3.98, -3.48, 1.94, 26.9], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 2.27, \"box_2d\": [0.0, 190.01, 281.24, 374.0], \"box_3d\": [1.65, 1.56, 3.98, -3.7, 1.78, 4.24], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 484, - "data_source": "Mono3DRefer", - "scene_name": "001284", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [438.34, 186.62, 519.64, 243.87], \"box_3d\": [1.39, 1.52, 3.56, -3.56, 1.8, 19.91], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [531.48, 182.03, 557.37, 202.27], \"box_3d\": [1.34, 1.64, 4.57, -4.63, 2.01, 51.14], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": -3.0, \"box_2d\": [0.0, 188.88, 231.35, 285.86], \"box_3d\": [1.53, 1.54, 4.35, -9.33, 1.85, 12.99], \"ry\": 2.67}, {\"category\": \"Car\", \"angle\": 2.89, \"box_2d\": [435.11, 181.59, 511.62, 209.09], \"box_3d\": [1.52, 1.41, 4.22, -8.1, 2.07, 42.67], \"ry\": 2.7}]\n```" - } - ] - }, - { - "id": 485, - "data_source": "Mono3DRefer", - "scene_name": "001285", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.0, \"box_2d\": [236.7, 194.84, 309.81, 227.72], \"box_3d\": [1.48, 1.79, 4.15, -17.13, 2.66, 36.85], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 486, - "data_source": "Mono3DRefer", - "scene_name": "001299", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.51, \"box_2d\": [564.58, 179.3, 605.64, 214.49], \"box_3d\": [1.43, 1.55, 3.91, -1.07, 1.72, 31.55], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -0.59, \"box_2d\": [969.3, 177.15, 1241.0, 374.0], \"box_3d\": [1.51, 1.69, 4.42, 5.36, 1.54, 5.52], \"ry\": 0.15}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [489.74, 174.43, 531.85, 203.18], \"box_3d\": [1.44, 1.58, 3.35, -5.23, 1.54, 38.13], \"ry\": 1.63}]\n```" - } - ] - }, - { - "id": 487, - "data_source": "Mono3DRefer", - "scene_name": "001302", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [344.42, 191.34, 389.25, 214.35], \"box_3d\": [1.48, 1.79, 4.15, -17.19, 2.83, 51.12], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 488, - "data_source": "Mono3DRefer", - "scene_name": "001311", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [663.07, 177.34, 729.99, 232.64], \"box_3d\": [1.57, 1.59, 3.75, 2.59, 1.73, 22.64], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 0.29, \"box_2d\": [329.7, 189.17, 415.25, 220.29], \"box_3d\": [1.52, 1.51, 3.94, -12.3, 2.39, 37.4], \"ry\": -0.02}, {\"category\": \"Car\", \"angle\": 0.45, \"box_2d\": [215.09, 190.27, 311.46, 222.74], \"box_3d\": [1.58, 1.56, 4.24, -17.95, 2.52, 37.39], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [632.99, 178.51, 659.51, 200.62], \"box_3d\": [1.49, 1.55, 3.93, 2.52, 1.91, 51.28], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [592.91, 178.26, 617.75, 201.89], \"box_3d\": [1.54, 1.63, 3.67, -0.34, 1.93, 49.52], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [259.77, 193.29, 474.86, 337.83], \"box_3d\": [1.45, 1.59, 3.59, -2.95, 1.75, 9.42], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -2.17, \"box_2d\": [848.87, 178.17, 1241.0, 374.0], \"box_3d\": [1.64, 1.64, 3.89, 2.72, 1.67, 3.83], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.94, \"box_2d\": [762.37, 171.1, 994.7, 362.52], \"box_3d\": [1.69, 1.39, 2.65, 2.65, 1.7, 7.79], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [676.2, 180.49, 771.98, 250.78], \"box_3d\": [1.5, 1.59, 3.79, 2.6, 1.7, 17.51], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [644.17, 178.82, 685.17, 214.02], \"box_3d\": [1.49, 1.57, 4.07, 2.4, 1.77, 32.83], \"ry\": -1.56}]\n```" - } - ] - }, - { - "id": 489, - "data_source": "Mono3DRefer", - "scene_name": "001313", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [628.51, 182.54, 712.5, 242.39], \"box_3d\": [1.38, 1.51, 4.1, 1.44, 1.65, 19.06], \"ry\": -1.66}, {\"category\": \"Car\", \"angle\": -2.01, \"box_2d\": [748.35, 161.41, 1078.23, 365.69], \"box_3d\": [1.6, 1.66, 3.55, 2.74, 1.53, 7.54], \"ry\": -1.67}]\n```" - } - ] - }, - { - "id": 490, - "data_source": "Mono3DRefer", - "scene_name": "001321", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [205.58, 191.44, 470.05, 343.25], \"box_3d\": [1.48, 1.87, 4.42, -3.29, 1.77, 9.65], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 491, - "data_source": "Mono3DRefer", - "scene_name": "001325", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [532.85, 180.03, 561.06, 203.5], \"box_3d\": [1.45, 1.53, 3.98, -4.11, 1.93, 47.2], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [452.04, 179.0, 522.51, 234.5], \"box_3d\": [1.65, 1.56, 3.98, -3.96, 1.86, 23.59], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 492, - "data_source": "Mono3DRefer", - "scene_name": "001327", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.6, \"box_2d\": [556.79, 167.86, 581.86, 193.45], \"box_3d\": [1.97, 1.87, 4.53, -3.27, 1.62, 58.39], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [355.71, 194.3, 503.1, 296.31], \"box_3d\": [1.42, 1.66, 4.11, -3.04, 1.85, 12.81], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 493, - "data_source": "Mono3DRefer", - "scene_name": "001328", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.64, \"box_2d\": [116.87, 189.54, 209.67, 221.62], \"box_3d\": [1.44, 1.47, 3.57, -21.18, 2.26, 34.2], \"ry\": 0.09}, {\"category\": \"Car\", \"angle\": 0.65, \"box_2d\": [159.9, 191.13, 233.32, 217.38], \"box_3d\": [1.28, 1.35, 3.07, -21.21, 2.24, 36.99], \"ry\": 0.13}]\n```" - } - ] - }, - { - "id": 494, - "data_source": "Mono3DRefer", - "scene_name": "001333", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [240.21, 189.76, 473.04, 353.44], \"box_3d\": [1.46, 1.6, 3.33, -2.72, 1.69, 8.38], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": -1.86, \"box_2d\": [734.55, 175.37, 984.05, 351.7], \"box_3d\": [1.64, 1.7, 3.54, 2.61, 1.69, 8.56], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [697.7, 184.68, 821.87, 272.94], \"box_3d\": [1.37, 1.51, 3.39, 2.55, 1.6, 13.18], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [523.97, 173.29, 568.27, 210.23], \"box_3d\": [1.55, 1.68, 4.42, -2.87, 1.6, 32.82], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 495, - "data_source": "Mono3DRefer", - "scene_name": "001337", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.02, \"box_2d\": [3.73, 182.35, 409.71, 374.0], \"box_3d\": [1.48, 1.64, 3.95, -3.35, 1.58, 6.88], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.92, \"box_2d\": [766.33, 193.96, 1118.85, 374.0], \"box_3d\": [1.35, 1.64, 3.84, 2.7, 1.6, 6.97], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.41, \"box_2d\": [677.33, 185.85, 774.34, 254.65], \"box_3d\": [1.42, 1.59, 4.19, 2.63, 1.76, 17.47], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [426.81, 181.22, 514.35, 245.4], \"box_3d\": [1.42, 1.64, 3.63, -3.41, 1.64, 18.01], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [662.33, 182.68, 725.51, 229.63], \"box_3d\": [1.38, 1.58, 3.66, 2.61, 1.71, 23.42], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.46, \"box_2d\": [651.8, 181.15, 704.35, 220.8], \"box_3d\": [1.42, 1.6, 3.78, 2.55, 1.75, 28.06], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 496, - "data_source": "Mono3DRefer", - "scene_name": "001339", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.92, \"box_2d\": [984.32, 203.17, 1241.0, 374.0], \"box_3d\": [1.4, 1.64, 3.73, 3.95, 1.68, 5.13], \"ry\": -1.29}, {\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [851.65, 189.06, 909.41, 231.37], \"box_3d\": [1.4, 1.61, 4.37, 10.1, 2.05, 27.28], \"ry\": -1.31}]\n```" - } - ] - }, - { - "id": 497, - "data_source": "Mono3DRefer", - "scene_name": "001343", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [511.41, 171.48, 557.34, 215.64], \"box_3d\": [1.97, 1.87, 4.53, -3.61, 1.94, 34.82], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [551.6, 174.93, 574.47, 196.69], \"box_3d\": [1.7, 1.53, 3.82, -3.82, 1.9, 58.92], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 498, - "data_source": "Mono3DRefer", - "scene_name": "001352", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.17, \"box_2d\": [11.14, 192.2, 196.35, 256.62], \"box_3d\": [1.5, 1.95, 4.6, -13.89, 2.09, 20.14], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [403.52, 182.12, 445.83, 208.27], \"box_3d\": [1.84, 2.04, 4.33, -13.81, 2.55, 53.93], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 499, - "data_source": "Mono3DRefer", - "scene_name": "001362", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.27, \"box_2d\": [427.79, 183.01, 510.67, 207.39], \"box_3d\": [1.36, 1.68, 4.49, -8.28, 1.98, 42.47], \"ry\": 0.08}]\n```" - } - ] - }, - { - "id": 500, - "data_source": "Mono3DRefer", - "scene_name": "001367", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.17, \"box_2d\": [0.0, 150.49, 262.49, 374.0], \"box_3d\": [1.74, 1.65, 4.18, -4.21, 1.64, 5.06], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 1.99, \"box_2d\": [322.34, 170.39, 497.9, 287.2], \"box_3d\": [1.64, 1.45, 3.48, -3.16, 1.62, 11.93], \"ry\": 1.74}, {\"category\": \"Car\", \"angle\": -2.0, \"box_2d\": [843.24, 172.26, 1241.0, 374.0], \"box_3d\": [1.61, 1.68, 4.04, 3.36, 1.63, 6.14], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [775.64, 181.29, 935.15, 299.47], \"box_3d\": [1.52, 1.58, 4.28, 3.65, 1.67, 11.59], \"ry\": -1.44}, {\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [597.9, 174.32, 693.8, 228.54], \"box_3d\": [1.54, 1.64, 4.08, 1.19, 1.61, 22.72], \"ry\": 1.99}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [671.46, 169.03, 721.44, 200.35], \"box_3d\": [1.5, 1.6, 4.17, 4.43, 1.34, 36.71], \"ry\": 1.92}]\n```" - } - ] - }, - { - "id": 501, - "data_source": "Mono3DRefer", - "scene_name": "001369", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.29, \"box_2d\": [0.0, 198.04, 136.72, 312.35], \"box_3d\": [1.47, 1.63, 4.11, -9.04, 1.89, 10.36], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 2.05, \"box_2d\": [151.89, 176.21, 306.54, 262.73], \"box_3d\": [1.8, 1.61, 3.83, -8.85, 1.9, 17.07], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 2.05, \"box_2d\": [275.15, 188.44, 375.68, 241.24], \"box_3d\": [1.45, 1.45, 3.4, -8.65, 1.95, 22.12], \"ry\": 1.68}, {\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [335.96, 178.46, 422.45, 229.9], \"box_3d\": [1.75, 1.71, 4.31, -8.53, 1.98, 26.96], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [425.51, 182.42, 478.7, 214.83], \"box_3d\": [1.53, 1.66, 4.01, -7.97, 2.03, 36.61], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [455.76, 181.61, 499.42, 208.89], \"box_3d\": [1.48, 1.6, 3.9, -7.64, 2.0, 41.79], \"ry\": 1.63}]\n```" - } - ] - }, - { - "id": 502, - "data_source": "Mono3DRefer", - "scene_name": "001370", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [585.35, 181.65, 657.55, 240.59], \"box_3d\": [1.36, 1.69, 3.38, 0.23, 1.6, 18.68], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.24, \"box_2d\": [0.0, 167.03, 151.38, 258.15], \"box_3d\": [1.59, 1.72, 3.86, -11.38, 1.52, 14.52], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -2.11, \"box_2d\": [313.44, 168.13, 470.42, 249.95], \"box_3d\": [1.59, 1.63, 3.64, -4.89, 1.52, 15.82], \"ry\": -2.41}, {\"category\": \"Car\", \"angle\": -2.27, \"box_2d\": [350.77, 176.23, 493.11, 235.61], \"box_3d\": [1.41, 1.64, 3.61, -4.99, 1.52, 18.84], \"ry\": -2.53}, {\"category\": \"Car\", \"angle\": -2.3, \"box_2d\": [391.69, 180.27, 523.45, 230.53], \"box_3d\": [1.39, 1.61, 4.09, -4.74, 1.64, 22.01], \"ry\": -2.51}, {\"category\": \"Car\", \"angle\": -2.18, \"box_2d\": [464.82, 174.33, 547.37, 213.55], \"box_3d\": [1.5, 1.57, 3.54, -4.32, 1.58, 29.43], \"ry\": -2.33}, {\"category\": \"Car\", \"angle\": -2.37, \"box_2d\": [507.32, 174.59, 569.78, 199.16], \"box_3d\": [1.4, 1.6, 3.55, -4.28, 1.52, 42.79], \"ry\": -2.47}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [689.73, 179.73, 756.9, 223.89], \"box_3d\": [1.36, 1.63, 4.02, 3.71, 1.6, 24.44], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [674.23, 171.57, 726.91, 212.49], \"box_3d\": [1.66, 1.72, 4.41, 3.88, 1.64, 31.81], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [588.46, 176.8, 629.42, 214.17], \"box_3d\": [1.52, 1.67, 3.61, -0.07, 1.71, 31.41], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.03, \"box_2d\": [262.53, 170.9, 341.7, 204.09], \"box_3d\": [1.47, 1.77, 4.25, -14.64, 1.42, 34.5], \"ry\": 1.63}]\n```" - } - ] - }, - { - "id": 503, - "data_source": "Mono3DRefer", - "scene_name": "001376", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.96, \"box_2d\": [775.91, 192.32, 1209.54, 374.0], \"box_3d\": [1.52, 1.51, 4.28, 2.69, 1.74, 6.35], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [396.94, 195.8, 513.24, 280.01], \"box_3d\": [1.42, 1.57, 4.1, -3.13, 1.96, 15.12], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.82, \"box_2d\": [685.79, 190.85, 858.72, 304.94], \"box_3d\": [1.43, 1.64, 4.14, 2.33, 1.76, 11.63], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [524.23, 186.07, 559.36, 215.81], \"box_3d\": [1.38, 1.53, 3.58, -3.42, 2.06, 36.29], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [631.37, 181.8, 664.18, 209.2], \"box_3d\": [1.49, 1.65, 4.06, 2.12, 2.02, 41.96], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 504, - "data_source": "Mono3DRefer", - "scene_name": "001382", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [1006.57, 195.04, 1209.1, 354.6], \"box_3d\": [1.4, 1.64, 3.73, 5.74, 1.71, 8.75], \"ry\": -1.05}, {\"category\": \"Car\", \"angle\": -1.61, \"box_2d\": [1001.98, 188.79, 1055.23, 228.78], \"box_3d\": [1.4, 1.61, 4.37, 16.59, 2.07, 28.79], \"ry\": -1.09}]\n```" - } - ] - }, - { - "id": 505, - "data_source": "Mono3DRefer", - "scene_name": "001385", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [537.18, 181.04, 569.55, 204.77], \"box_3d\": [1.45, 1.77, 4.04, -3.65, 1.99, 46.82], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 506, - "data_source": "Mono3DRefer", - "scene_name": "001390", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [632.19, 175.52, 657.01, 195.48], \"box_3d\": [1.41, 1.57, 3.23, 2.5, 1.62, 53.12], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [683.75, 174.84, 717.55, 196.65], \"box_3d\": [1.45, 1.61, 4.2, 6.29, 1.61, 50.59], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": -2.45, \"box_2d\": [800.07, 174.38, 866.28, 201.72], \"box_3d\": [1.48, 1.36, 3.51, 12.62, 1.59, 41.02], \"ry\": -2.16}]\n```" - } - ] - }, - { - "id": 507, - "data_source": "Mono3DRefer", - "scene_name": "001392", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [450.38, 175.33, 502.11, 214.16], \"box_3d\": [1.62, 1.75, 4.3, -6.0, 1.76, 32.59], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": 1.96, \"box_2d\": [188.08, 184.42, 393.95, 301.9], \"box_3d\": [1.44, 1.58, 3.35, -4.55, 1.63, 10.72], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [305.32, 168.75, 450.38, 267.74], \"box_3d\": [1.74, 1.71, 4.41, -4.65, 1.7, 15.01], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [415.55, 182.2, 485.46, 230.82], \"box_3d\": [1.41, 1.72, 4.41, -5.16, 1.73, 23.64], \"ry\": 1.47}]\n```" - } - ] - }, - { - "id": 508, - "data_source": "Mono3DRefer", - "scene_name": "001399", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [542.45, 179.21, 571.74, 201.0], \"box_3d\": [1.45, 1.77, 4.04, -3.7, 1.91, 50.71], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 509, - "data_source": "Mono3DRefer", - "scene_name": "001403", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [652.16, 179.52, 699.38, 216.18], \"box_3d\": [1.38, 1.49, 3.32, 2.56, 1.66, 29.1], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.82, \"box_2d\": [715.73, 178.7, 839.32, 259.68], \"box_3d\": [1.5, 1.62, 3.89, 3.4, 1.64, 15.51], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [705.91, 175.32, 747.82, 201.42], \"box_3d\": [1.44, 1.56, 3.96, 6.76, 1.61, 42.24], \"ry\": -1.64}, {\"category\": \"Car\", \"angle\": -2.83, \"box_2d\": [986.77, 179.33, 1083.81, 214.13], \"box_3d\": [1.48, 1.36, 3.51, 19.14, 1.79, 32.62], \"ry\": -2.3}]\n```" - } - ] - }, - { - "id": 510, - "data_source": "Mono3DRefer", - "scene_name": "001408", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [527.82, 181.82, 563.96, 208.46], \"box_3d\": [1.44, 1.72, 4.14, -3.7, 1.98, 41.89], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 511, - "data_source": "Mono3DRefer", - "scene_name": "001411", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [434.86, 181.69, 527.89, 240.01], \"box_3d\": [1.49, 1.53, 3.88, -3.6, 1.76, 20.62], \"ry\": 1.69}, {\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [647.03, 171.83, 715.5, 237.88], \"box_3d\": [1.57, 1.56, 2.62, 1.74, 1.57, 18.64], \"ry\": -1.54}]\n```" - } - ] - }, - { - "id": 512, - "data_source": "Mono3DRefer", - "scene_name": "001412", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.28, \"box_2d\": [0.0, 214.12, 261.55, 374.0], \"box_3d\": [1.51, 1.55, 3.92, -3.44, 1.82, 3.53], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": -1.93, \"box_2d\": [742.49, 189.82, 1062.92, 371.91], \"box_3d\": [1.39, 1.54, 4.6, 2.75, 1.62, 8.13], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": 1.98, \"box_2d\": [155.46, 199.75, 441.46, 356.55], \"box_3d\": [1.4, 1.62, 4.26, -3.59, 1.81, 9.16], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [357.32, 189.23, 474.17, 274.35], \"box_3d\": [1.46, 1.52, 3.48, -3.83, 1.82, 14.6], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [428.37, 188.62, 515.55, 248.83], \"box_3d\": [1.39, 1.6, 3.8, -3.62, 1.84, 19.27], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [707.9, 181.45, 843.89, 268.96], \"box_3d\": [1.46, 1.6, 4.52, 3.07, 1.64, 14.47], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [691.19, 178.3, 771.33, 238.3], \"box_3d\": [1.48, 1.56, 3.46, 3.17, 1.64, 19.68], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 2.86, \"box_2d\": [895.57, 158.42, 1024.41, 216.08], \"box_3d\": [2.05, 1.8, 3.95, 12.92, 1.56, 26.92], \"ry\": -2.98}, {\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [641.52, 178.65, 681.88, 206.59], \"box_3d\": [1.51, 1.53, 3.38, 2.89, 1.86, 41.25], \"ry\": -1.74}, {\"category\": \"Car\", \"angle\": -1.88, \"box_2d\": [596.39, 182.68, 628.9, 201.41], \"box_3d\": [1.39, 1.54, 3.42, 0.15, 2.16, 55.96], \"ry\": -1.87}]\n```" - } - ] - }, - { - "id": 513, - "data_source": "Mono3DRefer", - "scene_name": "001418", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [188.22, 187.57, 395.15, 337.78], \"box_3d\": [1.6, 1.73, 3.62, -4.13, 1.83, 9.86], \"ry\": 1.41}, {\"category\": \"Car\", \"angle\": 2.95, \"box_2d\": [586.07, 172.2, 673.55, 204.2], \"box_3d\": [1.73, 1.74, 4.71, 1.07, 1.72, 41.07], \"ry\": 2.97}]\n```" - } - ] - }, - { - "id": 514, - "data_source": "Mono3DRefer", - "scene_name": "001424", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [434.48, 178.98, 466.1, 197.6], \"box_3d\": [1.58, 1.82, 3.87, -14.09, 2.13, 63.8], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 2.06, \"box_2d\": [165.42, 186.36, 263.29, 232.83], \"box_3d\": [1.48, 1.61, 3.39, -13.88, 1.98, 25.44], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 2.11, \"box_2d\": [109.89, 188.5, 225.55, 231.07], \"box_3d\": [1.44, 1.79, 4.32, -16.85, 2.07, 27.7], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [336.43, 179.42, 391.38, 209.13], \"box_3d\": [1.6, 1.78, 3.86, -13.98, 1.98, 41.13], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.96, \"box_2d\": [301.91, 179.62, 357.3, 208.63], \"box_3d\": [1.65, 1.77, 3.85, -16.8, 2.07, 43.36], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 515, - "data_source": "Mono3DRefer", - "scene_name": "001425", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.64, \"box_2d\": [98.99, 171.33, 340.61, 266.79], \"box_3d\": [1.9, 1.78, 4.25, -8.32, 1.9, 15.55], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": 0.21, \"box_2d\": [397.78, 177.52, 483.92, 211.34], \"box_3d\": [1.52, 1.64, 3.7, -7.99, 1.77, 34.17], \"ry\": -0.02}]\n```" - } - ] - }, - { - "id": 516, - "data_source": "Mono3DRefer", - "scene_name": "001427", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [302.18, 175.25, 357.58, 200.23], \"box_3d\": [1.34, 1.73, 3.98, -16.0, 1.5, 41.33], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 517, - "data_source": "Mono3DRefer", - "scene_name": "001430", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.39, \"box_2d\": [608.29, 187.36, 662.29, 222.46], \"box_3d\": [1.44, 1.66, 4.53, 1.2, 2.13, 32.99], \"ry\": -1.35}]\n```" - } - ] - }, - { - "id": 518, - "data_source": "Mono3DRefer", - "scene_name": "001434", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.76, \"box_2d\": [551.83, 189.55, 838.96, 316.56], \"box_3d\": [1.46, 1.55, 4.28, 1.34, 1.74, 10.6], \"ry\": -0.64}, {\"category\": \"Car\", \"angle\": -0.51, \"box_2d\": [355.95, 193.33, 559.27, 274.05], \"box_3d\": [1.49, 1.66, 4.16, -3.35, 2.0, 16.15], \"ry\": -0.71}, {\"category\": \"Car\", \"angle\": 2.79, \"box_2d\": [266.08, 182.44, 447.62, 260.04], \"box_3d\": [1.85, 1.68, 4.34, -6.84, 2.14, 19.62], \"ry\": 2.45}, {\"category\": \"Car\", \"angle\": -1.02, \"box_2d\": [683.78, 191.9, 949.71, 345.13], \"box_3d\": [1.39, 1.59, 3.17, 2.49, 1.65, 8.52], \"ry\": -0.75}]\n```" - } - ] - }, - { - "id": 519, - "data_source": "Mono3DRefer", - "scene_name": "001438", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.17, \"box_2d\": [0.0, 192.31, 243.37, 374.0], \"box_3d\": [1.49, 1.63, 3.68, -5.85, 1.75, 8.0], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": -2.17, \"box_2d\": [854.28, 187.04, 1241.0, 374.0], \"box_3d\": [1.49, 1.61, 3.55, 3.03, 1.61, 4.67], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": 1.3, \"box_2d\": [462.73, 180.1, 516.73, 212.71], \"box_3d\": [1.41, 1.57, 3.53, -5.67, 1.77, 33.54], \"ry\": 1.13}, {\"category\": \"Car\", \"angle\": 2.63, \"box_2d\": [539.28, 175.31, 603.94, 197.37], \"box_3d\": [1.47, 1.72, 4.22, -2.7, 1.66, 50.65], \"ry\": 2.58}, {\"category\": \"Car\", \"angle\": 0.76, \"box_2d\": [339.93, 181.39, 418.71, 208.14], \"box_3d\": [1.42, 1.71, 4.14, -12.93, 1.92, 40.33], \"ry\": 0.45}, {\"category\": \"Car\", \"angle\": 2.72, \"box_2d\": [644.28, 172.98, 712.52, 193.76], \"box_3d\": [1.48, 1.87, 4.74, 5.11, 1.52, 54.19], \"ry\": 2.81}]\n```" - } - ] - }, - { - "id": 520, - "data_source": "Mono3DRefer", - "scene_name": "001453", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.04, \"box_2d\": [240.97, 179.18, 304.63, 209.17], \"box_3d\": [1.58, 1.64, 3.88, -18.79, 1.94, 40.32], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 2.0, \"box_2d\": [285.89, 179.97, 343.28, 203.83], \"box_3d\": [1.43, 1.81, 4.16, -18.72, 1.89, 45.85], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 2.06, \"box_2d\": [334.38, 177.53, 384.57, 198.44], \"box_3d\": [1.47, 1.67, 4.31, -18.47, 1.83, 53.33], \"ry\": 1.72}]\n```" - } - ] - }, - { - "id": 521, - "data_source": "Mono3DRefer", - "scene_name": "001456", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.04, \"box_2d\": [48.56, 180.91, 418.21, 374.0], \"box_3d\": [1.62, 1.75, 4.3, -3.74, 1.72, 8.1], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 2.02, \"box_2d\": [50.73, 182.85, 296.62, 290.75], \"box_3d\": [1.47, 1.73, 4.48, -7.13, 1.66, 12.36], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": -1.23, \"box_2d\": [389.86, 185.34, 467.96, 230.36], \"box_3d\": [1.39, 1.59, 3.17, -6.09, 1.83, 24.43], \"ry\": -1.48}, {\"category\": \"Car\", \"angle\": -1.1, \"box_2d\": [318.04, 185.7, 418.64, 232.18], \"box_3d\": [1.46, 1.55, 4.28, -8.43, 1.94, 25.46], \"ry\": -1.41}]\n```" - } - ] - }, - { - "id": 522, - "data_source": "Mono3DRefer", - "scene_name": "001461", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.16, \"box_2d\": [50.24, 187.58, 194.39, 238.41], \"box_3d\": [1.49, 1.88, 4.41, -16.25, 2.02, 24.3], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [377.21, 181.32, 417.91, 205.64], \"box_3d\": [1.48, 1.61, 3.39, -13.63, 2.04, 46.36], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [333.84, 184.69, 385.43, 208.17], \"box_3d\": [1.44, 1.79, 4.32, -16.53, 2.25, 47.77], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [436.57, 181.29, 468.72, 200.74], \"box_3d\": [1.6, 1.78, 3.86, -13.49, 2.33, 62.0], \"ry\": 1.61}]\n```" - } - ] - }, - { - "id": 523, - "data_source": "Mono3DRefer", - "scene_name": "001462", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.51, \"box_2d\": [563.67, 182.53, 604.0, 217.39], \"box_3d\": [1.43, 1.55, 3.91, -1.14, 1.87, 32.09], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -0.49, \"box_2d\": [906.8, 175.39, 1241.0, 355.86], \"box_3d\": [1.51, 1.69, 4.42, 5.33, 1.54, 6.87], \"ry\": 0.15}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [492.23, 177.96, 532.64, 205.7], \"box_3d\": [1.44, 1.58, 3.35, -5.32, 1.73, 39.45], \"ry\": 1.62}]\n```" - } - ] - }, - { - "id": 524, - "data_source": "Mono3DRefer", - "scene_name": "001471", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [573.95, 178.19, 610.71, 213.59], \"box_3d\": [1.66, 1.73, 3.05, -0.87, 1.94, 35.78], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 525, - "data_source": "Mono3DRefer", - "scene_name": "001475", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [406.34, 179.01, 595.22, 293.96], \"box_3d\": [1.54, 1.64, 4.08, -1.61, 1.65, 11.87], \"ry\": 1.81}, {\"category\": \"Car\", \"angle\": -1.83, \"box_2d\": [822.62, 189.37, 1163.93, 374.0], \"box_3d\": [1.46, 1.57, 3.58, 2.92, 1.64, 6.42], \"ry\": -1.42}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [793.95, 174.61, 967.55, 300.92], \"box_3d\": [1.58, 1.67, 3.43, 3.81, 1.63, 10.84], \"ry\": -1.45}, {\"category\": \"Car\", \"angle\": -2.49, \"box_2d\": [1137.22, 142.67, 1241.0, 219.69], \"box_3d\": [1.84, 1.94, 3.68, 15.78, 1.12, 18.19], \"ry\": -1.78}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [553.91, 173.8, 620.25, 218.82], \"box_3d\": [1.5, 1.6, 4.17, -0.78, 1.56, 26.2], \"ry\": 1.74}, {\"category\": \"Car\", \"angle\": 2.21, \"box_2d\": [872.9, 163.66, 957.56, 202.2], \"box_3d\": [1.52, 1.58, 3.28, 12.65, 1.16, 29.95], \"ry\": 2.61}, {\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [741.74, 170.96, 787.85, 209.39], \"box_3d\": [1.58, 1.68, 3.93, 6.72, 1.53, 31.79], \"ry\": -1.43}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [605.65, 171.86, 650.37, 204.28], \"box_3d\": [1.58, 1.54, 3.98, 0.96, 1.55, 37.27], \"ry\": 1.79}]\n```" - } - ] - }, - { - "id": 526, - "data_source": "Mono3DRefer", - "scene_name": "001482", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.87, \"box_2d\": [294.41, 183.95, 436.78, 230.6], \"box_3d\": [1.47, 1.74, 4.26, -8.25, 1.87, 24.52], \"ry\": 3.09}]\n```" - } - ] - }, - { - "id": 527, - "data_source": "Mono3DRefer", - "scene_name": "001489", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [648.04, 186.87, 677.18, 210.23], \"box_3d\": [1.41, 1.58, 4.36, 3.39, 2.36, 47.41], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 528, - "data_source": "Mono3DRefer", - "scene_name": "001491", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.8, \"box_2d\": [727.07, 182.91, 839.42, 258.82], \"box_3d\": [1.41, 1.57, 3.23, 3.5, 1.64, 15.27], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -2.04, \"box_2d\": [885.25, 180.43, 1068.33, 261.45], \"box_3d\": [1.45, 1.61, 4.2, 7.41, 1.62, 15.17], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -2.48, \"box_2d\": [812.22, 175.23, 880.66, 203.12], \"box_3d\": [1.48, 1.36, 3.51, 13.12, 1.64, 40.26], \"ry\": -2.17}, {\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [630.79, 174.48, 658.92, 197.32], \"box_3d\": [1.5, 1.62, 3.89, 2.36, 1.64, 49.97], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [672.89, 171.09, 739.82, 224.97], \"box_3d\": [1.69, 1.58, 3.95, 3.2, 1.66, 24.86], \"ry\": -1.62}]\n```" - } - ] - }, - { - "id": 529, - "data_source": "Mono3DRefer", - "scene_name": "001492", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [329.92, 192.14, 378.46, 216.61], \"box_3d\": [1.48, 1.79, 4.15, -17.1, 2.82, 48.35], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 530, - "data_source": "Mono3DRefer", - "scene_name": "001493", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.19, \"box_2d\": [0.0, 191.2, 71.26, 230.18], \"box_3d\": [1.42, 1.68, 3.38, -23.7, 2.19, 28.97], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 2.17, \"box_2d\": [54.0, 189.93, 145.19, 222.49], \"box_3d\": [1.38, 1.52, 4.0, -24.02, 2.22, 34.08], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 2.06, \"box_2d\": [135.36, 184.35, 196.67, 214.77], \"box_3d\": [1.57, 1.47, 3.35, -24.47, 2.23, 39.83], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 2.0, \"box_2d\": [183.53, 186.05, 234.1, 210.75], \"box_3d\": [1.43, 1.48, 3.33, -24.79, 2.27, 44.65], \"ry\": 1.49}, {\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [266.72, 182.1, 306.8, 201.63], \"box_3d\": [1.48, 1.35, 3.93, -25.75, 2.23, 57.55], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 531, - "data_source": "Mono3DRefer", - "scene_name": "001494", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.21, \"box_2d\": [142.43, 185.25, 295.32, 256.92], \"box_3d\": [1.48, 1.71, 4.65, -9.4, 1.8, 17.72], \"ry\": -1.69}, {\"category\": \"Car\", \"angle\": -1.04, \"box_2d\": [0.0, 164.12, 177.92, 261.19], \"box_3d\": [2.04, 1.66, 4.62, -12.62, 1.89, 17.58], \"ry\": -1.66}, {\"category\": \"Car\", \"angle\": -1.44, \"box_2d\": [382.55, 171.72, 447.59, 217.6], \"box_3d\": [1.42, 1.64, 3.95, -6.61, 1.42, 24.65], \"ry\": -1.7}, {\"category\": \"Car\", \"angle\": -1.89, \"box_2d\": [700.32, 169.13, 815.28, 231.64], \"box_3d\": [1.51, 1.72, 4.5, 3.87, 1.45, 19.91], \"ry\": -1.7}, {\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [547.89, 161.41, 586.57, 197.65], \"box_3d\": [1.7, 1.8, 4.5, -2.24, 1.19, 36.35], \"ry\": -1.66}, {\"category\": \"Car\", \"angle\": -1.95, \"box_2d\": [772.34, 162.44, 893.35, 227.17], \"box_3d\": [1.64, 1.73, 4.43, 6.19, 1.41, 20.72], \"ry\": -1.67}, {\"category\": \"Car\", \"angle\": -2.06, \"box_2d\": [854.24, 168.97, 996.72, 223.41], \"box_3d\": [1.38, 1.8, 4.49, 8.87, 1.32, 20.86], \"ry\": -1.67}]\n```" - } - ] - }, - { - "id": 532, - "data_source": "Mono3DRefer", - "scene_name": "001498", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [647.89, 171.83, 691.35, 208.94], \"box_3d\": [1.69, 1.58, 3.95, 2.82, 1.67, 35.2], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [676.77, 179.34, 735.32, 222.4], \"box_3d\": [1.41, 1.57, 3.23, 3.29, 1.65, 25.5], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.89, \"box_2d\": [781.87, 176.89, 869.81, 223.52], \"box_3d\": [1.45, 1.61, 4.2, 7.26, 1.61, 24.8], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -2.46, \"box_2d\": [801.44, 175.04, 867.88, 202.43], \"box_3d\": [1.48, 1.36, 3.51, 12.68, 1.63, 40.95], \"ry\": -2.16}]\n```" - } - ] - }, - { - "id": 533, - "data_source": "Mono3DRefer", - "scene_name": "001503", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.35, \"box_2d\": [760.43, 175.03, 819.05, 200.55], \"box_3d\": [1.48, 1.36, 3.51, 10.86, 1.64, 43.88], \"ry\": -2.11}]\n```" - } - ] - }, - { - "id": 534, - "data_source": "Mono3DRefer", - "scene_name": "001506", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [580.85, 176.13, 621.32, 214.63], \"box_3d\": [1.54, 1.63, 3.67, -0.41, 1.71, 31.1], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [597.81, 172.27, 630.83, 206.71], \"box_3d\": [1.77, 1.69, 4.32, 0.2, 1.77, 39.59], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -2.0, \"box_2d\": [852.91, 183.72, 1241.0, 374.0], \"box_3d\": [1.54, 1.55, 3.51, 3.04, 1.64, 5.34], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [682.62, 167.57, 769.88, 242.63], \"box_3d\": [1.74, 1.62, 4.13, 2.89, 1.65, 18.97], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.2, \"box_2d\": [0.0, 192.44, 205.57, 327.74], \"box_3d\": [1.51, 1.65, 3.94, -7.81, 1.83, 10.39], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [667.35, 175.74, 718.45, 219.64], \"box_3d\": [1.53, 1.46, 3.26, 3.01, 1.65, 26.98], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [646.03, 173.8, 676.15, 201.56], \"box_3d\": [1.6, 1.56, 3.45, 3.03, 1.68, 43.68], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 535, - "data_source": "Mono3DRefer", - "scene_name": "001513", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.47, \"box_2d\": [0.0, 177.3, 225.34, 356.15], \"box_3d\": [1.53, 1.7, 4.32, -6.41, 1.6, 6.99], \"ry\": 3.09}, {\"category\": \"Car\", \"angle\": 0.45, \"box_2d\": [112.05, 179.8, 402.68, 277.52], \"box_3d\": [1.54, 1.78, 4.2, -6.03, 1.69, 12.58], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": -2.69, \"box_2d\": [174.85, 178.05, 363.01, 252.99], \"box_3d\": [1.46, 1.42, 3.25, -7.07, 1.58, 15.04], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": 0.21, \"box_2d\": [335.56, 182.25, 494.16, 235.51], \"box_3d\": [1.5, 1.66, 4.4, -5.87, 1.81, 21.94], \"ry\": -0.05}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [670.21, 185.83, 748.47, 243.06], \"box_3d\": [1.41, 1.62, 4.19, 2.67, 1.81, 20.5], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -0.17, \"box_2d\": [742.16, 182.39, 843.3, 224.9], \"box_3d\": [1.42, 1.59, 3.21, 6.35, 1.78, 25.48], \"ry\": 0.07}]\n```" - } - ] - }, - { - "id": 536, - "data_source": "Mono3DRefer", - "scene_name": "001520", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.24, \"box_2d\": [0.0, 189.49, 107.76, 229.52], \"box_3d\": [1.5, 1.62, 4.08, -23.35, 2.23, 30.25], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 2.15, \"box_2d\": [105.11, 188.48, 185.96, 224.38], \"box_3d\": [1.65, 1.57, 3.82, -23.25, 2.46, 36.23], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 2.1, \"box_2d\": [170.23, 190.37, 238.59, 216.92], \"box_3d\": [1.37, 1.36, 4.44, -23.2, 2.41, 41.4], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 2.12, \"box_2d\": [214.09, 187.77, 270.15, 213.41], \"box_3d\": [1.51, 1.47, 3.64, -23.35, 2.49, 45.87], \"ry\": 1.65}]\n```" - } - ] - }, - { - "id": 537, - "data_source": "Mono3DRefer", - "scene_name": "001521", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [0.0, 195.4, 282.03, 374.0], \"box_3d\": [1.44, 1.58, 3.72, -4.87, 1.73, 7.76], \"ry\": 1.36}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [411.26, 190.43, 455.68, 223.31], \"box_3d\": [1.39, 1.52, 3.56, -8.25, 2.25, 33.81], \"ry\": 1.48}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [268.57, 193.58, 394.03, 281.18], \"box_3d\": [1.46, 1.59, 3.35, -5.47, 1.91, 14.43], \"ry\": 1.44}]\n```" - } - ] - }, - { - "id": 538, - "data_source": "Mono3DRefer", - "scene_name": "001526", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.21, \"box_2d\": [676.2, 178.45, 807.47, 223.0], \"box_3d\": [1.45, 1.6, 4.23, 4.48, 1.67, 25.08], \"ry\": -0.03}, {\"category\": \"Car\", \"angle\": 3.1, \"box_2d\": [646.61, 178.3, 709.31, 201.86], \"box_3d\": [1.48, 1.72, 3.99, 4.39, 1.86, 47.26], \"ry\": -3.1}]\n```" - } - ] - }, - { - "id": 539, - "data_source": "Mono3DRefer", - "scene_name": "001527", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.67, \"box_2d\": [945.17, 179.56, 1005.96, 200.46], \"box_3d\": [1.42, 1.58, 3.48, 25.81, 1.91, 51.03], \"ry\": 3.13}, {\"category\": \"Car\", \"angle\": 2.16, \"box_2d\": [75.82, 197.6, 179.08, 240.53], \"box_3d\": [1.49, 1.81, 3.36, -18.94, 2.52, 28.43], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 540, - "data_source": "Mono3DRefer", - "scene_name": "001532", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.51, \"box_2d\": [565.18, 192.31, 596.28, 218.05], \"box_3d\": [1.44, 1.66, 4.53, -1.81, 2.7, 45.09], \"ry\": -1.55}]\n```" - } - ] - }, - { - "id": 541, - "data_source": "Mono3DRefer", - "scene_name": "001533", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.86, 718.86), (cx, cy) = (607.19, 185.22). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [577.02, 173.77, 645.64, 236.29], \"box_3d\": [1.6, 1.76, 3.84, 0.08, 1.3, 20.5], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.01, \"box_2d\": [0.0, 186.54, 244.4, 290.95], \"box_3d\": [1.5, 1.7, 3.97, -8.29, 1.55, 12.65], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.17, \"box_2d\": [238.75, 183.35, 355.11, 245.51], \"box_3d\": [1.49, 1.66, 3.37, -8.24, 1.46, 19.29], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.34, \"box_2d\": [366.65, 176.16, 486.26, 259.98], \"box_3d\": [1.63, 1.67, 3.92, -3.92, 1.45, 16.05], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.29, \"box_2d\": [349.52, 182.3, 421.21, 221.67], \"box_3d\": [1.42, 1.61, 4.05, -8.7, 1.31, 28.35], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.41, \"box_2d\": [457.17, 177.82, 516.85, 221.43], \"box_3d\": [1.56, 1.69, 3.63, -4.63, 1.29, 27.83], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.44, \"box_2d\": [492.32, 180.92, 530.02, 207.86], \"box_3d\": [1.38, 1.58, 3.49, -5.21, 1.15, 38.96], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.47, \"box_2d\": [516.35, 179.8, 547.62, 202.67], \"box_3d\": [1.43, 1.64, 4.33, -4.99, 1.08, 47.64], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.24, \"box_2d\": [308.76, 184.02, 377.03, 217.37], \"box_3d\": [1.48, 1.69, 4.52, -12.79, 1.44, 34.93], \"ry\": -1.59}]\n```" - } - ] - }, - { - "id": 542, - "data_source": "Mono3DRefer", - "scene_name": "001539", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [574.57, 182.91, 670.33, 274.3], \"box_3d\": [1.55, 1.63, 3.32, 0.17, 1.76, 14.09], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.02, \"box_2d\": [219.7, 200.99, 305.08, 237.4], \"box_3d\": [1.37, 1.82, 3.96, -15.22, 2.68, 31.76], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 2.38, \"box_2d\": [119.68, 205.96, 213.89, 243.45], \"box_3d\": [1.38, 1.79, 2.94, -18.56, 2.82, 30.27], \"ry\": 1.84}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [363.3, 194.03, 420.68, 222.85], \"box_3d\": [1.4, 1.75, 4.75, -12.08, 2.64, 40.19], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 2.61, \"box_2d\": [1045.08, 186.53, 1235.23, 243.78], \"box_3d\": [1.59, 1.73, 4.26, 15.62, 2.03, 21.49], \"ry\": -3.05}, {\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [305.72, 197.02, 357.52, 222.11], \"box_3d\": [1.46, 1.81, 4.16, -18.18, 3.1, 47.25], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 543, - "data_source": "Mono3DRefer", - "scene_name": "001541", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.2, \"box_2d\": [0.0, 160.87, 222.17, 308.93], \"box_3d\": [2.17, 1.71, 4.33, -8.97, 2.02, 12.77], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [255.1, 183.2, 373.8, 249.87], \"box_3d\": [1.73, 1.6, 4.11, -8.54, 2.05, 21.14], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [383.79, 185.63, 453.65, 223.97], \"box_3d\": [1.47, 1.63, 4.11, -8.02, 2.03, 30.5], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [440.16, 176.51, 491.11, 213.54], \"box_3d\": [1.8, 1.61, 3.83, -7.43, 2.01, 37.31], \"ry\": 1.64}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [471.95, 183.79, 510.1, 210.02], \"box_3d\": [1.45, 1.45, 3.4, -6.96, 2.1, 42.31], \"ry\": 1.64}]\n```" - } - ] - }, - { - "id": 544, - "data_source": "Mono3DRefer", - "scene_name": "001551", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.81, \"box_2d\": [672.23, 180.0, 784.83, 225.63], \"box_3d\": [1.43, 1.55, 3.91, 4.1, 1.7, 24.88], \"ry\": -0.65}, {\"category\": \"Car\", \"angle\": 2.0, \"box_2d\": [221.03, 184.55, 421.05, 298.4], \"box_3d\": [1.45, 1.47, 3.5, -4.26, 1.64, 11.1], \"ry\": 1.64}, {\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [448.75, 179.87, 526.85, 232.45], \"box_3d\": [1.44, 1.58, 3.35, -3.61, 1.66, 21.58], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [477.66, 171.77, 547.55, 225.25], \"box_3d\": [1.74, 1.71, 4.41, -3.42, 1.73, 25.85], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.6, \"box_2d\": [517.64, 178.26, 556.55, 210.23], \"box_3d\": [1.41, 1.72, 4.41, -3.47, 1.69, 34.59], \"ry\": 1.5}]\n```" - } - ] - }, - { - "id": 545, - "data_source": "Mono3DRefer", - "scene_name": "001552", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.35, \"box_2d\": [288.93, 196.54, 490.45, 352.28], \"box_3d\": [1.5, 1.55, 4.21, -2.69, 1.87, 9.61], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.82, \"box_2d\": [712.97, 168.02, 918.88, 338.6], \"box_3d\": [1.82, 1.65, 3.67, 2.44, 1.79, 9.6], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [693.43, 185.45, 810.47, 274.68], \"box_3d\": [1.52, 1.51, 4.28, 2.7, 1.81, 14.84], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [478.68, 191.54, 544.59, 242.61], \"box_3d\": [1.42, 1.57, 4.1, -3.12, 2.06, 23.26], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [657.22, 189.26, 745.7, 250.49], \"box_3d\": [1.43, 1.64, 4.14, 2.34, 1.91, 19.74], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [541.72, 186.35, 568.98, 210.47], \"box_3d\": [1.38, 1.53, 3.58, -3.35, 2.23, 44.35], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [629.65, 179.42, 657.01, 202.11], \"box_3d\": [1.49, 1.65, 4.06, 2.25, 1.96, 50.04], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 546, - "data_source": "Mono3DRefer", - "scene_name": "001561", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [544.88, 178.84, 565.51, 196.55], \"box_3d\": [1.45, 1.53, 3.98, -4.69, 1.98, 61.92], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [507.95, 178.26, 547.3, 211.85], \"box_3d\": [1.65, 1.56, 3.98, -4.28, 1.94, 37.7], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 547, - "data_source": "Mono3DRefer", - "scene_name": "001563", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.0, \"box_2d\": [948.33, 179.34, 1083.19, 241.05], \"box_3d\": [1.48, 1.51, 4.35, 11.02, 1.67, 19.55], \"ry\": 2.51}, {\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [1041.44, 180.36, 1157.77, 230.42], \"box_3d\": [1.42, 1.68, 4.29, 15.56, 1.68, 22.91], \"ry\": 2.51}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [1145.95, 179.27, 1226.34, 218.06], \"box_3d\": [1.36, 1.4, 3.8, 21.94, 1.62, 27.48], \"ry\": 2.52}]\n```" - } - ] - }, - { - "id": 548, - "data_source": "Mono3DRefer", - "scene_name": "001564", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.86, 718.86), (cx, cy) = (607.19, 185.22). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.2, \"box_2d\": [280.24, 184.38, 381.04, 237.25], \"box_3d\": [1.4, 1.53, 3.74, -8.17, 1.4, 21.43], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.3, \"box_2d\": [374.59, 183.73, 435.93, 220.24], \"box_3d\": [1.41, 1.52, 3.8, -8.51, 1.37, 30.38], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.47, \"box_2d\": [512.79, 180.32, 546.93, 204.84], \"box_3d\": [1.36, 1.61, 3.76, -4.54, 1.08, 42.11], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.5, \"box_2d\": [529.69, 178.25, 555.62, 198.21], \"box_3d\": [1.37, 1.59, 3.89, -4.67, 0.88, 51.8], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [587.63, 176.59, 614.36, 201.68], \"box_3d\": [1.58, 1.68, 3.95, -0.45, 1.03, 47.62], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [696.75, 184.66, 804.05, 257.92], \"box_3d\": [1.37, 1.54, 3.56, 2.93, 1.38, 15.62], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.23, \"box_2d\": [252.39, 178.91, 449.85, 307.04], \"box_3d\": [1.57, 1.61, 3.42, -3.58, 1.48, 10.55], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [640.39, 176.26, 679.55, 205.73], \"box_3d\": [1.46, 1.66, 4.04, 2.68, 1.01, 37.99], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -0.98, \"box_2d\": [0.0, 184.64, 221.61, 296.83], \"box_3d\": [1.45, 1.57, 3.6, -7.85, 1.46, 11.26], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.36, \"box_2d\": [405.09, 180.7, 504.5, 246.4], \"box_3d\": [1.47, 1.61, 4.16, -3.8, 1.36, 18.34], \"ry\": -1.56}]\n```" - } - ] - }, - { - "id": 549, - "data_source": "Mono3DRefer", - "scene_name": "001570", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (707.05, 707.05), (cx, cy) = (604.08, 180.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.71, \"box_2d\": [975.65, 170.06, 1223.0, 337.28], \"box_3d\": [1.31, 1.61, 3.78, 5.66, 1.22, 6.45], \"ry\": -0.01}, {\"category\": \"Car\", \"angle\": -0.24, \"box_2d\": [713.78, 168.52, 850.33, 211.11], \"box_3d\": [1.41, 1.63, 4.42, 6.18, 1.02, 25.09], \"ry\": 0.01}]\n```" - } - ] - }, - { - "id": 550, - "data_source": "Mono3DRefer", - "scene_name": "001575", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.58, \"box_2d\": [491.95, 187.1, 531.67, 228.79], \"box_3d\": [1.68, 1.63, 3.9, -4.37, 2.34, 32.02], \"ry\": 1.45}, {\"category\": \"Car\", \"angle\": 1.45, \"box_2d\": [470.38, 188.77, 505.81, 213.86], \"box_3d\": [1.47, 1.7, 4.39, -7.88, 2.52, 46.2], \"ry\": 1.28}]\n```" - } - ] - }, - { - "id": 551, - "data_source": "Mono3DRefer", - "scene_name": "001577", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.43, \"box_2d\": [359.82, 179.94, 425.4, 203.66], \"box_3d\": [1.6, 1.66, 4.49, -15.55, 2.13, 51.74], \"ry\": 2.14}, {\"category\": \"Car\", \"angle\": 2.9, \"box_2d\": [491.31, 172.41, 541.74, 194.71], \"box_3d\": [1.78, 1.67, 3.85, -7.78, 1.77, 59.92], \"ry\": 2.77}, {\"category\": \"Car\", \"angle\": 2.95, \"box_2d\": [443.6, 178.49, 494.3, 197.17], \"box_3d\": [1.59, 1.76, 4.18, -12.63, 2.12, 64.51], \"ry\": 2.76}]\n```" - } - ] - }, - { - "id": 552, - "data_source": "Mono3DRefer", - "scene_name": "001578", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.29, \"box_2d\": [781.21, 181.18, 868.84, 234.35], \"box_3d\": [1.45, 1.51, 3.7, 6.33, 1.71, 21.74], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 2.95, \"box_2d\": [655.2, 173.5, 749.38, 206.86], \"box_3d\": [1.51, 1.65, 4.21, 4.34, 1.57, 34.49], \"ry\": 3.07}, {\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [575.24, 176.89, 603.89, 203.29], \"box_3d\": [1.43, 1.55, 3.91, -1.16, 1.68, 41.43], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.22, \"box_2d\": [797.65, 184.73, 1051.5, 332.66], \"box_3d\": [1.52, 1.61, 4.23, 3.82, 1.7, 9.73], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 553, - "data_source": "Mono3DRefer", - "scene_name": "001580", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [579.33, 182.18, 637.23, 229.54], \"box_3d\": [1.36, 1.69, 3.38, -0.1, 1.67, 22.88], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -2.19, \"box_2d\": [271.96, 183.99, 440.07, 255.83], \"box_3d\": [1.45, 1.61, 3.96, -5.91, 1.72, 16.51], \"ry\": -2.53}, {\"category\": \"Car\", \"angle\": -2.21, \"box_2d\": [340.72, 187.76, 475.06, 244.18], \"box_3d\": [1.36, 1.57, 3.73, -5.55, 1.78, 19.48], \"ry\": -2.48}, {\"category\": \"Car\", \"angle\": 1.0, \"box_2d\": [401.06, 181.15, 488.34, 223.47], \"box_3d\": [1.48, 1.57, 3.48, -6.32, 1.81, 27.23], \"ry\": 0.78}, {\"category\": \"Car\", \"angle\": -1.47, \"box_2d\": [505.27, 187.23, 566.64, 227.74], \"box_3d\": [1.25, 1.74, 3.66, -2.53, 1.78, 25.03], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [669.93, 179.59, 751.31, 236.35], \"box_3d\": [1.46, 1.72, 3.95, 2.75, 1.67, 20.78], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [630.77, 179.16, 657.78, 200.21], \"box_3d\": [1.38, 1.55, 3.6, 2.3, 1.83, 49.64], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 2.08, \"box_2d\": [105.31, 189.85, 264.4, 260.12], \"box_3d\": [1.35, 1.59, 3.41, -9.39, 1.76, 16.17], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [589.65, 178.27, 628.69, 213.9], \"box_3d\": [1.52, 1.67, 3.61, -0.09, 1.78, 32.86], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 2.1, \"box_2d\": [88.49, 185.03, 196.54, 231.02], \"box_3d\": [1.32, 1.43, 3.25, -14.8, 1.73, 22.96], \"ry\": 1.53}]\n```" - } - ] - }, - { - "id": 554, - "data_source": "Mono3DRefer", - "scene_name": "001582", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [453.27, 182.24, 544.5, 254.44], \"box_3d\": [1.46, 1.63, 4.47, -2.54, 1.7, 17.17], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [519.58, 168.82, 584.9, 215.89], \"box_3d\": [1.7, 1.82, 4.45, -2.22, 1.58, 28.46], \"ry\": 1.66}, {\"category\": \"Car\", \"angle\": -0.84, \"box_2d\": [620.59, 171.61, 671.8, 200.04], \"box_3d\": [1.57, 1.56, 2.62, 2.06, 1.52, 41.37], \"ry\": -0.79}]\n```" - } - ] - }, - { - "id": 555, - "data_source": "Mono3DRefer", - "scene_name": "001583", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.59, \"box_2d\": [36.4, 178.67, 323.82, 286.16], \"box_3d\": [1.5, 1.58, 3.47, -6.49, 1.61, 11.05], \"ry\": -3.11}, {\"category\": \"Car\", \"angle\": 0.41, \"box_2d\": [186.72, 157.67, 431.3, 255.6], \"box_3d\": [1.95, 1.64, 4.57, -6.38, 1.67, 15.48], \"ry\": 0.03}, {\"category\": \"Car\", \"angle\": -2.7, \"box_2d\": [153.89, 172.97, 375.94, 271.23], \"box_3d\": [1.66, 1.58, 3.31, -6.26, 1.69, 13.23], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [700.09, 179.46, 826.08, 262.58], \"box_3d\": [1.44, 1.66, 3.25, 2.85, 1.58, 14.28], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -2.79, \"box_2d\": [254.13, 161.69, 422.78, 243.01], \"box_3d\": [1.87, 1.61, 3.53, -6.62, 1.64, 17.73], \"ry\": -3.14}, {\"category\": \"Car\", \"angle\": 0.43, \"box_2d\": [310.93, 175.39, 457.48, 236.24], \"box_3d\": [1.59, 1.56, 3.5, 0.52, 1.68, 19.82], \"ry\": 0.13}, {\"category\": \"Car\", \"angle\": -2.82, \"box_2d\": [316.35, 170.99, 471.34, 229.45], \"box_3d\": [1.72, 1.65, 4.33, 0.0, 1.69, 22.54], \"ry\": -3.1}, {\"category\": \"Car\", \"angle\": 0.28, \"box_2d\": [399.71, 171.28, 506.84, 210.62], \"box_3d\": [1.57, 1.54, 4.1, -6.55, 1.53, 30.21], \"ry\": 0.07}, {\"category\": \"Car\", \"angle\": -2.95, \"box_2d\": [433.66, 164.2, 519.77, 204.83], \"box_3d\": [1.89, 1.59, 3.87, -6.46, 1.51, 35.07], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [649.54, 174.79, 694.8, 210.17], \"box_3d\": [1.45, 1.59, 3.59, 2.65, 1.56, 31.78], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [677.9, 175.42, 759.76, 238.11], \"box_3d\": [1.54, 1.67, 3.72, 2.84, 1.63, 19.77], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [638.05, 169.71, 671.54, 199.45], \"box_3d\": [1.66, 1.74, 3.65, 2.57, 1.51, 42.53], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.86, \"box_2d\": [734.87, 155.48, 1001.33, 345.26], \"box_3d\": [1.76, 1.71, 4.35, 2.75, 1.63, 8.92], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -2.46, \"box_2d\": [0.0, 189.52, 271.67, 333.04], \"box_3d\": [1.49, 1.61, 3.87, -6.35, 1.72, 8.5], \"ry\": -3.09}]\n```" - } - ] - }, - { - "id": 556, - "data_source": "Mono3DRefer", - "scene_name": "001584", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.95, \"box_2d\": [785.11, 193.9, 1221.92, 374.0], \"box_3d\": [1.41, 1.64, 3.77, 2.66, 1.63, 6.01], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [477.25, 184.13, 532.08, 220.48], \"box_3d\": [1.43, 1.76, 3.97, -4.52, 1.94, 31.21], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 2.02, \"box_2d\": [132.55, 199.4, 379.72, 334.14], \"box_3d\": [1.49, 1.61, 3.94, -4.9, 1.94, 10.56], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [309.39, 191.21, 450.68, 274.0], \"box_3d\": [1.48, 1.66, 3.9, -4.8, 1.91, 15.49], \"ry\": 1.61}]\n```" - } - ] - }, - { - "id": 557, - "data_source": "Mono3DRefer", - "scene_name": "001588", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.0, \"box_2d\": [314.33, 185.35, 369.1, 211.79], \"box_3d\": [1.45, 1.74, 3.42, -15.75, 2.21, 42.43], \"ry\": 1.65}]\n```" - } - ] - }, - { - "id": 558, - "data_source": "Mono3DRefer", - "scene_name": "001593", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.16, \"box_2d\": [0.0, 215.25, 377.01, 374.0], \"box_3d\": [1.44, 1.63, 3.32, -2.58, 1.75, 3.63], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -2.31, \"box_2d\": [904.49, 203.93, 1241.0, 374.0], \"box_3d\": [1.55, 1.55, 3.27, 2.77, 1.75, 3.14], \"ry\": -1.64}, {\"category\": \"Car\", \"angle\": -1.85, \"box_2d\": [732.71, 198.08, 987.43, 374.0], \"box_3d\": [1.48, 1.6, 3.79, 2.48, 1.83, 8.24], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [470.51, 192.98, 554.49, 261.94], \"box_3d\": [1.56, 1.55, 3.64, -2.49, 2.13, 18.91], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [519.79, 178.96, 583.62, 241.11], \"box_3d\": [1.91, 1.66, 4.22, -1.92, 2.12, 24.41], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [559.51, 179.97, 611.5, 226.96], \"box_3d\": [1.83, 1.69, 4.28, -0.99, 2.14, 30.38], \"ry\": 1.65}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [619.92, 190.28, 660.62, 216.34], \"box_3d\": [1.51, 1.63, 4.03, 1.95, 2.66, 45.8], \"ry\": 1.86}]\n```" - } - ] - }, - { - "id": 559, - "data_source": "Mono3DRefer", - "scene_name": "001598", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [529.66, 171.82, 555.18, 194.65], \"box_3d\": [1.65, 1.56, 3.98, -5.09, 1.6, 54.53], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 560, - "data_source": "Mono3DRefer", - "scene_name": "001604", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [591.78, 176.65, 634.94, 217.45], \"box_3d\": [1.56, 1.65, 3.69, 0.11, 1.73, 29.7], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 1.97, \"box_2d\": [315.91, 183.97, 388.42, 220.4], \"box_3d\": [1.4, 1.54, 3.67, -10.75, 1.88, 30.23], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [362.22, 180.22, 422.8, 214.34], \"box_3d\": [1.56, 1.7, 4.41, -10.68, 1.94, 35.63], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -0.89, \"box_2d\": [671.6, 173.81, 751.11, 210.59], \"box_3d\": [1.53, 1.57, 3.54, 4.47, 1.58, 31.82], \"ry\": -0.75}, {\"category\": \"Car\", \"angle\": -0.93, \"box_2d\": [668.46, 176.2, 745.19, 208.99], \"box_3d\": [1.5, 1.64, 3.99, 4.74, 1.68, 35.2], \"ry\": -0.8}, {\"category\": \"Car\", \"angle\": -0.87, \"box_2d\": [649.87, 176.46, 724.45, 204.46], \"box_3d\": [1.4, 1.65, 4.18, 4.15, 1.62, 38.66], \"ry\": -0.76}]\n```" - } - ] - }, - { - "id": 561, - "data_source": "Mono3DRefer", - "scene_name": "001606", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [569.32, 177.9, 615.36, 214.86], \"box_3d\": [1.46, 1.63, 4.47, -0.73, 1.7, 31.13], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [594.12, 169.32, 637.85, 200.14], \"box_3d\": [1.7, 1.82, 4.45, 0.39, 1.53, 42.31], \"ry\": 1.75}, {\"category\": \"Car\", \"angle\": -2.86, \"box_2d\": [309.3, 173.8, 495.92, 236.16], \"box_3d\": [1.47, 1.68, 4.26, -5.21, 1.52, 18.36], \"ry\": -3.14}]\n```" - } - ] - }, - { - "id": 562, - "data_source": "Mono3DRefer", - "scene_name": "001610", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.03, \"box_2d\": [796.85, 184.16, 1241.0, 374.0], \"box_3d\": [1.49, 1.71, 3.58, 2.84, 1.6, 5.86], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.28, \"box_2d\": [258.92, 201.75, 435.77, 307.55], \"box_3d\": [1.39, 1.81, 3.46, -4.23, 1.94, 12.09], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [683.79, 181.03, 782.86, 255.45], \"box_3d\": [1.47, 1.55, 3.37, 2.61, 1.66, 16.15], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.36, \"box_2d\": [407.04, 183.36, 498.75, 248.85], \"box_3d\": [1.72, 1.8, 4.16, -4.56, 2.05, 21.38], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [486.83, 183.19, 530.45, 218.86], \"box_3d\": [1.62, 1.66, 3.81, -4.94, 2.14, 35.34], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [518.1, 184.21, 548.86, 206.4], \"box_3d\": [1.46, 1.67, 4.92, -5.42, 2.29, 51.42], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [648.45, 175.54, 695.7, 214.56], \"box_3d\": [1.61, 1.67, 3.62, 2.65, 1.75, 31.89], \"ry\": -1.6}]\n```" - } - ] - }, - { - "id": 563, - "data_source": "Mono3DRefer", - "scene_name": "001619", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.83, \"box_2d\": [727.62, 181.95, 929.59, 305.85], \"box_3d\": [1.41, 1.58, 4.36, 2.84, 1.55, 10.54], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 564, - "data_source": "Mono3DRefer", - "scene_name": "001626", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -3.02, \"box_2d\": [448.62, 185.48, 517.58, 209.22], \"box_3d\": [1.47, 1.74, 4.26, -8.31, 2.33, 47.36], \"ry\": 3.09}]\n```" - } - ] - }, - { - "id": 565, - "data_source": "Mono3DRefer", - "scene_name": "001629", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [617.16, 181.63, 703.89, 260.51], \"box_3d\": [1.46, 1.53, 3.17, 0.92, 1.65, 15.14], \"ry\": -1.59}]\n```" - } - ] - }, - { - "id": 566, - "data_source": "Mono3DRefer", - "scene_name": "001634", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.27, \"box_2d\": [987.33, 189.54, 1241.0, 374.0], \"box_3d\": [1.41, 1.57, 3.16, 2.88, 1.5, 2.67], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": 2.26, \"box_2d\": [0.0, 214.97, 197.03, 374.0], \"box_3d\": [1.53, 1.56, 3.58, -5.99, 2.04, 7.19], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [430.63, 188.45, 485.45, 226.37], \"box_3d\": [1.42, 1.54, 3.5, -6.24, 2.09, 29.78], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [650.53, 174.55, 695.9, 209.69], \"box_3d\": [1.51, 1.75, 3.78, 2.83, 1.62, 33.32], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [458.42, 186.28, 505.31, 216.88], \"box_3d\": [1.36, 1.51, 4.02, -6.22, 2.04, 35.23], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [489.59, 179.45, 526.62, 209.94], \"box_3d\": [1.74, 1.65, 4.11, -6.13, 2.15, 43.57], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -0.97, \"box_2d\": [66.41, 204.45, 153.77, 236.54], \"box_3d\": [1.43, 1.8, 3.96, -25.92, 3.15, 37.54], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.02, \"box_2d\": [118.77, 202.56, 197.3, 233.74], \"box_3d\": [1.4, 1.6, 3.97, -23.46, 3.02, 37.57], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.05, \"box_2d\": [159.73, 198.14, 243.17, 231.88], \"box_3d\": [1.49, 1.7, 4.51, -20.84, 2.86, 36.98], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.11, \"box_2d\": [216.31, 198.79, 289.44, 232.43], \"box_3d\": [1.5, 1.56, 4.45, -18.33, 2.9, 37.19], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 567, - "data_source": "Mono3DRefer", - "scene_name": "001639", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.44, \"box_2d\": [795.49, 174.34, 860.8, 201.49], \"box_3d\": [1.48, 1.36, 3.51, 12.41, 1.59, 41.3], \"ry\": -2.15}]\n```" - } - ] - }, - { - "id": 568, - "data_source": "Mono3DRefer", - "scene_name": "001646", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.92, \"box_2d\": [669.02, 179.16, 768.17, 218.35], \"box_3d\": [1.4, 1.65, 4.18, 4.3, 1.67, 28.4], \"ry\": -0.77}, {\"category\": \"Car\", \"angle\": -0.92, \"box_2d\": [660.41, 176.69, 745.31, 215.51], \"box_3d\": [1.59, 1.65, 3.94, 4.11, 1.78, 31.82], \"ry\": -0.79}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [594.28, 178.84, 636.54, 218.8], \"box_3d\": [1.56, 1.65, 3.69, 0.18, 1.82, 30.3], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.1, \"box_2d\": [154.91, 190.02, 286.82, 248.18], \"box_3d\": [1.4, 1.54, 3.67, -10.66, 1.91, 19.98], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [256.97, 183.17, 356.95, 232.68], \"box_3d\": [1.56, 1.7, 4.41, -10.59, 1.94, 25.48], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -0.96, \"box_2d\": [707.05, 174.7, 820.99, 230.5], \"box_3d\": [1.53, 1.57, 3.54, 4.63, 1.6, 21.62], \"ry\": -0.75}, {\"category\": \"Car\", \"angle\": -0.93, \"box_2d\": [695.77, 177.37, 805.38, 224.96], \"box_3d\": [1.5, 1.64, 3.99, 4.88, 1.67, 24.93], \"ry\": -0.74}]\n```" - } - ] - }, - { - "id": 569, - "data_source": "Mono3DRefer", - "scene_name": "001652", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.44, \"box_2d\": [603.95, 186.24, 646.19, 216.13], \"box_3d\": [1.44, 1.66, 4.53, 0.83, 2.17, 38.19], \"ry\": -1.42}]\n```" - } - ] - }, - { - "id": 570, - "data_source": "Mono3DRefer", - "scene_name": "001656", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.2, \"box_2d\": [853.47, 200.96, 1241.0, 374.0], \"box_3d\": [1.38, 1.58, 3.66, 2.5, 1.57, 3.28], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.26, \"box_2d\": [727.09, 185.84, 999.42, 357.62], \"box_3d\": [1.42, 1.6, 3.78, 2.43, 1.59, 8.05], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 2.21, \"box_2d\": [517.99, 176.68, 685.76, 257.05], \"box_3d\": [1.57, 1.65, 4.1, -0.09, 1.68, 16.35], \"ry\": 2.2}]\n```" - } - ] - }, - { - "id": 571, - "data_source": "Mono3DRefer", - "scene_name": "001662", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.92, \"box_2d\": [875.07, 106.85, 1241.0, 374.0], \"box_3d\": [1.79, 1.72, 4.35, 3.65, 1.49, 5.18], \"ry\": 1.5}, {\"category\": \"Car\", \"angle\": 2.07, \"box_2d\": [0.0, 207.76, 337.87, 374.0], \"box_3d\": [1.45, 1.49, 4.07, -3.94, 1.84, 6.42], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [351.92, 197.74, 469.76, 274.84], \"box_3d\": [1.27, 1.54, 3.58, -3.94, 1.83, 14.65], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [424.68, 179.27, 505.3, 242.2], \"box_3d\": [1.63, 1.65, 3.91, -4.13, 1.83, 20.9], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [603.45, 177.11, 637.32, 207.61], \"box_3d\": [1.46, 1.53, 3.17, 0.46, 1.69, 36.52], \"ry\": -1.61}]\n```" - } - ] - }, - { - "id": 572, - "data_source": "Mono3DRefer", - "scene_name": "001666", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.95, \"box_2d\": [762.07, 184.98, 1125.99, 374.0], \"box_3d\": [1.56, 1.57, 3.98, 2.7, 1.7, 7.0], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -2.84, \"box_2d\": [358.14, 177.26, 452.16, 218.74], \"box_3d\": [1.61, 1.57, 3.35, -8.33, 1.81, 29.37], \"ry\": -3.11}, {\"category\": \"Car\", \"angle\": 1.39, \"box_2d\": [686.54, 166.43, 729.36, 198.46], \"box_3d\": [1.79, 1.72, 4.35, 5.73, 1.46, 42.86], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": 1.58, \"box_2d\": [564.78, 178.69, 590.74, 203.86], \"box_3d\": [1.45, 1.49, 4.07, -1.97, 1.82, 43.99], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 573, - "data_source": "Mono3DRefer", - "scene_name": "001675", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.0, \"box_2d\": [245.71, 192.25, 326.34, 227.95], \"box_3d\": [1.46, 1.86, 4.01, -14.81, 2.39, 33.14], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 574, - "data_source": "Mono3DRefer", - "scene_name": "001677", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.16, \"box_2d\": [0.0, 195.97, 226.01, 293.89], \"box_3d\": [1.48, 1.6, 3.9, -9.14, 1.97, 13.61], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [376.64, 182.08, 441.96, 220.0], \"box_3d\": [1.61, 1.67, 4.56, -9.23, 2.05, 33.42], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 575, - "data_source": "Mono3DRefer", - "scene_name": "001685", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.8, \"box_2d\": [815.22, 172.08, 911.67, 236.95], \"box_3d\": [1.55, 1.57, 3.64, 6.59, 1.56, 19.23], \"ry\": -1.48}, {\"category\": \"Car\", \"angle\": -1.93, \"box_2d\": [899.78, 174.45, 1017.82, 231.57], \"box_3d\": [1.33, 1.47, 3.83, 8.91, 1.39, 18.85], \"ry\": -1.49}, {\"category\": \"Car\", \"angle\": 0.64, \"box_2d\": [0.0, 188.25, 140.87, 245.0], \"box_3d\": [1.76, 1.56, 4.01, -18.08, 2.3, 23.87], \"ry\": -0.0}]\n```" - } - ] - }, - { - "id": 576, - "data_source": "Mono3DRefer", - "scene_name": "001691", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [694.29, 181.46, 721.01, 205.97], \"box_3d\": [1.54, 1.63, 3.67, 6.5, 2.13, 47.99], \"ry\": -1.4}, {\"category\": \"Car\", \"angle\": -1.91, \"box_2d\": [775.56, 182.64, 1040.16, 349.6], \"box_3d\": [1.58, 1.67, 3.84, 3.27, 1.72, 8.88], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [548.44, 183.54, 614.62, 229.98], \"box_3d\": [1.53, 1.67, 3.78, -0.99, 1.93, 26.06], \"ry\": 1.73}, {\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [753.01, 153.67, 876.6, 281.05], \"box_3d\": [2.05, 1.5, 3.83, 3.65, 1.77, 13.6], \"ry\": -1.49}]\n```" - } - ] - }, - { - "id": 577, - "data_source": "Mono3DRefer", - "scene_name": "001698", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [536.09, 179.83, 562.46, 201.92], \"box_3d\": [1.45, 1.53, 3.98, -4.2, 1.95, 50.02], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [469.01, 178.69, 529.97, 227.77], \"box_3d\": [1.65, 1.56, 3.98, -4.0, 1.87, 26.42], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 578, - "data_source": "Mono3DRefer", - "scene_name": "001703", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.54, \"box_2d\": [529.6, 182.07, 563.56, 216.64], \"box_3d\": [1.68, 1.63, 3.9, -3.38, 2.17, 37.61], \"ry\": 1.45}, {\"category\": \"Car\", \"angle\": 1.44, \"box_2d\": [500.24, 183.64, 528.7, 202.99], \"box_3d\": [1.47, 1.7, 4.39, -7.74, 2.35, 58.02], \"ry\": 1.31}]\n```" - } - ] - }, - { - "id": 579, - "data_source": "Mono3DRefer", - "scene_name": "001715", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [541.08, 179.47, 571.06, 201.81], \"box_3d\": [1.44, 1.72, 4.14, -3.66, 1.91, 49.29], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 580, - "data_source": "Mono3DRefer", - "scene_name": "001727", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.02, \"box_2d\": [200.64, 188.13, 300.08, 235.77], \"box_3d\": [1.6, 1.79, 3.91, -13.42, 2.21, 27.12], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 581, - "data_source": "Mono3DRefer", - "scene_name": "001728", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.08, \"box_2d\": [91.01, 200.84, 222.95, 259.01], \"box_3d\": [1.35, 1.5, 3.64, -12.48, 2.19, 20.08], \"ry\": 1.53}]\n```" - } - ] - }, - { - "id": 582, - "data_source": "Mono3DRefer", - "scene_name": "001729", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.2, \"box_2d\": [1025.57, 211.79, 1241.0, 374.0], \"box_3d\": [1.4, 1.64, 3.73, 3.34, 1.65, 3.0], \"ry\": -1.41}, {\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [786.34, 187.45, 849.86, 232.24], \"box_3d\": [1.4, 1.61, 4.37, 7.26, 1.95, 25.65], \"ry\": -1.42}]\n```" - } - ] - }, - { - "id": 583, - "data_source": "Mono3DRefer", - "scene_name": "001733", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [634.65, 179.26, 673.28, 211.63], \"box_3d\": [1.61, 1.67, 3.62, 2.24, 1.96, 38.07], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.82, \"box_2d\": [703.35, 182.68, 865.92, 289.41], \"box_3d\": [1.49, 1.71, 3.58, 2.68, 1.67, 12.08], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.39, \"box_2d\": [380.31, 193.92, 479.25, 258.68], \"box_3d\": [1.39, 1.81, 3.46, -4.47, 1.97, 18.21], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [656.98, 182.59, 723.24, 235.11], \"box_3d\": [1.47, 1.55, 3.37, 2.36, 1.79, 22.28], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.4, \"box_2d\": [448.34, 182.73, 515.19, 232.26], \"box_3d\": [1.72, 1.8, 4.16, -4.84, 2.11, 27.54], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [497.47, 184.78, 534.26, 215.11], \"box_3d\": [1.62, 1.66, 3.81, -5.38, 2.33, 41.43], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 584, - "data_source": "Mono3DRefer", - "scene_name": "001739", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.54, \"box_2d\": [537.25, 189.24, 569.33, 222.81], \"box_3d\": [1.68, 1.63, 3.9, -3.18, 2.61, 39.54], \"ry\": 1.46}, {\"category\": \"Car\", \"angle\": 1.46, \"box_2d\": [506.07, 189.61, 530.59, 207.34], \"box_3d\": [1.47, 1.7, 4.39, -8.27, 3.0, 64.72], \"ry\": 1.33}]\n```" - } - ] - }, - { - "id": 585, - "data_source": "Mono3DRefer", - "scene_name": "001750", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.19, \"box_2d\": [0.0, 192.42, 148.33, 258.77], \"box_3d\": [1.48, 1.61, 3.39, -13.83, 2.03, 18.61], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 2.23, \"box_2d\": [0.0, 191.38, 107.95, 249.14], \"box_3d\": [1.44, 1.79, 4.32, -16.82, 2.01, 20.69], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [278.39, 183.08, 348.83, 219.49], \"box_3d\": [1.6, 1.78, 3.86, -13.99, 2.1, 34.2], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [241.18, 181.83, 311.79, 216.76], \"box_3d\": [1.65, 1.77, 3.85, -16.82, 2.12, 36.51], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [410.39, 179.97, 446.93, 201.04], \"box_3d\": [1.58, 1.82, 3.87, -14.2, 2.15, 56.62], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 586, - "data_source": "Mono3DRefer", - "scene_name": "001752", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [427.03, 183.84, 459.93, 201.11], \"box_3d\": [1.46, 1.86, 4.01, -14.84, 2.46, 64.43], \"ry\": 1.61}]\n```" - } - ] - }, - { - "id": 587, - "data_source": "Mono3DRefer", - "scene_name": "001759", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [635.76, 174.97, 667.37, 200.28], \"box_3d\": [1.5, 1.62, 3.89, 2.54, 1.66, 45.3], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [684.66, 170.61, 764.99, 233.79], \"box_3d\": [1.69, 1.58, 3.95, 3.28, 1.66, 21.5], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": -1.85, \"box_2d\": [755.79, 185.2, 908.81, 283.88], \"box_3d\": [1.41, 1.57, 3.23, 3.54, 1.64, 12.19], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -2.14, \"box_2d\": [946.12, 182.23, 1209.36, 288.27], \"box_3d\": [1.45, 1.61, 4.2, 7.46, 1.62, 12.16], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -2.52, \"box_2d\": [820.71, 175.86, 891.13, 204.14], \"box_3d\": [1.48, 1.36, 3.51, 13.47, 1.67, 39.73], \"ry\": -2.2}]\n```" - } - ] - }, - { - "id": 588, - "data_source": "Mono3DRefer", - "scene_name": "001762", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.14, \"box_2d\": [81.75, 190.51, 214.47, 243.57], \"box_3d\": [1.56, 1.7, 4.35, -15.52, 2.2, 24.48], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [317.82, 183.64, 379.79, 217.04], \"box_3d\": [1.6, 1.79, 3.91, -13.45, 2.18, 37.31], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 589, - "data_source": "Mono3DRefer", - "scene_name": "001772", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.23, \"box_2d\": [0.0, 194.3, 349.61, 374.0], \"box_3d\": [1.55, 1.68, 4.42, -2.79, 1.7, 3.16], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": -1.85, \"box_2d\": [717.45, 176.93, 914.92, 314.61], \"box_3d\": [1.6, 1.53, 4.18, 2.69, 1.67, 10.56], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [672.86, 178.84, 781.32, 258.58], \"box_3d\": [1.56, 1.57, 3.98, 2.46, 1.71, 16.33], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": 1.41, \"box_2d\": [663.95, 167.76, 697.64, 193.79], \"box_3d\": [1.79, 1.72, 4.35, 5.06, 1.47, 52.23], \"ry\": 1.5}]\n```" - } - ] - }, - { - "id": 590, - "data_source": "Mono3DRefer", - "scene_name": "001778", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [487.04, 175.06, 560.72, 233.28], \"box_3d\": [1.46, 1.63, 4.47, -2.39, 1.55, 20.61], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [534.48, 165.26, 593.34, 206.92], \"box_3d\": [1.7, 1.82, 4.45, -1.98, 1.42, 31.89], \"ry\": 1.69}, {\"category\": \"Car\", \"angle\": -0.62, \"box_2d\": [607.24, 166.76, 657.74, 193.53], \"box_3d\": [1.57, 1.56, 2.62, 1.34, 1.22, 43.74], \"ry\": -0.59}, {\"category\": \"Car\", \"angle\": -2.55, \"box_2d\": [0.0, 176.38, 278.1, 332.98], \"box_3d\": [1.47, 1.68, 4.26, -6.11, 1.53, 7.69], \"ry\": 3.08}]\n```" - } - ] - }, - { - "id": 591, - "data_source": "Mono3DRefer", - "scene_name": "001779", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.4, \"box_2d\": [0.0, 194.78, 220.47, 374.0], \"box_3d\": [1.47, 1.67, 4.28, -6.37, 1.72, 6.91], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": -2.79, \"box_2d\": [48.82, 190.67, 366.54, 285.48], \"box_3d\": [1.38, 1.67, 4.37, -6.59, 1.71, 12.17], \"ry\": 3.0}, {\"category\": \"Car\", \"angle\": -1.01, \"box_2d\": [1194.01, 166.76, 1241.0, 374.0], \"box_3d\": [1.54, 1.58, 3.2, 4.71, 1.49, 3.14], \"ry\": -0.07}, {\"category\": \"Car\", \"angle\": -2.13, \"box_2d\": [1004.94, 155.99, 1192.67, 258.77], \"box_3d\": [1.54, 1.44, 2.54, 8.04, 1.31, 12.17], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -2.3, \"box_2d\": [1115.1, 151.35, 1241.0, 254.6], \"box_3d\": [1.59, 1.8, 3.41, 10.53, 1.26, 12.06], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 0.56, \"box_2d\": [157.98, 184.38, 341.44, 240.76], \"box_3d\": [1.42, 1.41, 4.3, -9.61, 1.75, 19.26], \"ry\": 0.11}]\n```" - } - ] - }, - { - "id": 592, - "data_source": "Mono3DRefer", - "scene_name": "001783", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.57, \"box_2d\": [605.42, 181.13, 634.74, 208.5], \"box_3d\": [1.5, 1.62, 4.19, 0.56, 2.0, 42.25], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 593, - "data_source": "Mono3DRefer", - "scene_name": "001784", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [188.9, 193.27, 446.76, 351.87], \"box_3d\": [1.44, 1.62, 3.67, -3.27, 1.73, 8.76], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [430.05, 181.92, 525.87, 252.98], \"box_3d\": [1.51, 1.62, 3.68, -3.12, 1.75, 17.46], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 594, - "data_source": "Mono3DRefer", - "scene_name": "001807", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [379.68, 184.9, 424.23, 209.63], \"box_3d\": [1.48, 1.71, 3.54, -13.26, 2.27, 46.1], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 595, - "data_source": "Mono3DRefer", - "scene_name": "001809", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [513.12, 177.54, 548.99, 208.95], \"box_3d\": [1.7, 1.72, 4.11, -4.52, 1.99, 41.53], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 596, - "data_source": "Mono3DRefer", - "scene_name": "001811", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.59, \"box_2d\": [680.88, 174.29, 845.64, 248.43], \"box_3d\": [1.62, 1.47, 3.66, 3.64, 1.67, 17.34], \"ry\": -0.38}, {\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [337.45, 200.21, 526.35, 353.54], \"box_3d\": [1.38, 1.43, 3.64, -2.0, 1.78, 8.87], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 597, - "data_source": "Mono3DRefer", - "scene_name": "001824", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.04, \"box_2d\": [199.72, 188.3, 291.62, 229.29], \"box_3d\": [1.56, 1.7, 4.35, -15.46, 2.26, 30.82], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [364.04, 182.24, 413.35, 210.1], \"box_3d\": [1.6, 1.79, 3.91, -13.5, 2.19, 44.15], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 598, - "data_source": "Mono3DRefer", - "scene_name": "001829", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [984.27, 173.21, 1030.24, 203.05], \"box_3d\": [1.46, 1.74, 3.99, 20.59, 1.49, 37.38], \"ry\": 2.16}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [791.69, 180.56, 888.5, 233.92], \"box_3d\": [1.42, 1.68, 4.29, 7.0, 1.67, 21.78], \"ry\": 2.15}, {\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [870.14, 181.23, 929.61, 219.4], \"box_3d\": [1.36, 1.4, 3.8, 11.32, 1.7, 28.13], \"ry\": 2.16}, {\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [934.38, 177.11, 971.43, 204.61], \"box_3d\": [1.44, 1.65, 2.96, 18.84, 1.69, 39.64], \"ry\": 2.07}, {\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [709.74, 180.08, 836.63, 253.17], \"box_3d\": [1.48, 1.51, 4.35, 3.94, 1.67, 17.04], \"ry\": 2.14}]\n```" - } - ] - }, - { - "id": 599, - "data_source": "Mono3DRefer", - "scene_name": "001832", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [301.96, 181.15, 511.37, 329.23], \"box_3d\": [1.57, 1.54, 4.01, -2.5, 1.69, 9.72], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 2.21, \"box_2d\": [0.0, 181.49, 313.33, 374.0], \"box_3d\": [1.65, 1.75, 3.39, -2.52, 1.69, 2.5], \"ry\": 1.47}]\n```" - } - ] - }, - { - "id": 600, - "data_source": "Mono3DRefer", - "scene_name": "001842", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.47, \"box_2d\": [610.43, 177.52, 638.81, 200.51], \"box_3d\": [1.54, 1.63, 3.67, 1.04, 1.89, 50.75], \"ry\": -1.45}, {\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [450.64, 183.63, 504.73, 215.6], \"box_3d\": [1.56, 1.59, 3.69, -6.88, 2.14, 37.71], \"ry\": 1.74}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [479.0, 175.05, 525.53, 209.84], \"box_3d\": [1.93, 1.59, 4.01, -6.29, 2.08, 42.34], \"ry\": 1.72}, {\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [657.28, 174.48, 685.44, 200.55], \"box_3d\": [1.58, 1.67, 3.84, 3.87, 1.71, 46.17], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": 2.76, \"box_2d\": [946.64, 163.36, 1032.23, 192.21], \"box_3d\": [1.39, 1.51, 3.49, 19.06, 0.95, 36.4], \"ry\": -3.04}]\n```" - } - ] - }, - { - "id": 601, - "data_source": "Mono3DRefer", - "scene_name": "001856", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.3, \"box_2d\": [268.24, 192.18, 488.76, 324.62], \"box_3d\": [1.33, 1.71, 4.09, -2.87, 1.64, 9.78], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [654.58, 169.64, 721.8, 219.87], \"box_3d\": [1.49, 1.56, 4.34, 2.45, 1.43, 23.82], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.41, \"box_2d\": [444.19, 178.7, 530.52, 241.17], \"box_3d\": [1.42, 1.53, 4.12, -3.1, 1.59, 18.7], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [705.57, 169.71, 876.51, 293.87], \"box_3d\": [1.59, 1.59, 3.89, 2.57, 1.58, 11.33], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [630.36, 171.44, 661.62, 197.3], \"box_3d\": [1.38, 1.64, 3.51, 2.03, 1.33, 40.58], \"ry\": -1.49}]\n```" - } - ] - }, - { - "id": 602, - "data_source": "Mono3DRefer", - "scene_name": "001860", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.4, \"box_2d\": [0.0, 174.15, 122.61, 318.73], \"box_3d\": [1.69, 1.77, 3.51, -8.71, 1.73, 9.32], \"ry\": -3.14}, {\"category\": \"Car\", \"angle\": -2.57, \"box_2d\": [0.0, 185.07, 254.96, 290.84], \"box_3d\": [1.61, 1.66, 3.86, -8.37, 1.84, 12.15], \"ry\": 3.11}, {\"category\": \"Car\", \"angle\": 3.04, \"box_2d\": [146.68, 180.65, 342.92, 259.2], \"box_3d\": [1.59, 1.71, 3.9, -8.47, 1.8, 16.74], \"ry\": 2.57}, {\"category\": \"Car\", \"angle\": 2.94, \"box_2d\": [193.59, 179.49, 364.02, 252.01], \"box_3d\": [1.73, 1.69, 3.88, -8.85, 1.93, 19.3], \"ry\": 2.52}, {\"category\": \"Car\", \"angle\": 3.02, \"box_2d\": [312.4, 179.28, 393.38, 223.97], \"box_3d\": [1.43, 1.57, 2.43, -8.84, 1.67, 24.68], \"ry\": 2.68}, {\"category\": \"Car\", \"angle\": 3.07, \"box_2d\": [338.95, 179.16, 445.81, 221.14], \"box_3d\": [1.48, 1.74, 3.77, -8.33, 1.74, 27.46], \"ry\": 2.78}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [456.44, 178.68, 496.4, 208.54], \"box_3d\": [1.55, 1.54, 3.06, -7.28, 1.88, 39.37], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 603, - "data_source": "Mono3DRefer", - "scene_name": "001866", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -3.0, \"box_2d\": [427.53, 187.28, 505.74, 214.12], \"box_3d\": [1.47, 1.74, 4.26, -8.35, 2.34, 42.15], \"ry\": 3.09}]\n```" - } - ] - }, - { - "id": 604, - "data_source": "Mono3DRefer", - "scene_name": "001867", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.28, \"box_2d\": [0.0, 184.99, 301.76, 374.0], \"box_3d\": [1.48, 1.64, 3.95, -3.3, 1.56, 3.66], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -2.17, \"box_2d\": [867.24, 207.12, 1241.0, 374.0], \"box_3d\": [1.35, 1.64, 3.84, 2.74, 1.6, 3.62], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 1.37, \"box_2d\": [694.62, 183.25, 824.53, 270.59], \"box_3d\": [1.42, 1.59, 4.19, 2.69, 1.64, 14.08], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [381.45, 182.87, 501.05, 263.9], \"box_3d\": [1.42, 1.64, 3.63, -3.34, 1.64, 14.71], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [672.15, 179.79, 748.54, 234.79], \"box_3d\": [1.38, 1.58, 3.66, 2.66, 1.58, 20.11], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.45, \"box_2d\": [659.26, 176.02, 719.76, 221.17], \"box_3d\": [1.42, 1.6, 3.78, 2.62, 1.55, 24.82], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 2.81, \"box_2d\": [776.87, 165.28, 864.85, 196.73], \"box_3d\": [1.57, 1.65, 4.1, 11.01, 1.21, 37.93], \"ry\": 3.09}]\n```" - } - ] - }, - { - "id": 605, - "data_source": "Mono3DRefer", - "scene_name": "001871", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [519.67, 182.57, 559.91, 212.03], \"box_3d\": [1.44, 1.72, 4.14, -3.69, 1.97, 38.19], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 606, - "data_source": "Mono3DRefer", - "scene_name": "001875", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.13, \"box_2d\": [12.47, 190.45, 259.48, 296.18], \"box_3d\": [1.49, 1.71, 3.94, -7.99, 1.83, 12.6], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [497.72, 177.27, 525.28, 197.6], \"box_3d\": [1.56, 1.73, 4.21, -7.94, 1.94, 58.3], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 607, - "data_source": "Mono3DRefer", - "scene_name": "001885", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.31, \"box_2d\": [221.36, 181.01, 312.24, 214.34], \"box_3d\": [1.6, 1.66, 4.49, -17.72, 2.04, 37.43], \"ry\": 1.87}, {\"category\": \"Car\", \"angle\": 2.53, \"box_2d\": [390.32, 173.24, 455.28, 201.61], \"box_3d\": [1.78, 1.67, 3.85, -12.27, 1.82, 47.35], \"ry\": 2.28}, {\"category\": \"Car\", \"angle\": 2.78, \"box_2d\": [338.79, 177.48, 404.59, 200.59], \"box_3d\": [1.59, 1.76, 4.18, -17.32, 1.95, 52.44], \"ry\": 2.46}, {\"category\": \"Car\", \"angle\": 3.06, \"box_2d\": [501.44, 169.9, 567.89, 192.15], \"box_3d\": [1.66, 1.57, 5.04, -5.93, 1.47, 56.47], \"ry\": 2.95}]\n```" - } - ] - }, - { - "id": 608, - "data_source": "Mono3DRefer", - "scene_name": "001887", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [482.73, 184.48, 537.83, 226.4], \"box_3d\": [1.39, 1.48, 3.76, -3.62, 1.83, 26.41], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 609, - "data_source": "Mono3DRefer", - "scene_name": "001889", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.15, \"box_2d\": [724.26, 175.05, 771.63, 198.45], \"box_3d\": [1.48, 1.36, 3.51, 9.07, 1.65, 47.77], \"ry\": -1.96}]\n```" - } - ] - }, - { - "id": 610, - "data_source": "Mono3DRefer", - "scene_name": "001892", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.52, \"box_2d\": [413.46, 177.05, 470.81, 197.31], \"box_3d\": [1.6, 1.66, 4.49, -13.88, 1.97, 59.8], \"ry\": 2.29}]\n```" - } - ] - }, - { - "id": 611, - "data_source": "Mono3DRefer", - "scene_name": "001897", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.9, \"box_2d\": [649.18, 177.54, 795.77, 276.22], \"box_3d\": [1.51, 1.53, 3.38, 1.82, 1.61, 12.97], \"ry\": -1.76}, {\"category\": \"Car\", \"angle\": -1.8, \"box_2d\": [550.4, 180.68, 610.5, 219.96], \"box_3d\": [1.39, 1.54, 3.42, -1.23, 1.7, 27.62], \"ry\": -1.84}, {\"category\": \"Car\", \"angle\": 2.5, \"box_2d\": [1143.65, 183.39, 1241.0, 374.0], \"box_3d\": [1.53, 1.54, 3.54, 6.35, 1.6, 5.38], \"ry\": -2.95}, {\"category\": \"Car\", \"angle\": 2.5, \"box_2d\": [496.23, 174.0, 561.69, 201.72], \"box_3d\": [1.56, 1.39, 3.73, -4.77, 1.64, 42.55], \"ry\": 2.39}, {\"category\": \"Car\", \"angle\": -0.82, \"box_2d\": [604.87, 172.05, 674.42, 201.95], \"box_3d\": [1.56, 1.67, 3.78, 1.63, 1.53, 39.52], \"ry\": -0.78}, {\"category\": \"Car\", \"angle\": -0.95, \"box_2d\": [658.77, 171.7, 720.53, 198.48], \"box_3d\": [1.5, 1.53, 4.04, 4.71, 1.45, 42.47], \"ry\": -0.84}, {\"category\": \"Car\", \"angle\": -0.94, \"box_2d\": [715.32, 172.65, 762.0, 195.12], \"box_3d\": [1.38, 1.28, 3.16, 8.16, 1.37, 45.8], \"ry\": -0.76}]\n```" - } - ] - }, - { - "id": 612, - "data_source": "Mono3DRefer", - "scene_name": "001899", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [549.66, 173.77, 573.91, 195.41], \"box_3d\": [1.7, 1.72, 4.11, -3.95, 1.8, 59.33], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 613, - "data_source": "Mono3DRefer", - "scene_name": "001907", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [574.49, 183.36, 657.44, 262.64], \"box_3d\": [1.55, 1.63, 3.32, 0.08, 1.79, 16.01], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.21, \"box_2d\": [0.0, 209.5, 155.71, 268.4], \"box_3d\": [1.37, 1.82, 3.96, -15.39, 2.52, 20.92], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.96, \"box_2d\": [278.1, 195.64, 363.74, 233.79], \"box_3d\": [1.4, 1.75, 4.75, -12.37, 2.45, 31.14], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 2.52, \"box_2d\": [1146.21, 183.13, 1241.0, 255.04], \"box_3d\": [1.59, 1.73, 4.26, 15.39, 1.84, 16.92], \"ry\": -3.03}, {\"category\": \"Car\", \"angle\": 2.02, \"box_2d\": [207.66, 200.97, 283.06, 233.69], \"box_3d\": [1.46, 1.81, 4.16, -18.74, 2.98, 37.23], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [370.1, 190.99, 411.91, 216.46], \"box_3d\": [1.62, 1.85, 3.57, -15.03, 2.9, 49.62], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 614, - "data_source": "Mono3DRefer", - "scene_name": "001917", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [641.56, 175.81, 665.76, 198.76], \"box_3d\": [1.54, 1.63, 3.67, 3.06, 1.77, 50.91], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": -1.82, \"box_2d\": [752.2, 161.92, 901.28, 286.87], \"box_3d\": [1.88, 1.57, 3.71, 3.62, 1.75, 12.82], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -1.86, \"box_2d\": [832.21, 188.3, 1152.3, 374.0], \"box_3d\": [1.5, 1.56, 3.69, 3.29, 1.68, 7.09], \"ry\": -1.44}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [708.16, 174.98, 777.78, 234.78], \"box_3d\": [1.7, 1.62, 3.76, 4.03, 1.79, 22.59], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [694.74, 180.7, 732.85, 212.36], \"box_3d\": [1.4, 1.53, 3.56, 4.83, 1.78, 34.11], \"ry\": -1.49}]\n```" - } - ] - }, - { - "id": 615, - "data_source": "Mono3DRefer", - "scene_name": "001932", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [466.43, 183.42, 530.19, 230.93], \"box_3d\": [1.39, 1.48, 3.76, -3.58, 1.75, 23.44], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.62, \"box_2d\": [548.06, 173.38, 577.82, 196.9], \"box_3d\": [1.46, 1.75, 4.23, -3.08, 1.53, 47.52], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 616, - "data_source": "Mono3DRefer", - "scene_name": "001933", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [585.69, 170.52, 607.19, 187.55], \"box_3d\": [1.26, 1.6, 3.56, -1.04, 1.11, 55.74], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 617, - "data_source": "Mono3DRefer", - "scene_name": "001934", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [428.46, 182.28, 477.41, 214.12], \"box_3d\": [1.51, 1.51, 4.15, -7.97, 2.0, 36.8], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [493.52, 178.09, 521.61, 196.88], \"box_3d\": [1.49, 1.71, 3.94, -8.45, 1.93, 59.7], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 618, - "data_source": "Mono3DRefer", - "scene_name": "001935", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [565.01, 171.09, 617.27, 221.53], \"box_3d\": [1.6, 1.66, 3.55, -0.73, 1.57, 24.9], \"ry\": -1.62}]\n```" - } - ] - }, - { - "id": 619, - "data_source": "Mono3DRefer", - "scene_name": "001943", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [463.13, 177.96, 502.2, 204.74], \"box_3d\": [1.48, 1.64, 3.59, -7.42, 1.79, 42.16], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 620, - "data_source": "Mono3DRefer", - "scene_name": "001946", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [746.63, 128.04, 783.73, 155.31], \"box_3d\": [1.5, 1.62, 4.19, 9.93, -1.2, 46.05], \"ry\": 1.95}]\n```" - } - ] - }, - { - "id": 621, - "data_source": "Mono3DRefer", - "scene_name": "001948", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.48, \"box_2d\": [0.0, 166.55, 221.04, 310.34], \"box_3d\": [1.9, 1.78, 4.25, -8.33, 1.85, 10.52], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": 0.25, \"box_2d\": [354.64, 177.1, 457.56, 216.95], \"box_3d\": [1.52, 1.64, 3.7, -8.21, 1.72, 29.15], \"ry\": -0.02}]\n```" - } - ] - }, - { - "id": 622, - "data_source": "Mono3DRefer", - "scene_name": "001953", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [1003.41, 183.87, 1115.25, 234.77], \"box_3d\": [1.42, 1.68, 4.29, 14.22, 1.79, 22.8], \"ry\": 2.46}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [1100.63, 183.02, 1176.29, 221.8], \"box_3d\": [1.36, 1.4, 3.8, 20.33, 1.77, 27.74], \"ry\": 2.46}, {\"category\": \"Car\", \"angle\": 1.98, \"box_2d\": [915.19, 181.89, 1046.53, 245.09], \"box_3d\": [1.48, 1.51, 4.35, 9.95, 1.74, 19.29], \"ry\": 2.45}]\n```" - } - ] - }, - { - "id": 623, - "data_source": "Mono3DRefer", - "scene_name": "001957", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.79, \"box_2d\": [214.51, 190.04, 399.14, 257.11], \"box_3d\": [1.41, 1.54, 3.6, -6.92, 1.83, 16.61], \"ry\": 3.1}]\n```" - } - ] - }, - { - "id": 624, - "data_source": "Mono3DRefer", - "scene_name": "001960", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [362.97, 180.11, 483.54, 262.14], \"box_3d\": [1.45, 1.64, 4.21, -3.76, 1.61, 15.05], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 1.47, \"box_2d\": [643.48, 172.14, 700.04, 221.6], \"box_3d\": [1.62, 1.58, 3.89, 2.09, 1.62, 25.76], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [492.92, 175.87, 545.87, 211.07], \"box_3d\": [1.45, 1.67, 4.43, -4.01, 1.61, 32.29], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [628.36, 177.46, 675.45, 212.25], \"box_3d\": [1.47, 1.59, 3.62, 1.81, 1.69, 32.63], \"ry\": -1.66}]\n```" - } - ] - }, - { - "id": 625, - "data_source": "Mono3DRefer", - "scene_name": "001961", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.26, \"box_2d\": [737.57, 173.98, 803.64, 200.84], \"box_3d\": [1.62, 1.72, 4.35, 10.17, 1.72, 46.1], \"ry\": -2.04}]\n```" - } - ] - }, - { - "id": 626, - "data_source": "Mono3DRefer", - "scene_name": "001966", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.34, \"box_2d\": [366.5, 179.65, 496.91, 259.94], \"box_3d\": [1.46, 1.73, 4.19, -3.66, 1.62, 15.39], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [702.28, 175.33, 773.89, 221.51], \"box_3d\": [1.42, 1.66, 4.18, 4.23, 1.53, 24.61], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [547.13, 186.07, 758.41, 374.0], \"box_3d\": [1.49, 1.6, 3.83, 0.3, 1.65, 7.44], \"ry\": -1.56}]\n```" - } - ] - }, - { - "id": 627, - "data_source": "Mono3DRefer", - "scene_name": "001974", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.24, \"box_2d\": [0.0, 200.26, 160.08, 315.75], \"box_3d\": [1.48, 1.6, 3.9, -9.2, 1.98, 11.44], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [356.61, 183.64, 428.84, 224.71], \"box_3d\": [1.61, 1.67, 4.56, -9.32, 2.1, 31.21], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 628, - "data_source": "Mono3DRefer", - "scene_name": "001975", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [633.55, 176.18, 660.99, 193.25], \"box_3d\": [1.44, 1.68, 4.39, 3.3, 1.75, 63.49], \"ry\": 1.79}]\n```" - } - ] - }, - { - "id": 629, - "data_source": "Mono3DRefer", - "scene_name": "001977", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [102.68, 214.47, 217.03, 293.16], \"box_3d\": [1.6, 1.65, 3.77, -11.39, 2.73, 18.09], \"ry\": -2.35}, {\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [32.18, 224.69, 102.97, 271.44], \"box_3d\": [1.37, 1.61, 4.26, -20.34, 3.42, 26.93], \"ry\": -2.33}, {\"category\": \"Car\", \"angle\": -2.18, \"box_2d\": [367.68, 211.68, 687.5, 374.0], \"box_3d\": [1.47, 1.58, 3.55, -1.1, 1.95, 7.69], \"ry\": -2.32}, {\"category\": \"Car\", \"angle\": -1.96, \"box_2d\": [239.87, 215.47, 463.1, 339.91], \"box_3d\": [1.37, 1.7, 4.38, -4.25, 2.14, 11.27], \"ry\": -2.32}, {\"category\": \"Car\", \"angle\": -1.85, \"box_2d\": [172.09, 218.75, 314.85, 309.1], \"box_3d\": [1.4, 1.56, 3.91, -7.61, 2.44, 14.71], \"ry\": -2.32}]\n```" - } - ] - }, - { - "id": 630, - "data_source": "Mono3DRefer", - "scene_name": "001983", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [647.8, 178.52, 706.49, 221.98], \"box_3d\": [1.41, 1.58, 4.36, 2.28, 1.63, 25.85], \"ry\": -1.6}]\n```" - } - ] - }, - { - "id": 631, - "data_source": "Mono3DRefer", - "scene_name": "002009", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.62, \"box_2d\": [867.24, 168.82, 1131.48, 247.6], \"box_3d\": [1.47, 1.77, 4.49, 7.82, 1.42, 14.83], \"ry\": 3.1}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [565.3, 184.71, 592.09, 208.34], \"box_3d\": [1.52, 1.62, 4.5, -2.22, 2.36, 49.83], \"ry\": -1.66}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [473.44, 187.03, 506.34, 210.57], \"box_3d\": [1.42, 1.66, 4.5, -7.89, 2.38, 47.57], \"ry\": 1.51}]\n```" - } - ] - }, - { - "id": 632, - "data_source": "Mono3DRefer", - "scene_name": "002016", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.13, \"box_2d\": [31.06, 192.25, 253.86, 288.31], \"box_3d\": [1.61, 1.67, 4.56, -9.46, 2.06, 15.06], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 633, - "data_source": "Mono3DRefer", - "scene_name": "002024", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.06, \"box_2d\": [751.67, 184.29, 1241.0, 374.0], \"box_3d\": [1.46, 1.58, 3.93, 2.5, 1.57, 5.91], \"ry\": -1.68}, {\"category\": \"Car\", \"angle\": 2.1, \"box_2d\": [0.0, 197.52, 257.14, 373.42], \"box_3d\": [1.5, 1.68, 4.03, -5.95, 1.85, 8.56], \"ry\": 1.5}, {\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [644.27, 172.08, 789.56, 277.84], \"box_3d\": [1.6, 1.7, 3.56, 1.69, 1.61, 12.85], \"ry\": -1.68}, {\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [601.25, 176.12, 702.66, 245.88], \"box_3d\": [1.52, 1.73, 3.89, 0.88, 1.63, 17.99], \"ry\": -1.72}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [290.17, 190.06, 385.31, 254.59], \"box_3d\": [1.44, 1.6, 3.79, -7.03, 1.93, 18.86], \"ry\": 1.41}, {\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [339.98, 187.05, 406.22, 233.97], \"box_3d\": [1.42, 1.51, 3.71, -8.0, 1.93, 24.52], \"ry\": 1.43}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [582.02, 182.76, 645.14, 229.57], \"box_3d\": [1.37, 1.51, 3.49, 0.01, 1.7, 23.23], \"ry\": -1.72}]\n```" - } - ] - }, - { - "id": 634, - "data_source": "Mono3DRefer", - "scene_name": "002029", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [686.57, 187.14, 768.04, 251.55], \"box_3d\": [1.51, 1.56, 3.81, 3.0, 1.92, 19.38], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [669.73, 183.87, 736.04, 235.03], \"box_3d\": [1.57, 1.68, 3.83, 3.05, 1.97, 24.63], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.6, \"box_2d\": [530.63, 180.97, 562.27, 212.61], \"box_3d\": [1.52, 1.5, 3.85, -3.24, 1.94, 36.93], \"ry\": 1.51}]\n```" - } - ] - }, - { - "id": 635, - "data_source": "Mono3DRefer", - "scene_name": "002031", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.97, \"box_2d\": [399.8, 184.42, 491.5, 215.45], \"box_3d\": [1.47, 1.74, 4.26, -8.26, 2.08, 36.4], \"ry\": 3.09}]\n```" - } - ] - }, - { - "id": 636, - "data_source": "Mono3DRefer", - "scene_name": "002034", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [383.55, 192.77, 517.49, 287.08], \"box_3d\": [1.46, 1.75, 4.23, -2.95, 1.89, 14.01], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 637, - "data_source": "Mono3DRefer", - "scene_name": "002036", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [784.58, 124.54, 822.25, 152.15], \"box_3d\": [1.5, 1.62, 4.19, 12.34, -1.4, 45.95], \"ry\": 2.0}]\n```" - } - ] - }, - { - "id": 638, - "data_source": "Mono3DRefer", - "scene_name": "002037", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.1, \"box_2d\": [130.14, 190.79, 233.14, 238.34], \"box_3d\": [1.63, 1.87, 3.47, -16.28, 2.35, 27.56], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [375.53, 186.9, 413.46, 205.73], \"box_3d\": [1.33, 1.72, 3.63, -16.29, 2.42, 54.64], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 639, - "data_source": "Mono3DRefer", - "scene_name": "002043", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.78, \"box_2d\": [292.69, 176.29, 411.41, 227.62], \"box_3d\": [1.61, 1.57, 3.35, -8.52, 1.75, 23.89], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": 1.37, \"box_2d\": [692.11, 167.81, 742.84, 204.79], \"box_3d\": [1.79, 1.72, 4.35, 5.47, 1.57, 37.41], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 1.59, \"box_2d\": [552.45, 178.81, 582.17, 207.62], \"box_3d\": [1.45, 1.49, 4.07, -2.28, 1.78, 38.7], \"ry\": 1.53}]\n```" - } - ] - }, - { - "id": 640, - "data_source": "Mono3DRefer", - "scene_name": "002047", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [268.12, 180.81, 313.32, 201.69], \"box_3d\": [1.32, 1.51, 3.55, -21.14, 1.85, 47.84], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": -2.03, \"box_2d\": [811.76, 186.75, 1230.61, 374.0], \"box_3d\": [1.38, 1.51, 4.1, 3.26, 1.54, 6.83], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 1.98, \"box_2d\": [236.18, 180.58, 287.64, 205.21], \"box_3d\": [1.41, 1.48, 3.63, -21.03, 1.89, 43.66], \"ry\": 1.53}]\n```" - } - ] - }, - { - "id": 641, - "data_source": "Mono3DRefer", - "scene_name": "002048", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [888.35, 184.43, 982.23, 250.23], \"box_3d\": [1.52, 1.54, 4.5, 8.57, 1.86, 19.44], \"ry\": -1.32}, {\"category\": \"Car\", \"angle\": -2.58, \"box_2d\": [335.96, 177.03, 541.82, 279.17], \"box_3d\": [1.59, 1.45, 3.15, -2.97, 1.68, 12.42], \"ry\": -2.81}]\n```" - } - ] - }, - { - "id": 642, - "data_source": "Mono3DRefer", - "scene_name": "002053", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.53, \"box_2d\": [529.73, 180.95, 557.03, 207.67], \"box_3d\": [1.68, 1.63, 3.9, -4.48, 2.23, 47.94], \"ry\": 1.44}]\n```" - } - ] - }, - { - "id": 643, - "data_source": "Mono3DRefer", - "scene_name": "002054", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.35, \"box_2d\": [738.71, 184.5, 918.99, 312.07], \"box_3d\": [1.49, 1.61, 3.74, 2.89, 1.67, 10.46], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [586.9, 175.22, 624.25, 205.01], \"box_3d\": [1.46, 1.63, 4.47, -0.21, 1.61, 38.08], \"ry\": 1.64}, {\"category\": \"Car\", \"angle\": -2.95, \"box_2d\": [400.06, 175.37, 531.04, 219.96], \"box_3d\": [1.47, 1.68, 4.26, -5.04, 1.59, 25.35], \"ry\": -3.14}]\n```" - } - ] - }, - { - "id": 644, - "data_source": "Mono3DRefer", - "scene_name": "002055", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [548.56, 175.36, 573.17, 197.58], \"box_3d\": [1.7, 1.72, 4.11, -3.92, 1.92, 57.85], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 645, - "data_source": "Mono3DRefer", - "scene_name": "002057", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.23, \"box_2d\": [716.0, 143.83, 898.53, 263.83], \"box_3d\": [1.79, 1.72, 4.35, 3.26, 1.38, 13.05], \"ry\": 1.47}, {\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [302.84, 189.59, 439.11, 282.76], \"box_3d\": [1.45, 1.49, 4.07, -4.42, 1.8, 13.8], \"ry\": 1.5}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [424.45, 184.97, 489.79, 231.15], \"box_3d\": [1.27, 1.54, 3.58, -4.67, 1.66, 22.23], \"ry\": 1.48}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [455.06, 173.09, 507.96, 217.74], \"box_3d\": [1.63, 1.65, 3.91, -5.07, 1.67, 28.66], \"ry\": 1.5}, {\"category\": \"Car\", \"angle\": -1.61, \"box_2d\": [580.67, 171.07, 607.98, 195.98], \"box_3d\": [1.46, 1.53, 3.17, -1.02, 1.38, 44.33], \"ry\": -1.63}]\n```" - } - ] - }, - { - "id": 646, - "data_source": "Mono3DRefer", - "scene_name": "002061", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [380.08, 169.25, 474.55, 252.42], \"box_3d\": [1.68, 1.63, 3.9, -4.15, 1.64, 16.75], \"ry\": 1.45}, {\"category\": \"Car\", \"angle\": 1.51, \"box_2d\": [431.92, 175.13, 479.1, 212.32], \"box_3d\": [1.47, 1.7, 4.39, -6.75, 1.59, 31.06], \"ry\": 1.3}]\n```" - } - ] - }, - { - "id": 647, - "data_source": "Mono3DRefer", - "scene_name": "002062", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [242.26, 180.42, 457.99, 311.86], \"box_3d\": [1.48, 1.64, 3.95, -3.42, 1.59, 10.17], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [717.95, 189.19, 906.73, 309.4], \"box_3d\": [1.35, 1.64, 3.84, 2.63, 1.62, 10.42], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.44, \"box_2d\": [663.8, 177.9, 740.18, 233.05], \"box_3d\": [1.42, 1.59, 4.19, 2.51, 1.58, 20.87], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [452.65, 175.69, 523.93, 228.73], \"box_3d\": [1.42, 1.64, 3.63, -3.55, 1.52, 21.36], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [654.28, 176.0, 708.02, 216.37], \"box_3d\": [1.38, 1.58, 3.66, 2.53, 1.51, 26.73], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 648, - "data_source": "Mono3DRefer", - "scene_name": "002066", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [554.2, 180.44, 580.57, 198.58], \"box_3d\": [1.44, 1.72, 4.14, -3.54, 2.09, 60.28], \"ry\": 1.63}]\n```" - } - ] - }, - { - "id": 649, - "data_source": "Mono3DRefer", - "scene_name": "002069", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [525.08, 172.78, 553.96, 198.3], \"box_3d\": [1.65, 1.56, 3.98, -4.77, 1.67, 49.01], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 650, - "data_source": "Mono3DRefer", - "scene_name": "002080", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.75, \"box_2d\": [300.23, 180.39, 491.9, 242.95], \"box_3d\": [1.47, 1.67, 4.28, -5.32, 1.68, 18.06], \"ry\": -3.04}, {\"category\": \"Car\", \"angle\": -3.02, \"box_2d\": [380.92, 179.54, 525.13, 225.54], \"box_3d\": [1.38, 1.67, 4.37, -5.01, 1.62, 23.32], \"ry\": 3.05}, {\"category\": \"Car\", \"angle\": -0.37, \"box_2d\": [807.2, 167.6, 1017.73, 258.35], \"box_3d\": [1.54, 1.58, 3.2, 5.41, 1.47, 13.29], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": -0.45, \"box_2d\": [827.8, 167.72, 1153.76, 278.04], \"box_3d\": [1.51, 1.72, 4.03, 5.56, 1.47, 10.99], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": 0.67, \"box_2d\": [0.0, 200.02, 369.8, 359.74], \"box_3d\": [1.38, 1.66, 4.38, -5.01, 1.71, 7.49], \"ry\": 0.09}, {\"category\": \"Car\", \"angle\": -0.91, \"box_2d\": [1091.07, 176.48, 1241.0, 374.0], \"box_3d\": [1.48, 1.86, 4.18, 4.66, 1.47, 2.93], \"ry\": 0.06}, {\"category\": \"Car\", \"angle\": -0.63, \"box_2d\": [874.81, 144.89, 1241.0, 374.0], \"box_3d\": [1.74, 1.66, 4.37, 4.76, 1.54, 6.27], \"ry\": -0.0}]\n```" - } - ] - }, - { - "id": 651, - "data_source": "Mono3DRefer", - "scene_name": "002093", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.16, \"box_2d\": [0.0, 191.61, 283.73, 374.0], \"box_3d\": [1.46, 1.63, 4.47, -3.19, 1.6, 3.43], \"ry\": 1.45}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [338.96, 165.87, 494.27, 267.45], \"box_3d\": [1.7, 1.82, 4.45, -3.67, 1.61, 14.39], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -1.35, \"box_2d\": [650.43, 170.42, 693.6, 203.93], \"box_3d\": [1.57, 1.56, 2.62, 3.04, 1.48, 35.34], \"ry\": -1.27}]\n```" - } - ] - }, - { - "id": 652, - "data_source": "Mono3DRefer", - "scene_name": "002121", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.96, \"box_2d\": [282.72, 188.64, 351.24, 220.12], \"box_3d\": [1.56, 1.91, 4.44, -16.0, 2.46, 39.57], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 653, - "data_source": "Mono3DRefer", - "scene_name": "002122", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.6, \"box_2d\": [154.17, 189.61, 384.47, 281.6], \"box_3d\": [1.41, 1.62, 3.6, -6.25, 1.76, 13.51], \"ry\": -1.03}, {\"category\": \"Car\", \"angle\": -0.54, \"box_2d\": [32.46, 174.2, 305.11, 276.94], \"box_3d\": [1.77, 1.96, 4.39, -8.86, 1.82, 14.85], \"ry\": -1.07}]\n```" - } - ] - }, - { - "id": 654, - "data_source": "Mono3DRefer", - "scene_name": "002127", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [98.75, 169.48, 409.42, 348.97], \"box_3d\": [1.5, 1.62, 4.19, -3.65, 1.5, 8.22], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [642.67, 189.17, 669.13, 210.63], \"box_3d\": [1.41, 1.58, 4.36, 3.24, 2.62, 51.88], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 655, - "data_source": "Mono3DRefer", - "scene_name": "002131", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.56, \"box_2d\": [0.0, 179.8, 142.72, 331.83], \"box_3d\": [1.77, 1.73, 4.53, -8.8, 1.89, 9.05], \"ry\": 2.96}, {\"category\": \"Car\", \"angle\": -1.37, \"box_2d\": [367.71, 185.74, 477.48, 272.48], \"box_3d\": [1.57, 1.55, 3.32, -3.83, 1.86, 15.08], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [439.55, 183.89, 519.71, 242.16], \"box_3d\": [1.56, 1.68, 3.91, -3.85, 1.91, 21.68], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [493.93, 180.3, 536.03, 209.29], \"box_3d\": [1.43, 1.75, 4.47, -5.01, 1.83, 38.28], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 656, - "data_source": "Mono3DRefer", - "scene_name": "002137", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.37, \"box_2d\": [880.29, 170.54, 1241.0, 374.0], \"box_3d\": [1.72, 1.67, 4.42, 2.98, 1.71, 3.15], \"ry\": -1.66}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [167.25, 200.54, 438.48, 374.0], \"box_3d\": [1.44, 1.56, 3.84, -3.02, 1.8, 7.85], \"ry\": 1.5}, {\"category\": \"Car\", \"angle\": -1.52, \"box_2d\": [467.12, 186.17, 519.23, 235.58], \"box_3d\": [1.5, 1.55, 4.21, -3.97, 1.98, 24.79], \"ry\": -1.68}, {\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [617.98, 174.75, 678.92, 231.86], \"box_3d\": [1.82, 1.65, 3.67, 1.23, 1.91, 25.06], \"ry\": -1.64}]\n```" - } - ] - }, - { - "id": 657, - "data_source": "Mono3DRefer", - "scene_name": "002141", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [802.48, 175.9, 864.2, 228.25], \"box_3d\": [1.56, 1.65, 3.69, 7.14, 1.67, 23.46], \"ry\": -1.34}, {\"category\": \"Car\", \"angle\": 2.46, \"box_2d\": [0.0, 195.51, 257.23, 374.0], \"box_3d\": [1.42, 1.67, 4.53, -5.54, 1.68, 6.55], \"ry\": 1.78}]\n```" - } - ] - }, - { - "id": 658, - "data_source": "Mono3DRefer", - "scene_name": "002148", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [450.44, 176.41, 515.85, 215.34], \"box_3d\": [1.42, 1.78, 3.62, -4.6, 1.24, 28.34], \"ry\": 1.63}]\n```" - } - ] - }, - { - "id": 659, - "data_source": "Mono3DRefer", - "scene_name": "002163", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.98, \"box_2d\": [249.18, 189.97, 324.18, 227.23], \"box_3d\": [1.45, 1.64, 3.61, -13.92, 2.23, 31.2], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [402.93, 182.01, 442.66, 204.2], \"box_3d\": [1.57, 1.9, 4.16, -14.04, 2.28, 54.23], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 660, - "data_source": "Mono3DRefer", - "scene_name": "002164", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.92, \"box_2d\": [663.46, 179.39, 746.37, 212.09], \"box_3d\": [1.48, 1.8, 4.31, 4.7, 1.83, 35.53], \"ry\": -0.79}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [655.0, 180.21, 688.72, 206.84], \"box_3d\": [1.69, 1.88, 4.5, 4.19, 2.2, 48.52], \"ry\": 1.74}]\n```" - } - ] - }, - { - "id": 661, - "data_source": "Mono3DRefer", - "scene_name": "002165", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.06, \"box_2d\": [187.06, 189.42, 253.17, 223.74], \"box_3d\": [1.54, 1.85, 2.47, -18.77, 2.37, 34.78], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 662, - "data_source": "Mono3DRefer", - "scene_name": "002168", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.75, \"box_2d\": [155.68, 184.56, 362.79, 251.57], \"box_3d\": [1.43, 1.6, 4.04, -8.11, 1.73, 16.84], \"ry\": 3.09}, {\"category\": \"Car\", \"angle\": 0.37, \"box_2d\": [214.96, 177.57, 394.01, 238.05], \"box_3d\": [1.53, 1.72, 4.15, -8.27, 1.69, 19.69], \"ry\": -0.02}, {\"category\": \"Car\", \"angle\": -2.93, \"box_2d\": [296.23, 176.86, 427.73, 224.43], \"box_3d\": [1.52, 1.57, 4.01, -8.45, 1.69, 24.72], \"ry\": 3.03}, {\"category\": \"Car\", \"angle\": -0.27, \"box_2d\": [650.61, 180.34, 753.69, 223.14], \"box_3d\": [1.48, 1.63, 3.43, 3.32, 1.77, 26.55], \"ry\": -0.15}, {\"category\": \"Car\", \"angle\": -1.89, \"box_2d\": [460.01, 175.88, 493.38, 195.74], \"box_3d\": [1.66, 1.66, 3.98, -11.63, 1.95, 62.68], \"ry\": -2.08}, {\"category\": \"Car\", \"angle\": -0.25, \"box_2d\": [664.39, 179.68, 779.91, 229.25], \"box_3d\": [1.55, 1.57, 3.51, 3.67, 1.8, 24.12], \"ry\": -0.1}, {\"category\": \"Car\", \"angle\": -0.41, \"box_2d\": [729.15, 181.04, 971.85, 269.47], \"box_3d\": [1.47, 1.74, 3.88, 4.31, 1.64, 13.38], \"ry\": -0.1}, {\"category\": \"Car\", \"angle\": 0.38, \"box_2d\": [77.95, 166.0, 318.79, 259.38], \"box_3d\": [1.68, 1.67, 3.79, -8.02, 1.58, 14.28], \"ry\": -0.13}]\n```" - } - ] - }, - { - "id": 663, - "data_source": "Mono3DRefer", - "scene_name": "002179", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [654.86, 183.49, 734.93, 240.49], \"box_3d\": [1.41, 1.58, 4.36, 2.25, 1.73, 20.47], \"ry\": -1.61}]\n```" - } - ] - }, - { - "id": 664, - "data_source": "Mono3DRefer", - "scene_name": "002192", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [464.74, 181.03, 521.27, 226.06], \"box_3d\": [1.7, 1.72, 4.11, -4.76, 2.05, 29.64], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 665, - "data_source": "Mono3DRefer", - "scene_name": "002193", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.99, \"box_2d\": [17.88, 213.87, 437.19, 374.0], \"box_3d\": [1.42, 1.57, 4.1, -2.91, 1.9, 6.5], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -2.25, \"box_2d\": [856.97, 213.09, 1241.0, 374.0], \"box_3d\": [1.43, 1.64, 4.14, 2.5, 1.69, 2.88], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [511.74, 186.31, 559.18, 225.95], \"box_3d\": [1.38, 1.53, 3.58, -2.84, 1.92, 27.69], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [649.77, 182.34, 693.71, 217.45], \"box_3d\": [1.49, 1.65, 4.06, 2.75, 1.94, 33.26], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [552.36, 183.11, 578.79, 207.86], \"box_3d\": [1.48, 1.54, 4.05, -2.83, 2.16, 46.18], \"ry\": -1.6}]\n```" - } - ] - }, - { - "id": 666, - "data_source": "Mono3DRefer", - "scene_name": "002196", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [368.72, 184.86, 406.03, 214.14], \"box_3d\": [1.45, 1.62, 3.83, -11.89, 2.11, 38.61], \"ry\": 1.35}]\n```" - } - ] - }, - { - "id": 667, - "data_source": "Mono3DRefer", - "scene_name": "002197", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [761.64, 170.95, 812.18, 214.78], \"box_3d\": [1.56, 1.65, 3.69, 6.68, 1.52, 27.66], \"ry\": -1.39}]\n```" - } - ] - }, - { - "id": 668, - "data_source": "Mono3DRefer", - "scene_name": "002200", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.73, \"box_2d\": [1011.89, 168.02, 1225.84, 236.69], \"box_3d\": [1.5, 1.7, 3.72, 11.75, 1.43, 16.95], \"ry\": -2.96}, {\"category\": \"Car\", \"angle\": 2.47, \"box_2d\": [542.29, 171.25, 591.66, 192.18], \"box_3d\": [1.56, 1.39, 3.73, -3.31, 1.45, 55.53], \"ry\": 2.41}, {\"category\": \"Car\", \"angle\": 2.33, \"box_2d\": [0.0, 222.33, 248.54, 374.0], \"box_3d\": [1.39, 1.6, 3.8, -3.69, 1.77, 3.73], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -2.17, \"box_2d\": [896.0, 186.77, 1241.0, 374.0], \"box_3d\": [1.48, 1.56, 3.46, 3.06, 1.58, 4.13], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.86, \"box_2d\": [653.18, 172.58, 723.26, 218.9], \"box_3d\": [1.51, 1.53, 3.38, 2.66, 1.53, 25.56], \"ry\": -1.76}, {\"category\": \"Car\", \"angle\": -1.94, \"box_2d\": [582.0, 174.14, 630.24, 200.39], \"box_3d\": [1.39, 1.54, 3.42, -0.29, 1.48, 40.29], \"ry\": -1.95}, {\"category\": \"Car\", \"angle\": 3.02, \"box_2d\": [816.64, 170.82, 978.82, 236.56], \"box_3d\": [1.53, 1.54, 3.54, 6.97, 1.51, 17.9], \"ry\": -2.9}]\n```" - } - ] - }, - { - "id": 669, - "data_source": "Mono3DRefer", - "scene_name": "002207", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.18, \"box_2d\": [1133.98, 169.65, 1241.0, 273.25], \"box_3d\": [1.65, 1.67, 3.64, 11.31, 1.6, 12.92], \"ry\": -0.47}, {\"category\": \"Car\", \"angle\": 2.21, \"box_2d\": [982.67, 170.48, 1120.21, 226.64], \"box_3d\": [1.59, 1.55, 3.93, 13.52, 1.54, 22.15], \"ry\": 2.75}]\n```" - } - ] - }, - { - "id": 670, - "data_source": "Mono3DRefer", - "scene_name": "002208", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.94, \"box_2d\": [415.51, 178.0, 492.83, 207.86], \"box_3d\": [1.49, 1.64, 3.7, -8.16, 1.78, 37.83], \"ry\": 3.13}]\n```" - } - ] - }, - { - "id": 671, - "data_source": "Mono3DRefer", - "scene_name": "002215", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.29, \"box_2d\": [1132.73, 174.81, 1241.0, 271.24], \"box_3d\": [1.65, 1.67, 3.64, 11.86, 1.69, 14.0], \"ry\": -0.59}, {\"category\": \"Car\", \"angle\": 2.22, \"box_2d\": [814.03, 176.34, 952.05, 239.25], \"box_3d\": [1.59, 1.55, 3.93, 7.66, 1.71, 20.24], \"ry\": 2.58}, {\"category\": \"Car\", \"angle\": 2.02, \"box_2d\": [1026.82, 180.55, 1135.07, 225.39], \"box_3d\": [1.47, 1.57, 4.24, 16.99, 1.77, 26.03], \"ry\": 2.6}]\n```" - } - ] - }, - { - "id": 672, - "data_source": "Mono3DRefer", - "scene_name": "002223", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [655.2, 172.36, 725.15, 237.79], \"box_3d\": [1.56, 1.65, 3.69, 2.08, 1.57, 19.14], \"ry\": -1.46}, {\"category\": \"Car\", \"angle\": -1.43, \"box_2d\": [1060.02, 174.41, 1241.0, 374.0], \"box_3d\": [1.39, 1.61, 3.65, 4.68, 1.4, 5.1], \"ry\": -0.72}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [467.49, 170.1, 547.23, 226.73], \"box_3d\": [1.76, 1.58, 3.68, -3.41, 1.7, 24.35], \"ry\": 1.75}, {\"category\": \"Car\", \"angle\": 2.26, \"box_2d\": [81.98, 185.55, 293.87, 271.89], \"box_3d\": [1.56, 1.62, 4.19, -8.88, 1.86, 15.58], \"ry\": 1.74}, {\"category\": \"Car\", \"angle\": 2.07, \"box_2d\": [399.8, 182.95, 495.72, 223.42], \"box_3d\": [1.42, 1.67, 4.53, -6.26, 1.84, 28.25], \"ry\": 1.86}, {\"category\": \"Car\", \"angle\": -0.83, \"box_2d\": [741.61, 167.68, 805.61, 194.46], \"box_3d\": [1.54, 1.63, 3.74, 9.79, 1.25, 43.2], \"ry\": -0.61}]\n```" - } - ] - }, - { - "id": 673, - "data_source": "Mono3DRefer", - "scene_name": "002227", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.65, \"box_2d\": [89.88, 188.9, 312.13, 267.65], \"box_3d\": [1.49, 1.64, 3.7, -8.45, 1.86, 15.05], \"ry\": 3.13}, {\"category\": \"Car\", \"angle\": -2.72, \"box_2d\": [176.26, 182.24, 369.57, 249.19], \"box_3d\": [1.52, 1.55, 4.02, -8.22, 1.78, 17.71], \"ry\": 3.13}, {\"category\": \"Car\", \"angle\": -2.85, \"box_2d\": [347.79, 173.0, 465.25, 222.27], \"box_3d\": [1.9, 1.78, 4.25, -8.24, 1.94, 29.33], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": 0.15, \"box_2d\": [462.91, 177.96, 522.62, 201.9], \"box_3d\": [1.52, 1.64, 3.7, -7.78, 1.89, 47.93], \"ry\": -0.01}]\n```" - } - ] - }, - { - "id": 674, - "data_source": "Mono3DRefer", - "scene_name": "002234", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.04, \"box_2d\": [80.93, 196.93, 423.54, 374.0], \"box_3d\": [1.44, 1.64, 3.78, -3.47, 1.75, 7.76], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [318.62, 173.38, 503.45, 302.09], \"box_3d\": [1.77, 1.71, 4.28, -3.12, 1.81, 12.13], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [714.72, 188.47, 804.21, 246.62], \"box_3d\": [1.45, 1.71, 3.93, 4.13, 1.93, 20.72], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [476.23, 181.54, 557.48, 246.86], \"box_3d\": [1.73, 1.65, 3.96, -2.68, 2.0, 21.3], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [692.25, 177.53, 763.74, 230.3], \"box_3d\": [1.78, 1.89, 4.37, 4.23, 1.97, 26.75], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [514.31, 182.33, 567.48, 232.38], \"box_3d\": [1.68, 1.51, 3.29, -2.48, 2.03, 26.15], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 675, - "data_source": "Mono3DRefer", - "scene_name": "002237", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.97, \"box_2d\": [0.0, 164.01, 419.95, 374.0], \"box_3d\": [1.83, 1.69, 4.28, -3.01, 1.81, 6.16], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [498.71, 187.41, 588.36, 244.75], \"box_3d\": [1.51, 1.63, 4.03, -1.93, 1.98, 21.73], \"ry\": 1.76}, {\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [561.64, 185.15, 627.18, 229.72], \"box_3d\": [1.57, 1.44, 3.6, -0.57, 2.07, 27.94], \"ry\": 1.88}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [618.3, 185.09, 671.17, 219.41], \"box_3d\": [1.47, 1.51, 3.5, 1.63, 2.06, 33.42], \"ry\": 1.9}, {\"category\": \"Car\", \"angle\": -1.27, \"box_2d\": [798.44, 167.78, 838.14, 190.57], \"box_3d\": [1.77, 1.75, 4.67, 16.9, 1.39, 58.45], \"ry\": -0.99}]\n```" - } - ] - }, - { - "id": 676, - "data_source": "Mono3DRefer", - "scene_name": "002266", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [589.66, 176.96, 616.81, 203.8], \"box_3d\": [1.65, 1.67, 3.64, -0.48, 1.93, 46.64], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.34, \"box_2d\": [354.99, 178.26, 511.4, 296.39], \"box_3d\": [1.61, 1.51, 4.5, -2.79, 1.72, 12.2], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.11, \"box_2d\": [67.12, 194.83, 215.51, 264.32], \"box_3d\": [1.61, 1.72, 3.65, -12.53, 2.25, 19.52], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 677, - "data_source": "Mono3DRefer", - "scene_name": "002269", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [506.99, 182.68, 550.57, 217.49], \"box_3d\": [1.45, 1.53, 3.98, -3.66, 1.91, 32.73], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [189.17, 184.26, 432.95, 342.19], \"box_3d\": [1.65, 1.56, 3.98, -3.72, 1.82, 9.67], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 678, - "data_source": "Mono3DRefer", - "scene_name": "002273", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.35, \"box_2d\": [334.67, 184.58, 458.58, 220.43], \"box_3d\": [1.36, 1.68, 4.49, -8.61, 1.86, 29.17], \"ry\": 0.07}]\n```" - } - ] - }, - { - "id": 679, - "data_source": "Mono3DRefer", - "scene_name": "002276", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.87, \"box_2d\": [134.1, 176.71, 349.75, 254.23], \"box_3d\": [1.77, 1.73, 4.53, -9.13, 1.9, 18.17], \"ry\": 2.96}, {\"category\": \"Car\", \"angle\": -1.44, \"box_2d\": [445.88, 184.47, 505.26, 235.51], \"box_3d\": [1.57, 1.55, 3.32, -4.51, 1.98, 24.36], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [471.74, 182.68, 524.25, 222.12], \"box_3d\": [1.56, 1.68, 3.91, -4.77, 1.99, 30.98], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 680, - "data_source": "Mono3DRefer", - "scene_name": "002277", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [183.89, 187.29, 458.65, 369.0], \"box_3d\": [1.46, 1.63, 4.47, -2.95, 1.66, 8.33], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [442.44, 168.28, 544.04, 239.29], \"box_3d\": [1.7, 1.82, 4.45, -3.06, 1.62, 19.62], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": -1.23, \"box_2d\": [645.37, 168.34, 691.54, 200.22], \"box_3d\": [1.57, 1.56, 2.62, 3.0, 1.37, 37.09], \"ry\": -1.15}]\n```" - } - ] - }, - { - "id": 681, - "data_source": "Mono3DRefer", - "scene_name": "002278", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [636.28, 179.53, 661.36, 197.37], \"box_3d\": [1.48, 1.36, 3.51, 3.31, 2.07, 62.2], \"ry\": -1.75}]\n```" - } - ] - }, - { - "id": 682, - "data_source": "Mono3DRefer", - "scene_name": "002281", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.99, \"box_2d\": [195.98, 172.77, 285.37, 217.8], \"box_3d\": [1.66, 1.54, 4.34, -14.75, 1.69, 29.01], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": -1.61, \"box_2d\": [585.92, 179.82, 642.42, 225.42], \"box_3d\": [1.36, 1.69, 3.38, 0.03, 1.6, 23.59], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -2.1, \"box_2d\": [224.6, 182.74, 421.77, 283.35], \"box_3d\": [1.5, 1.58, 3.57, -5.11, 1.69, 12.56], \"ry\": -2.48}, {\"category\": \"Car\", \"angle\": -2.14, \"box_2d\": [316.55, 164.63, 483.12, 259.41], \"box_3d\": [1.8, 1.63, 3.79, -4.64, 1.67, 15.52], \"ry\": -2.43}, {\"category\": \"Car\", \"angle\": -2.2, \"box_2d\": [398.62, 181.37, 523.21, 231.52], \"box_3d\": [1.41, 1.62, 4.19, -4.76, 1.7, 22.49], \"ry\": -2.41}, {\"category\": \"Car\", \"angle\": -2.28, \"box_2d\": [416.75, 178.07, 526.9, 223.07], \"box_3d\": [1.49, 1.66, 3.97, -5.02, 1.7, 25.8], \"ry\": -2.47}, {\"category\": \"Car\", \"angle\": -2.23, \"box_2d\": [460.56, 178.91, 538.39, 210.73], \"box_3d\": [1.39, 1.62, 3.72, -5.22, 1.69, 33.58], \"ry\": -2.39}, {\"category\": \"Car\", \"angle\": -2.35, \"box_2d\": [485.43, 175.3, 553.38, 203.82], \"box_3d\": [1.49, 1.66, 3.55, -5.01, 1.65, 39.47], \"ry\": -2.48}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [655.97, 178.0, 687.76, 203.0], \"box_3d\": [1.4, 1.66, 4.08, 3.62, 1.72, 43.0], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.61, \"box_2d\": [588.42, 174.96, 624.37, 206.78], \"box_3d\": [1.52, 1.67, 3.61, -0.26, 1.65, 36.6], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": 2.13, \"box_2d\": [0.0, 186.75, 201.22, 287.94], \"box_3d\": [1.39, 1.64, 3.55, -8.58, 1.64, 11.97], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 2.15, \"box_2d\": [55.4, 179.27, 212.1, 233.37], \"box_3d\": [1.46, 1.72, 4.6, -14.35, 1.67, 22.04], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 683, - "data_source": "Mono3DRefer", - "scene_name": "002283", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [594.78, 175.36, 625.78, 205.99], \"box_3d\": [1.65, 1.67, 3.64, 0.02, 1.81, 41.07], \"ry\": -1.55}]\n```" - } - ] - }, - { - "id": 684, - "data_source": "Mono3DRefer", - "scene_name": "002284", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [395.74, 184.06, 436.09, 206.9], \"box_3d\": [1.48, 1.71, 3.54, -13.33, 2.27, 49.63], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 685, - "data_source": "Mono3DRefer", - "scene_name": "002285", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [601.63, 179.27, 662.66, 228.87], \"box_3d\": [1.36, 1.69, 3.38, 0.58, 1.57, 21.8], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.61, \"box_2d\": [945.8, 175.5, 1057.13, 208.18], \"box_3d\": [1.45, 1.67, 4.32, 18.26, 1.6, 33.81], \"ry\": 3.1}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [408.55, 180.29, 450.33, 205.94], \"box_3d\": [1.59, 1.72, 3.86, -11.76, 2.08, 47.11], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [584.53, 176.23, 625.26, 213.42], \"box_3d\": [1.52, 1.67, 3.61, -0.28, 1.68, 31.57], \"ry\": -1.59}]\n```" - } - ] - }, - { - "id": 686, - "data_source": "Mono3DRefer", - "scene_name": "002288", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.34, \"box_2d\": [257.66, 181.94, 402.01, 236.27], \"box_3d\": [1.52, 1.64, 3.7, -8.37, 1.82, 21.66], \"ry\": -0.03}, {\"category\": \"Car\", \"angle\": -2.74, \"box_2d\": [321.48, 182.87, 436.08, 225.48], \"box_3d\": [1.5, 1.62, 3.69, -8.55, 1.89, 26.72], \"ry\": -3.05}, {\"category\": \"Car\", \"angle\": 1.46, \"box_2d\": [877.75, 163.38, 910.72, 185.49], \"box_3d\": [1.51, 1.7, 4.45, 20.43, 0.88, 52.03], \"ry\": 1.83}]\n```" - } - ] - }, - { - "id": 687, - "data_source": "Mono3DRefer", - "scene_name": "002294", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [707.47, 185.64, 904.87, 352.89], \"box_3d\": [1.46, 1.53, 3.17, 1.92, 1.62, 8.04], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": -2.64, \"box_2d\": [271.89, 188.95, 427.28, 260.75], \"box_3d\": [1.59, 1.45, 3.15, -6.18, 1.99, 17.12], \"ry\": -2.98}]\n```" - } - ] - }, - { - "id": 688, - "data_source": "Mono3DRefer", - "scene_name": "002304", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [589.17, 168.99, 620.81, 200.2], \"box_3d\": [1.65, 1.67, 3.64, -0.27, 1.47, 40.34], \"ry\": -1.55}]\n```" - } - ] - }, - { - "id": 689, - "data_source": "Mono3DRefer", - "scene_name": "002321", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.52, \"box_2d\": [142.03, 193.13, 355.09, 253.13], \"box_3d\": [1.36, 1.68, 4.49, -8.93, 1.9, 17.93], \"ry\": 0.06}]\n```" - } - ] - }, - { - "id": 690, - "data_source": "Mono3DRefer", - "scene_name": "002337", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.2, \"box_2d\": [13.05, 198.84, 175.58, 258.02], \"box_3d\": [1.46, 1.86, 4.01, -14.99, 2.28, 21.22], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 691, - "data_source": "Mono3DRefer", - "scene_name": "002339", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [479.6, 177.54, 520.17, 209.5], \"box_3d\": [1.5, 1.46, 3.7, -5.47, 1.75, 35.98], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [345.27, 183.95, 451.24, 253.09], \"box_3d\": [1.54, 1.54, 3.72, -5.27, 1.84, 18.26], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [687.9, 176.71, 785.05, 253.82], \"box_3d\": [1.57, 1.54, 3.22, 2.74, 1.67, 16.45], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [660.14, 169.3, 733.04, 233.81], \"box_3d\": [1.74, 1.6, 3.86, 2.45, 1.67, 21.54], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [649.47, 168.02, 706.13, 221.63], \"box_3d\": [1.83, 1.69, 4.44, 2.42, 1.69, 27.03], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [535.61, 169.81, 558.95, 193.69], \"box_3d\": [1.86, 1.57, 3.83, -5.08, 1.65, 58.6], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 692, - "data_source": "Mono3DRefer", - "scene_name": "002343", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [527.73, 171.1, 554.77, 195.17], \"box_3d\": [1.65, 1.56, 3.98, -4.92, 1.55, 51.84], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 693, - "data_source": "Mono3DRefer", - "scene_name": "002350", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [492.09, 184.73, 564.66, 241.72], \"box_3d\": [1.6, 1.68, 3.94, -2.51, 1.99, 22.67], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [574.93, 185.35, 601.67, 206.72], \"box_3d\": [1.42, 1.66, 4.11, -1.53, 2.34, 51.58], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": -0.19, \"box_2d\": [1054.67, 123.02, 1241.0, 261.46], \"box_3d\": [1.66, 1.15, 3.39, 7.68, 1.07, 9.2], \"ry\": 0.49}, {\"category\": \"Car\", \"angle\": -1.28, \"box_2d\": [311.65, 183.53, 510.78, 320.59], \"box_3d\": [1.77, 1.75, 4.67, -2.99, 1.96, 11.81], \"ry\": -1.52}]\n```" - } - ] - }, - { - "id": 694, - "data_source": "Mono3DRefer", - "scene_name": "002373", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [572.87, 181.15, 602.03, 207.64], \"box_3d\": [1.43, 1.55, 3.91, -1.28, 1.91, 41.41], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.29, \"box_2d\": [764.29, 186.73, 945.57, 301.74], \"box_3d\": [1.52, 1.61, 4.23, 3.76, 1.78, 11.98], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.31, \"box_2d\": [763.27, 184.92, 839.62, 233.48], \"box_3d\": [1.45, 1.51, 3.7, 6.21, 1.87, 23.92], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 2.96, \"box_2d\": [649.14, 178.45, 737.48, 209.83], \"box_3d\": [1.51, 1.65, 4.21, 4.16, 1.82, 36.66], \"ry\": 3.07}]\n```" - } - ] - }, - { - "id": 695, - "data_source": "Mono3DRefer", - "scene_name": "002374", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.54, \"box_2d\": [534.9, 186.36, 567.67, 220.11], \"box_3d\": [1.68, 1.63, 3.9, -3.24, 2.44, 38.99], \"ry\": 1.46}, {\"category\": \"Car\", \"angle\": 1.45, \"box_2d\": [504.53, 187.58, 530.61, 205.8], \"box_3d\": [1.47, 1.7, 4.39, -8.06, 2.77, 62.49], \"ry\": 1.32}]\n```" - } - ] - }, - { - "id": 696, - "data_source": "Mono3DRefer", - "scene_name": "002378", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.4, \"box_2d\": [949.76, 216.7, 1241.0, 374.0], \"box_3d\": [1.47, 1.59, 3.62, 2.81, 1.72, 2.45], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [163.93, 174.08, 442.86, 343.45], \"box_3d\": [1.56, 1.68, 3.69, -3.33, 1.6, 8.54], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.25, \"box_2d\": [746.81, 196.73, 1060.64, 374.0], \"box_3d\": [1.44, 1.64, 4.5, 2.75, 1.77, 8.05], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [429.73, 170.1, 516.15, 238.41], \"box_3d\": [1.46, 1.51, 3.24, -3.21, 1.42, 17.2], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.44, \"box_2d\": [678.54, 176.83, 750.15, 227.25], \"box_3d\": [1.38, 1.59, 4.19, 3.04, 1.52, 22.02], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [487.54, 173.02, 549.18, 217.68], \"box_3d\": [1.38, 1.69, 3.74, -3.06, 1.41, 24.41], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [651.79, 171.93, 682.57, 199.2], \"box_3d\": [1.51, 1.56, 3.81, 3.28, 1.48, 42.23], \"ry\": -1.55}]\n```" - } - ] - }, - { - "id": 697, - "data_source": "Mono3DRefer", - "scene_name": "002380", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.66, \"box_2d\": [1089.78, 182.76, 1241.0, 374.0], \"box_3d\": [1.63, 1.53, 3.45, 6.43, 1.71, 6.31], \"ry\": 0.11}, {\"category\": \"Car\", \"angle\": 2.7, \"box_2d\": [813.71, 166.58, 1078.66, 264.51], \"box_3d\": [1.76, 1.71, 4.4, 6.43, 1.67, 14.19], \"ry\": 3.12}, {\"category\": \"Car\", \"angle\": 0.56, \"box_2d\": [0.0, 130.71, 311.81, 299.42], \"box_3d\": [2.22, 1.68, 4.08, -6.77, 1.69, 10.48], \"ry\": -0.0}, {\"category\": \"Car\", \"angle\": -2.67, \"box_2d\": [111.08, 183.42, 335.34, 268.53], \"box_3d\": [1.4, 1.46, 3.23, -6.87, 1.61, 12.94], \"ry\": 3.13}, {\"category\": \"Car\", \"angle\": -2.83, \"box_2d\": [283.89, 182.46, 458.16, 239.12], \"box_3d\": [1.46, 1.61, 4.31, -6.59, 1.75, 20.06], \"ry\": 3.14}, {\"category\": \"Car\", \"angle\": -3.02, \"box_2d\": [362.72, 170.56, 498.93, 225.98], \"box_3d\": [1.8, 1.77, 4.41, -6.17, 1.75, 25.11], \"ry\": 3.02}, {\"category\": \"Car\", \"angle\": -2.94, \"box_2d\": [447.88, 177.06, 512.43, 206.44], \"box_3d\": [1.58, 1.54, 3.3, -7.26, 1.83, 40.34], \"ry\": -3.12}]\n```" - } - ] - }, - { - "id": 698, - "data_source": "Mono3DRefer", - "scene_name": "002400", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.19, \"box_2d\": [429.22, 179.66, 503.36, 209.07], \"box_3d\": [1.52, 1.64, 3.7, -7.79, 1.91, 39.17], \"ry\": -0.01}, {\"category\": \"Car\", \"angle\": -2.43, \"box_2d\": [0.0, 196.69, 160.51, 334.04], \"box_3d\": [1.52, 1.55, 4.02, -8.15, 1.86, 8.99], \"ry\": 3.13}, {\"category\": \"Car\", \"angle\": -2.74, \"box_2d\": [233.99, 173.71, 408.13, 244.73], \"box_3d\": [1.9, 1.78, 4.25, -8.19, 1.95, 20.6], \"ry\": -3.12}]\n```" - } - ] - }, - { - "id": 699, - "data_source": "Mono3DRefer", - "scene_name": "002416", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.74, \"box_2d\": [478.94, 171.24, 665.38, 253.54], \"box_3d\": [1.85, 1.68, 4.34, -0.96, 1.83, 18.07], \"ry\": 2.68}, {\"category\": \"Car\", \"angle\": -2.78, \"box_2d\": [0.0, 207.32, 56.07, 337.21], \"box_3d\": [1.51, 1.53, 4.46, -9.5, 2.0, 8.47], \"ry\": 2.68}, {\"category\": \"Car\", \"angle\": -2.89, \"box_2d\": [0.0, 197.63, 127.45, 271.59], \"box_3d\": [1.43, 1.56, 3.85, -12.73, 2.01, 15.32], \"ry\": 2.71}, {\"category\": \"Car\", \"angle\": -2.93, \"box_2d\": [9.0, 194.87, 164.2, 248.66], \"box_3d\": [1.42, 1.62, 3.5, -15.62, 2.13, 21.68], \"ry\": 2.74}, {\"category\": \"Car\", \"angle\": -3.02, \"box_2d\": [116.2, 190.21, 213.18, 223.95], \"box_3d\": [1.41, 1.54, 3.6, -20.28, 2.23, 32.96], \"ry\": 2.71}, {\"category\": \"Car\", \"angle\": -1.21, \"box_2d\": [1057.63, 198.69, 1241.0, 374.0], \"box_3d\": [1.39, 1.59, 3.17, 4.73, 1.59, 4.81], \"ry\": -0.46}, {\"category\": \"Car\", \"angle\": -0.88, \"box_2d\": [812.22, 188.06, 1241.0, 374.0], \"box_3d\": [1.46, 1.55, 4.28, 4.23, 1.64, 7.14], \"ry\": -0.36}, {\"category\": \"Car\", \"angle\": -0.51, \"box_2d\": [563.99, 185.76, 798.6, 277.65], \"box_3d\": [1.49, 1.66, 4.16, 1.33, 1.77, 13.76], \"ry\": -0.42}]\n```" - } - ] - }, - { - "id": 700, - "data_source": "Mono3DRefer", - "scene_name": "002419", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.22, \"box_2d\": [606.09, 205.57, 1050.45, 374.0], \"box_3d\": [1.47, 1.58, 3.55, 1.38, 1.81, 6.02], \"ry\": -2.0}, {\"category\": \"Car\", \"angle\": -1.93, \"box_2d\": [483.73, 205.82, 709.94, 343.72], \"box_3d\": [1.37, 1.7, 4.38, -0.5, 1.93, 10.39], \"ry\": -1.98}, {\"category\": \"Car\", \"angle\": -1.83, \"box_2d\": [419.61, 207.08, 551.12, 299.25], \"box_3d\": [1.4, 1.56, 3.91, -2.64, 2.16, 14.29], \"ry\": -2.02}, {\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [369.6, 201.81, 459.54, 274.11], \"box_3d\": [1.6, 1.65, 3.77, -5.33, 2.43, 19.2], \"ry\": -2.02}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [126.44, 207.56, 205.7, 274.91], \"box_3d\": [1.86, 1.63, 4.2, -14.66, 3.1, 23.96], \"ry\": 1.17}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [28.31, 221.86, 147.85, 321.42], \"box_3d\": [1.45, 1.46, 3.41, -9.92, 2.49, 13.85], \"ry\": 1.06}, {\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [324.66, 210.13, 374.42, 250.03], \"box_3d\": [1.37, 1.61, 4.26, -11.06, 3.03, 30.38], \"ry\": -2.0}, {\"category\": \"Car\", \"angle\": 1.6, \"box_2d\": [195.81, 213.05, 227.98, 242.88], \"box_3d\": [1.51, 1.55, 3.92, -23.82, 4.01, 43.19], \"ry\": 1.1}]\n```" - } - ] - }, - { - "id": 701, - "data_source": "Mono3DRefer", - "scene_name": "002423", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.88, \"box_2d\": [249.97, 180.59, 425.05, 238.96], \"box_3d\": [1.47, 1.53, 4.21, -7.38, 1.71, 19.72], \"ry\": 3.05}, {\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [649.24, 170.26, 756.05, 251.45], \"box_3d\": [1.64, 1.62, 4.2, 1.98, 1.62, 16.9], \"ry\": -1.65}, {\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [376.39, 187.28, 500.8, 276.45], \"box_3d\": [1.35, 1.47, 3.43, -3.0, 1.64, 13.07], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 702, - "data_source": "Mono3DRefer", - "scene_name": "002428", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [443.45, 186.31, 534.42, 250.49], \"box_3d\": [1.46, 1.75, 4.23, -3.12, 1.85, 19.15], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 703, - "data_source": "Mono3DRefer", - "scene_name": "002431", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.59, \"box_2d\": [510.05, 181.22, 561.21, 226.86], \"box_3d\": [1.41, 1.59, 4.47, -2.57, 1.71, 24.93], \"ry\": 1.48}]\n```" - } - ] - }, - { - "id": 704, - "data_source": "Mono3DRefer", - "scene_name": "002432", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.98, \"box_2d\": [278.1, 186.3, 350.03, 222.26], \"box_3d\": [1.48, 1.71, 3.54, -13.23, 2.11, 32.38], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 705, - "data_source": "Mono3DRefer", - "scene_name": "002443", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.88, \"box_2d\": [752.11, 190.12, 1046.88, 374.0], \"box_3d\": [1.41, 1.64, 3.77, 2.64, 1.63, 7.66], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [483.02, 183.4, 534.41, 217.69], \"box_3d\": [1.43, 1.76, 3.97, -4.58, 1.93, 32.87], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.97, \"box_2d\": [204.95, 196.82, 403.8, 309.55], \"box_3d\": [1.49, 1.61, 3.94, -4.93, 1.94, 12.15], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [339.74, 189.52, 462.96, 263.06], \"box_3d\": [1.48, 1.66, 3.9, -4.83, 1.91, 17.11], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 2.48, \"box_2d\": [0.0, 241.13, 96.61, 374.0], \"box_3d\": [1.31, 1.58, 4.23, -5.38, 1.89, 4.18], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 706, - "data_source": "Mono3DRefer", - "scene_name": "002445", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.68, \"box_2d\": [917.88, 181.82, 1229.36, 281.3], \"box_3d\": [1.35, 1.49, 4.0, 7.0, 1.51, 11.11], \"ry\": -0.13}, {\"category\": \"Car\", \"angle\": -0.64, \"box_2d\": [882.41, 180.78, 1170.94, 261.5], \"box_3d\": [1.32, 1.62, 4.57, 7.57, 1.49, 13.34], \"ry\": -0.14}]\n```" - } - ] - }, - { - "id": 707, - "data_source": "Mono3DRefer", - "scene_name": "002447", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.57, \"box_2d\": [592.3, 183.29, 613.76, 202.68], \"box_3d\": [1.41, 1.59, 4.47, -0.57, 2.24, 56.14], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 708, - "data_source": "Mono3DRefer", - "scene_name": "002451", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [794.92, 171.12, 870.61, 235.39], \"box_3d\": [1.56, 1.65, 3.69, 5.87, 1.54, 19.47], \"ry\": -1.35}, {\"category\": \"Car\", \"angle\": 2.15, \"box_2d\": [301.59, 184.71, 487.48, 262.98], \"box_3d\": [1.42, 1.67, 4.53, -4.54, 1.71, 15.88], \"ry\": 1.87}, {\"category\": \"Car\", \"angle\": -0.42, \"box_2d\": [1153.25, 160.44, 1241.0, 198.87], \"box_3d\": [1.54, 1.63, 3.74, 25.09, 1.05, 29.95], \"ry\": 0.27}]\n```" - } - ] - }, - { - "id": 709, - "data_source": "Mono3DRefer", - "scene_name": "002460", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [766.36, 192.81, 894.4, 279.8], \"box_3d\": [1.52, 1.54, 4.5, 4.53, 2.0, 15.61], \"ry\": -1.53}]\n```" - } - ] - }, - { - "id": 710, - "data_source": "Mono3DRefer", - "scene_name": "002489", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [692.67, 189.13, 834.5, 288.41], \"box_3d\": [1.41, 1.64, 3.77, 2.46, 1.73, 12.62], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [486.53, 188.88, 532.75, 219.14], \"box_3d\": [1.43, 1.76, 3.97, -5.21, 2.3, 37.71], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [322.62, 196.84, 445.95, 273.66], \"box_3d\": [1.49, 1.61, 3.94, -5.17, 2.1, 16.93], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [391.93, 193.03, 481.52, 249.34], \"box_3d\": [1.48, 1.66, 3.9, -5.21, 2.14, 22.0], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 2.13, \"box_2d\": [0.0, 215.5, 304.82, 374.0], \"box_3d\": [1.31, 1.58, 4.23, -5.52, 1.94, 8.87], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 711, - "data_source": "Mono3DRefer", - "scene_name": "002491", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [456.92, 185.19, 534.95, 239.96], \"box_3d\": [1.52, 1.73, 4.28, -3.53, 1.93, 22.79], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [538.36, 181.11, 569.89, 204.79], \"box_3d\": [1.44, 1.7, 3.65, -3.58, 1.99, 46.44], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 712, - "data_source": "Mono3DRefer", - "scene_name": "002494", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.52, \"box_2d\": [0.0, 167.96, 257.76, 295.9], \"box_3d\": [1.9, 1.78, 4.25, -8.36, 1.86, 11.77], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": 0.24, \"box_2d\": [366.94, 176.63, 465.05, 214.76], \"box_3d\": [1.52, 1.64, 3.7, -8.15, 1.71, 30.41], \"ry\": -0.02}]\n```" - } - ] - }, - { - "id": 713, - "data_source": "Mono3DRefer", - "scene_name": "002501", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [600.33, 177.68, 659.13, 225.47], \"box_3d\": [1.36, 1.69, 3.38, 0.53, 1.53, 22.56], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.62, \"box_2d\": [925.2, 172.39, 1029.1, 203.19], \"box_3d\": [1.45, 1.67, 4.32, 18.15, 1.45, 35.82], \"ry\": 3.09}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [414.61, 176.48, 455.5, 201.06], \"box_3d\": [1.59, 1.72, 3.86, -11.87, 1.85, 49.07], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [583.98, 172.97, 623.94, 209.44], \"box_3d\": [1.52, 1.67, 3.61, -0.31, 1.55, 32.15], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 714, - "data_source": "Mono3DRefer", - "scene_name": "002511", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [150.54, 200.33, 405.94, 359.46], \"box_3d\": [1.39, 1.48, 3.76, -3.73, 1.78, 8.74], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [506.35, 181.17, 556.37, 217.1], \"box_3d\": [1.46, 1.75, 4.23, -3.44, 1.84, 31.93], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 715, - "data_source": "Mono3DRefer", - "scene_name": "002512", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [483.47, 176.78, 559.18, 214.51], \"box_3d\": [1.37, 1.78, 4.65, -3.46, 1.55, 28.78], \"ry\": 1.73}]\n```" - } - ] - }, - { - "id": 716, - "data_source": "Mono3DRefer", - "scene_name": "002514", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.68, \"box_2d\": [33.19, 200.59, 308.52, 300.67], \"box_3d\": [1.42, 1.62, 3.5, -7.07, 1.92, 11.83], \"ry\": 3.07}, {\"category\": \"Car\", \"angle\": -2.89, \"box_2d\": [330.36, 186.95, 453.38, 232.82], \"box_3d\": [1.41, 1.54, 3.6, -7.16, 1.9, 23.79], \"ry\": 3.1}]\n```" - } - ] - }, - { - "id": 717, - "data_source": "Mono3DRefer", - "scene_name": "002521", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.09, \"box_2d\": [147.55, 187.15, 240.63, 226.59], \"box_3d\": [1.67, 1.82, 4.42, -19.46, 2.38, 33.94], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [321.35, 181.72, 373.23, 209.66], \"box_3d\": [1.63, 1.87, 3.47, -16.19, 2.19, 44.56], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 718, - "data_source": "Mono3DRefer", - "scene_name": "002542", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.48, \"box_2d\": [692.41, 170.14, 722.84, 194.95], \"box_3d\": [1.54, 1.63, 3.67, 6.39, 1.4, 47.17], \"ry\": -1.34}, {\"category\": \"Car\", \"angle\": 2.22, \"box_2d\": [0.0, 152.41, 285.24, 327.12], \"box_3d\": [1.93, 1.59, 4.01, -6.42, 1.74, 10.05], \"ry\": 1.66}, {\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [734.83, 172.65, 870.66, 267.42], \"box_3d\": [1.58, 1.67, 3.84, 3.54, 1.61, 14.09], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [603.83, 170.26, 649.62, 201.14], \"box_3d\": [1.53, 1.67, 3.78, 0.9, 1.42, 37.84], \"ry\": 1.79}, {\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [730.85, 151.85, 810.14, 240.13], \"box_3d\": [2.05, 1.5, 3.83, 4.04, 1.59, 18.8], \"ry\": -1.48}]\n```" - } - ] - }, - { - "id": 719, - "data_source": "Mono3DRefer", - "scene_name": "002560", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.44, \"box_2d\": [0.0, 206.47, 183.94, 366.57], \"box_3d\": [1.41, 1.54, 3.6, -6.67, 1.8, 7.38], \"ry\": 3.12}]\n```" - } - ] - }, - { - "id": 720, - "data_source": "Mono3DRefer", - "scene_name": "002565", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (707.05, 707.05), (cx, cy) = (604.08, 180.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.07, \"box_2d\": [0.0, 189.33, 383.1, 369.0], \"box_3d\": [1.5, 1.78, 3.69, -3.26, 1.61, 5.53], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [454.42, 168.74, 486.36, 192.16], \"box_3d\": [1.45, 1.75, 4.23, -8.78, 0.71, 46.45], \"ry\": 1.46}]\n```" - } - ] - }, - { - "id": 721, - "data_source": "Mono3DRefer", - "scene_name": "002578", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [783.32, 169.41, 859.64, 234.34], \"box_3d\": [1.56, 1.65, 3.69, 5.51, 1.5, 19.29], \"ry\": -1.36}, {\"category\": \"Car\", \"angle\": 2.13, \"box_2d\": [330.03, 183.31, 497.68, 254.07], \"box_3d\": [1.42, 1.67, 4.53, -4.51, 1.69, 17.22], \"ry\": 1.88}, {\"category\": \"Car\", \"angle\": -0.42, \"box_2d\": [1089.41, 159.13, 1205.81, 196.0], \"box_3d\": [1.54, 1.63, 3.74, 23.52, 0.99, 31.73], \"ry\": 0.21}]\n```" - } - ] - }, - { - "id": 722, - "data_source": "Mono3DRefer", - "scene_name": "002582", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.06, \"box_2d\": [0.0, 166.55, 366.42, 374.0], \"box_3d\": [1.65, 1.75, 3.39, -2.54, 1.66, 3.51], \"ry\": 1.47}, {\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [340.46, 180.95, 517.05, 309.12], \"box_3d\": [1.57, 1.54, 4.01, -2.55, 1.7, 10.93], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 723, - "data_source": "Mono3DRefer", - "scene_name": "002589", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.62, \"box_2d\": [9.54, 198.4, 289.73, 275.98], \"box_3d\": [1.36, 1.68, 4.49, -8.93, 1.9, 14.12], \"ry\": 0.06}]\n```" - } - ] - }, - { - "id": 724, - "data_source": "Mono3DRefer", - "scene_name": "002590", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [663.02, 165.64, 682.55, 183.83], \"box_3d\": [1.47, 1.55, 3.67, 5.31, 0.91, 60.93], \"ry\": -1.46}]\n```" - } - ] - }, - { - "id": 725, - "data_source": "Mono3DRefer", - "scene_name": "002591", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [396.65, 185.33, 505.3, 283.9], \"box_3d\": [1.65, 1.75, 3.39, -3.02, 1.91, 14.13], \"ry\": 1.45}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [459.58, 184.8, 532.48, 244.06], \"box_3d\": [1.57, 1.54, 4.01, -3.35, 1.95, 21.6], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 726, - "data_source": "Mono3DRefer", - "scene_name": "002593", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.93, \"box_2d\": [398.21, 177.85, 481.78, 209.94], \"box_3d\": [1.49, 1.64, 3.7, -8.29, 1.76, 35.27], \"ry\": 3.13}]\n```" - } - ] - }, - { - "id": 727, - "data_source": "Mono3DRefer", - "scene_name": "002603", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [343.13, 180.59, 390.05, 206.41], \"box_3d\": [1.63, 1.87, 3.47, -16.14, 2.16, 47.93], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 2.05, \"box_2d\": [187.54, 184.89, 268.8, 220.66], \"box_3d\": [1.67, 1.82, 4.42, -19.45, 2.31, 36.92], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 728, - "data_source": "Mono3DRefer", - "scene_name": "002624", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [486.97, 171.95, 548.27, 227.95], \"box_3d\": [1.97, 1.87, 4.53, -3.51, 1.97, 27.93], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [546.11, 176.05, 572.37, 200.8], \"box_3d\": [1.7, 1.53, 3.82, -3.65, 1.95, 52.04], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 729, - "data_source": "Mono3DRefer", - "scene_name": "002630", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.0, \"box_2d\": [337.08, 185.68, 387.47, 210.16], \"box_3d\": [1.45, 1.74, 3.42, -15.67, 2.29, 45.73], \"ry\": 1.67}, {\"category\": \"Car\", \"angle\": 2.22, \"box_2d\": [0.0, 195.63, 135.09, 248.8], \"box_3d\": [1.38, 1.8, 3.41, -16.28, 2.1, 21.58], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 730, - "data_source": "Mono3DRefer", - "scene_name": "002631", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.02, \"box_2d\": [838.4, 181.67, 940.9, 229.84], \"box_3d\": [1.47, 1.57, 4.24, 9.56, 1.79, 24.59], \"ry\": 2.38}, {\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [970.6, 182.83, 1045.52, 217.01], \"box_3d\": [1.4, 1.68, 4.45, 18.02, 1.88, 32.63], \"ry\": 2.38}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [1039.38, 182.2, 1096.98, 212.49], \"box_3d\": [1.45, 1.62, 3.77, 23.71, 1.96, 37.33], \"ry\": 2.38}, {\"category\": \"Car\", \"angle\": -1.48, \"box_2d\": [1222.82, 183.36, 1241.0, 222.28], \"box_3d\": [1.5, 1.61, 4.07, 26.61, 1.95, 29.77], \"ry\": -0.76}, {\"category\": \"Car\", \"angle\": -1.39, \"box_2d\": [1077.91, 178.2, 1204.52, 261.88], \"box_3d\": [1.65, 1.67, 3.64, 11.99, 1.79, 16.25], \"ry\": -0.76}, {\"category\": \"Car\", \"angle\": 2.33, \"box_2d\": [565.21, 180.46, 735.59, 261.65], \"box_3d\": [1.59, 1.55, 3.93, 0.96, 1.79, 16.34], \"ry\": 2.39}]\n```" - } - ] - }, - { - "id": 731, - "data_source": "Mono3DRefer", - "scene_name": "002641", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [329.48, 171.29, 496.03, 296.55], \"box_3d\": [1.7, 1.63, 4.08, -3.08, 1.71, 11.93], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -2.0, \"box_2d\": [807.42, 208.98, 1241.0, 374.0], \"box_3d\": [1.37, 1.59, 3.22, 2.41, 1.67, 4.58], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.4, \"box_2d\": [456.77, 184.11, 541.8, 241.73], \"box_3d\": [1.4, 1.67, 3.72, -2.98, 1.73, 19.86], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": 1.49, \"box_2d\": [633.29, 176.01, 671.42, 207.14], \"box_3d\": [1.49, 1.61, 3.74, 2.08, 1.67, 36.71], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 732, - "data_source": "Mono3DRefer", - "scene_name": "002647", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.94, \"box_2d\": [407.26, 190.18, 601.82, 292.78], \"box_3d\": [1.39, 1.59, 3.17, -1.65, 1.71, 11.93], \"ry\": -1.07}, {\"category\": \"Car\", \"angle\": -0.78, \"box_2d\": [296.6, 190.6, 527.39, 288.78], \"box_3d\": [1.46, 1.55, 4.28, -3.52, 1.83, 13.5], \"ry\": -1.04}, {\"category\": \"Car\", \"angle\": -0.56, \"box_2d\": [73.33, 195.1, 291.12, 272.49], \"box_3d\": [1.49, 1.66, 4.16, -9.93, 2.07, 17.03], \"ry\": -1.08}, {\"category\": \"Car\", \"angle\": 2.67, \"box_2d\": [0.0, 185.04, 165.0, 265.27], \"box_3d\": [1.85, 1.68, 4.34, -14.41, 2.2, 18.93], \"ry\": 2.03}]\n```" - } - ] - }, - { - "id": 733, - "data_source": "Mono3DRefer", - "scene_name": "002654", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [113.15, 190.34, 363.12, 314.76], \"box_3d\": [1.51, 1.81, 4.07, -5.5, 1.83, 11.25], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 734, - "data_source": "Mono3DRefer", - "scene_name": "002655", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [515.84, 170.26, 559.03, 212.13], \"box_3d\": [1.97, 1.87, 4.53, -3.64, 1.88, 36.59], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [552.96, 175.65, 575.09, 196.77], \"box_3d\": [1.7, 1.53, 3.82, -3.85, 1.96, 60.66], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 735, - "data_source": "Mono3DRefer", - "scene_name": "002656", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.24, \"box_2d\": [0.0, 187.79, 137.74, 273.61], \"box_3d\": [1.62, 1.86, 4.31, -12.77, 1.98, 15.87], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 2.03, \"box_2d\": [249.62, 182.09, 316.03, 212.32], \"box_3d\": [1.67, 1.82, 4.42, -19.37, 2.24, 42.87], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [378.38, 178.63, 417.66, 201.15], \"box_3d\": [1.63, 1.87, 3.47, -16.03, 2.08, 54.66], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 736, - "data_source": "Mono3DRefer", - "scene_name": "002657", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [0.0, 170.69, 379.3, 374.0], \"box_3d\": [1.5, 1.62, 4.19, -3.62, 1.52, 6.75], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [644.4, 189.71, 671.65, 211.9], \"box_3d\": [1.41, 1.58, 4.36, 3.29, 2.62, 50.35], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 737, - "data_source": "Mono3DRefer", - "scene_name": "002662", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.03, \"box_2d\": [132.11, 176.8, 240.88, 228.43], \"box_3d\": [1.66, 1.54, 4.34, -14.9, 1.82, 25.6], \"ry\": 1.5}, {\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [583.51, 181.73, 639.74, 227.68], \"box_3d\": [1.36, 1.69, 3.38, -0.03, 1.67, 23.5], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -2.0, \"box_2d\": [90.82, 186.16, 350.61, 332.3], \"box_3d\": [1.5, 1.58, 3.57, -5.1, 1.68, 9.21], \"ry\": -2.49}, {\"category\": \"Car\", \"angle\": -2.08, \"box_2d\": [238.08, 163.14, 446.21, 288.74], \"box_3d\": [1.8, 1.63, 3.79, -4.65, 1.69, 12.16], \"ry\": -2.44}, {\"category\": \"Car\", \"angle\": -2.16, \"box_2d\": [362.22, 185.95, 505.62, 246.35], \"box_3d\": [1.41, 1.62, 4.19, -4.79, 1.77, 19.11], \"ry\": -2.4}, {\"category\": \"Car\", \"angle\": -2.24, \"box_2d\": [385.87, 182.46, 510.69, 235.1], \"box_3d\": [1.49, 1.66, 3.97, -5.13, 1.81, 22.45], \"ry\": -2.47}, {\"category\": \"Car\", \"angle\": -2.22, \"box_2d\": [443.05, 183.75, 528.98, 219.53], \"box_3d\": [1.39, 1.62, 3.72, -5.27, 1.86, 30.23], \"ry\": -2.39}, {\"category\": \"Car\", \"angle\": -2.34, \"box_2d\": [474.13, 182.35, 548.14, 213.9], \"box_3d\": [1.49, 1.66, 3.55, -5.01, 1.99, 36.15], \"ry\": -2.48}, {\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [693.99, 185.16, 776.28, 240.64], \"box_3d\": [1.47, 1.73, 4.22, 3.64, 1.87, 21.89], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [658.13, 184.45, 693.82, 212.11], \"box_3d\": [1.4, 1.66, 4.08, 3.55, 2.06, 39.7], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [590.26, 179.45, 626.9, 211.37], \"box_3d\": [1.52, 1.67, 3.61, -0.15, 1.86, 36.51], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": 2.23, \"box_2d\": [0.0, 183.8, 144.76, 249.72], \"box_3d\": [1.46, 1.72, 4.6, -14.39, 1.77, 18.59], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 738, - "data_source": "Mono3DRefer", - "scene_name": "002684", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [393.44, 192.22, 515.28, 278.61], \"box_3d\": [1.42, 1.66, 4.11, -3.03, 1.86, 14.65], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.6, \"box_2d\": [559.04, 166.94, 583.27, 191.71], \"box_3d\": [1.97, 1.87, 4.53, -3.22, 1.53, 60.27], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 739, - "data_source": "Mono3DRefer", - "scene_name": "002686", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [876.76, 111.88, 917.21, 140.79], \"box_3d\": [1.5, 1.62, 4.19, 18.09, -2.14, 45.44], \"ry\": 2.13}]\n```" - } - ] - }, - { - "id": 740, - "data_source": "Mono3DRefer", - "scene_name": "002694", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 3.12, \"box_2d\": [380.17, 170.39, 468.41, 204.67], \"box_3d\": [1.49, 1.46, 3.84, -8.55, 1.4, 33.16], \"ry\": 2.87}, {\"category\": \"Car\", \"angle\": 1.21, \"box_2d\": [682.64, 159.13, 728.43, 192.93], \"box_3d\": [2.21, 1.76, 4.11, 6.48, 1.34, 49.52], \"ry\": 1.34}, {\"category\": \"Car\", \"angle\": -2.01, \"box_2d\": [711.53, 165.59, 757.22, 191.46], \"box_3d\": [1.75, 1.57, 4.07, 8.73, 1.28, 51.09], \"ry\": -1.84}, {\"category\": \"Car\", \"angle\": -2.05, \"box_2d\": [750.47, 166.08, 806.36, 189.05], \"box_3d\": [1.54, 1.71, 5.03, 11.87, 1.12, 51.28], \"ry\": -1.82}, {\"category\": \"Car\", \"angle\": -2.19, \"box_2d\": [922.97, 163.51, 970.47, 185.27], \"box_3d\": [1.53, 1.41, 3.4, 24.62, 0.89, 52.89], \"ry\": -1.76}, {\"category\": \"Car\", \"angle\": -2.35, \"box_2d\": [962.93, 160.64, 1016.78, 184.23], \"box_3d\": [1.69, 1.61, 3.36, 28.2, 0.83, 53.68], \"ry\": -1.87}]\n```" - } - ] - }, - { - "id": 741, - "data_source": "Mono3DRefer", - "scene_name": "002696", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.98, \"box_2d\": [82.36, 190.62, 401.91, 374.0], \"box_3d\": [1.63, 1.65, 3.91, -3.76, 1.87, 8.13], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [616.7, 180.68, 669.4, 228.9], \"box_3d\": [1.46, 1.53, 3.17, 0.99, 1.73, 23.69], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 742, - "data_source": "Mono3DRefer", - "scene_name": "002699", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.97, \"box_2d\": [77.15, 103.72, 483.73, 373.0], \"box_3d\": [2.11, 2.04, 4.62, -2.81, 1.54, 7.63], \"ry\": 1.62}]\n```" - } - ] - }, - { - "id": 743, - "data_source": "Mono3DRefer", - "scene_name": "002709", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.16, \"box_2d\": [0.0, 59.71, 413.85, 373.0], \"box_3d\": [2.11, 2.04, 4.62, -2.93, 1.72, 4.51], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 744, - "data_source": "Mono3DRefer", - "scene_name": "002712", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.83, \"box_2d\": [735.4, 178.83, 917.4, 292.0], \"box_3d\": [1.4, 1.61, 3.77, 3.0, 1.51, 10.95], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [698.0, 177.73, 780.48, 244.14], \"box_3d\": [1.41, 1.57, 3.16, 2.92, 1.54, 17.06], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [360.46, 192.18, 448.9, 250.86], \"box_3d\": [1.53, 1.56, 3.58, -6.05, 2.14, 21.48], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [490.26, 188.07, 523.21, 213.03], \"box_3d\": [1.42, 1.54, 3.5, -6.32, 2.38, 44.28], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [638.21, 179.02, 668.34, 203.09], \"box_3d\": [1.51, 1.75, 3.78, 2.8, 1.93, 47.77], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [503.61, 188.22, 533.73, 209.63], \"box_3d\": [1.36, 1.51, 4.02, -6.29, 2.45, 49.86], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -2.14, \"box_2d\": [827.6, 170.16, 1241.0, 374.0], \"box_3d\": [1.51, 1.85, 4.93, 2.89, 1.51, 4.16], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 745, - "data_source": "Mono3DRefer", - "scene_name": "002713", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [592.74, 173.21, 624.03, 204.13], \"box_3d\": [1.65, 1.67, 3.64, -0.08, 1.7, 40.7], \"ry\": -1.55}]\n```" - } - ] - }, - { - "id": 746, - "data_source": "Mono3DRefer", - "scene_name": "002715", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.96, \"box_2d\": [316.83, 181.03, 378.52, 211.91], \"box_3d\": [1.58, 1.82, 3.87, -14.25, 2.04, 39.33], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 2.23, \"box_2d\": [0.0, 189.3, 113.74, 266.56], \"box_3d\": [1.6, 1.78, 3.86, -13.92, 2.02, 16.98], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 747, - "data_source": "Mono3DRefer", - "scene_name": "002727", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [315.46, 187.8, 491.49, 316.26], \"box_3d\": [1.46, 1.6, 3.33, -2.75, 1.7, 10.16], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [714.11, 175.14, 902.41, 313.64], \"box_3d\": [1.64, 1.7, 3.54, 2.59, 1.7, 10.41], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [686.38, 182.92, 791.18, 258.62], \"box_3d\": [1.37, 1.51, 3.39, 2.51, 1.59, 15.0], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [526.56, 175.33, 568.41, 210.22], \"box_3d\": [1.55, 1.68, 4.42, -2.97, 1.69, 34.63], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 748, - "data_source": "Mono3DRefer", - "scene_name": "002730", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.46, \"box_2d\": [0.0, 197.92, 216.48, 366.49], \"box_3d\": [1.41, 1.63, 3.75, -6.23, 1.7, 7.06], \"ry\": 3.12}, {\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [663.13, 180.63, 708.76, 214.58], \"box_3d\": [1.41, 1.65, 4.28, 3.33, 1.78, 32.61], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 0.2, \"box_2d\": [397.25, 178.46, 526.28, 219.63], \"box_3d\": [1.44, 1.8, 4.44, -5.49, 1.68, 26.95], \"ry\": 0.0}, {\"category\": \"Car\", \"angle\": 0.31, \"box_2d\": [332.99, 180.82, 493.82, 232.21], \"box_3d\": [1.5, 1.8, 4.5, 0.0, 1.77, 22.53], \"ry\": 0.05}, {\"category\": \"Car\", \"angle\": 0.37, \"box_2d\": [310.14, 179.92, 487.51, 240.31], \"box_3d\": [1.53, 1.8, 4.22, 0.43, 1.74, 19.48], \"ry\": 0.09}, {\"category\": \"Car\", \"angle\": 0.41, \"box_2d\": [282.07, 182.4, 468.85, 249.72], \"box_3d\": [1.5, 1.46, 3.95, -5.54, 1.75, 17.1], \"ry\": 0.1}, {\"category\": \"Car\", \"angle\": 0.38, \"box_2d\": [216.05, 180.46, 426.17, 262.42], \"box_3d\": [1.53, 1.64, 3.6, -5.8, 1.71, 14.66], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": -2.54, \"box_2d\": [0.0, 184.39, 270.58, 313.75], \"box_3d\": [1.53, 1.51, 3.38, -6.55, 1.7, 9.45], \"ry\": -3.14}]\n```" - } - ] - }, - { - "id": 749, - "data_source": "Mono3DRefer", - "scene_name": "002732", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.08, \"box_2d\": [79.32, 199.43, 182.17, 240.6], \"box_3d\": [1.42, 1.7, 4.14, -19.31, 2.56, 29.23], \"ry\": -1.66}, {\"category\": \"Car\", \"angle\": -0.98, \"box_2d\": [2.68, 198.18, 119.27, 242.24], \"box_3d\": [1.52, 1.71, 4.11, -21.94, 2.6, 29.0], \"ry\": -1.62}]\n```" - } - ] - }, - { - "id": 750, - "data_source": "Mono3DRefer", - "scene_name": "002741", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.19, \"box_2d\": [112.12, 187.04, 463.44, 374.0], \"box_3d\": [1.77, 1.75, 4.67, -3.11, 1.96, 8.11], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [456.44, 187.69, 546.71, 257.75], \"box_3d\": [1.6, 1.68, 3.94, -2.77, 2.02, 18.98], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [564.04, 186.37, 592.97, 209.52], \"box_3d\": [1.42, 1.66, 4.11, -2.08, 2.34, 47.93], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 751, - "data_source": "Mono3DRefer", - "scene_name": "002751", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.53, \"box_2d\": [565.25, 193.54, 600.37, 226.43], \"box_3d\": [1.43, 1.55, 3.91, -1.31, 2.48, 35.22], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.31, \"box_2d\": [993.86, 175.85, 1241.0, 374.0], \"box_3d\": [1.51, 1.65, 4.21, 5.1, 1.52, 4.9], \"ry\": 3.09}, {\"category\": \"Car\", \"angle\": -0.55, \"box_2d\": [892.9, 184.25, 1241.0, 347.35], \"box_3d\": [1.48, 1.49, 4.16, 5.25, 1.62, 7.41], \"ry\": 0.05}, {\"category\": \"Car\", \"angle\": -0.2, \"box_2d\": [751.82, 186.07, 991.3, 264.82], \"box_3d\": [1.51, 1.69, 4.42, 5.23, 1.81, 15.06], \"ry\": 0.12}, {\"category\": \"Car\", \"angle\": 2.7, \"box_2d\": [839.46, 179.43, 1178.51, 304.14], \"box_3d\": [1.56, 1.5, 3.8, 5.26, 1.67, 9.93], \"ry\": -3.11}]\n```" - } - ] - }, - { - "id": 752, - "data_source": "Mono3DRefer", - "scene_name": "002776", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [481.89, 184.32, 544.53, 229.67], \"box_3d\": [1.52, 1.73, 4.28, -3.57, 1.97, 27.0], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [544.18, 181.01, 572.7, 202.59], \"box_3d\": [1.44, 1.7, 3.65, -3.61, 2.03, 50.78], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 753, - "data_source": "Mono3DRefer", - "scene_name": "002797", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [704.01, 175.35, 875.15, 294.91], \"box_3d\": [1.49, 1.56, 4.34, 2.5, 1.55, 11.26], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.18, \"box_2d\": [0.0, 188.36, 415.81, 374.0], \"box_3d\": [1.42, 1.53, 4.12, -2.99, 1.59, 6.27], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [646.47, 175.94, 691.56, 213.76], \"box_3d\": [1.38, 1.64, 3.51, 2.23, 1.52, 28.28], \"ry\": -1.51}]\n```" - } - ] - }, - { - "id": 754, - "data_source": "Mono3DRefer", - "scene_name": "002798", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.68, \"box_2d\": [851.24, 180.89, 1241.0, 352.76], \"box_3d\": [1.49, 1.66, 4.16, 4.84, 1.6, 7.53], \"ry\": -0.13}, {\"category\": \"Car\", \"angle\": 2.68, \"box_2d\": [698.6, 159.51, 984.04, 280.92], \"box_3d\": [1.85, 1.68, 4.34, 3.82, 1.66, 12.32], \"ry\": 2.97}, {\"category\": \"Car\", \"angle\": -2.42, \"box_2d\": [0.0, 210.33, 89.43, 374.0], \"box_3d\": [1.51, 1.53, 4.46, -6.99, 1.87, 5.62], \"ry\": 2.99}, {\"category\": \"Car\", \"angle\": -2.72, \"box_2d\": [18.53, 193.36, 289.43, 283.68], \"box_3d\": [1.43, 1.56, 3.85, -8.12, 1.84, 13.07], \"ry\": 3.01}, {\"category\": \"Car\", \"angle\": -2.9, \"box_2d\": [205.42, 189.3, 354.98, 245.58], \"box_3d\": [1.42, 1.62, 3.5, -9.14, 1.91, 20.14], \"ry\": 2.96}, {\"category\": \"Car\", \"angle\": -2.97, \"box_2d\": [331.78, 185.5, 422.05, 219.26], \"box_3d\": [1.41, 1.54, 3.6, -10.35, 1.99, 32.12], \"ry\": 3.01}]\n```" - } - ] - }, - { - "id": 755, - "data_source": "Mono3DRefer", - "scene_name": "002800", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.0, \"box_2d\": [0.0, 188.99, 333.46, 373.0], \"box_3d\": [1.57, 1.67, 4.14, 0.0, 1.69, 7.74], \"ry\": 1.49}]\n```" - } - ] - }, - { - "id": 756, - "data_source": "Mono3DRefer", - "scene_name": "002818", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [660.5, 176.51, 718.74, 220.45], \"box_3d\": [1.51, 1.75, 3.78, 2.87, 1.67, 27.02], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [424.29, 191.91, 485.48, 230.78], \"box_3d\": [1.36, 1.51, 4.02, -6.13, 2.16, 28.75], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [468.65, 182.4, 513.81, 218.53], \"box_3d\": [1.74, 1.65, 4.11, -6.11, 2.25, 37.31], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [486.77, 183.01, 524.1, 211.22], \"box_3d\": [1.61, 1.64, 4.0, -6.35, 2.24, 43.98], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -0.88, \"box_2d\": [0.0, 211.22, 66.36, 249.94], \"box_3d\": [1.43, 1.8, 3.96, -25.9, 3.18, 31.14], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -0.93, \"box_2d\": [10.37, 209.64, 117.1, 248.65], \"box_3d\": [1.4, 1.6, 3.97, -23.48, 3.08, 31.15], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -0.97, \"box_2d\": [58.25, 205.12, 171.9, 247.63], \"box_3d\": [1.49, 1.7, 4.51, -20.82, 2.95, 30.56], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [378.88, 194.87, 454.8, 245.01], \"box_3d\": [1.42, 1.54, 3.5, -6.21, 2.18, 23.4], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 757, - "data_source": "Mono3DRefer", - "scene_name": "002828", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [321.49, 185.25, 386.17, 216.6], \"box_3d\": [1.55, 1.88, 4.59, -13.79, 2.24, 39.05], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 758, - "data_source": "Mono3DRefer", - "scene_name": "002842", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [472.0, 185.61, 537.75, 234.78], \"box_3d\": [1.45, 1.53, 3.98, -3.44, 1.9, 23.98], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 759, - "data_source": "Mono3DRefer", - "scene_name": "002844", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.18, \"box_2d\": [38.34, 198.36, 166.25, 248.85], \"box_3d\": [1.67, 1.82, 4.42, -19.53, 2.73, 27.97], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.99, \"box_2d\": [266.37, 192.33, 331.41, 226.56], \"box_3d\": [1.63, 1.87, 3.47, -16.24, 2.69, 37.77], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 760, - "data_source": "Mono3DRefer", - "scene_name": "002852", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.05, \"box_2d\": [0.0, 196.41, 359.18, 374.0], \"box_3d\": [1.7, 1.43, 3.2, -3.33, 1.93, 5.84], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.52, \"box_2d\": [513.87, 184.43, 539.47, 209.12], \"box_3d\": [1.68, 1.63, 3.9, -6.08, 2.54, 52.2], \"ry\": 1.4}]\n```" - } - ] - }, - { - "id": 761, - "data_source": "Mono3DRefer", - "scene_name": "002862", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.54, \"box_2d\": [525.69, 179.78, 560.13, 214.91], \"box_3d\": [1.68, 1.63, 3.9, -3.51, 2.05, 36.94], \"ry\": 1.45}, {\"category\": \"Car\", \"angle\": 1.45, \"box_2d\": [497.41, 181.69, 526.61, 201.74], \"box_3d\": [1.47, 1.7, 4.39, -7.63, 2.16, 55.77], \"ry\": 1.31}]\n```" - } - ] - }, - { - "id": 762, - "data_source": "Mono3DRefer", - "scene_name": "002867", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.97, \"box_2d\": [265.55, 181.29, 333.72, 216.73], \"box_3d\": [1.48, 1.61, 3.39, -13.87, 1.87, 32.34], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 2.03, \"box_2d\": [213.79, 183.27, 296.8, 216.44], \"box_3d\": [1.44, 1.79, 4.32, -16.75, 1.95, 34.25], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [377.03, 176.64, 422.16, 201.88], \"box_3d\": [1.6, 1.78, 3.86, -13.98, 1.87, 48.05], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [346.24, 176.66, 391.17, 201.61], \"box_3d\": [1.65, 1.77, 3.85, -16.72, 1.93, 50.1], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 763, - "data_source": "Mono3DRefer", - "scene_name": "002876", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.18, \"box_2d\": [667.92, 175.39, 767.79, 209.34], \"box_3d\": [1.45, 1.6, 4.23, 4.8, 1.59, 32.63], \"ry\": -0.04}, {\"category\": \"Car\", \"angle\": 3.11, \"box_2d\": [649.02, 175.08, 702.96, 195.39], \"box_3d\": [1.48, 1.72, 3.99, 4.95, 1.67, 54.72], \"ry\": -3.09}]\n```" - } - ] - }, - { - "id": 764, - "data_source": "Mono3DRefer", - "scene_name": "002882", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.04, \"box_2d\": [0.0, 201.04, 410.64, 374.0], \"box_3d\": [1.52, 1.73, 4.28, -3.41, 1.86, 6.77], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [499.73, 185.21, 554.13, 224.07], \"box_3d\": [1.44, 1.7, 3.65, -3.35, 1.97, 29.39], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 765, - "data_source": "Mono3DRefer", - "scene_name": "002884", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [394.18, 171.44, 478.03, 247.47], \"box_3d\": [1.68, 1.63, 3.9, -4.3, 1.68, 18.14], \"ry\": 1.44}, {\"category\": \"Car\", \"angle\": 1.5, \"box_2d\": [433.4, 175.57, 479.41, 211.04], \"box_3d\": [1.47, 1.7, 4.39, -7.0, 1.62, 32.46], \"ry\": 1.29}]\n```" - } - ] - }, - { - "id": 766, - "data_source": "Mono3DRefer", - "scene_name": "002891", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [421.33, 182.55, 456.81, 202.43], \"box_3d\": [1.55, 1.88, 4.59, -14.08, 2.36, 59.61], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 767, - "data_source": "Mono3DRefer", - "scene_name": "002893", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [791.51, 187.33, 910.34, 266.63], \"box_3d\": [1.52, 1.54, 4.5, 5.31, 1.88, 16.58], \"ry\": -1.49}, {\"category\": \"Car\", \"angle\": -2.4, \"box_2d\": [0.0, 160.59, 336.65, 334.4], \"box_3d\": [1.59, 1.45, 3.15, -4.78, 1.5, 7.6], \"ry\": -2.95}]\n```" - } - ] - }, - { - "id": 768, - "data_source": "Mono3DRefer", - "scene_name": "002901", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.07, \"box_2d\": [218.46, 186.14, 296.08, 220.91], \"box_3d\": [1.45, 1.74, 3.42, -15.96, 2.08, 32.75], \"ry\": 1.62}]\n```" - } - ] - }, - { - "id": 769, - "data_source": "Mono3DRefer", - "scene_name": "002908", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [254.58, 173.21, 322.13, 202.09], \"box_3d\": [1.34, 1.73, 3.98, -16.0, 1.39, 36.02], \"ry\": 1.53}]\n```" - } - ] - }, - { - "id": 770, - "data_source": "Mono3DRefer", - "scene_name": "002913", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.35, \"box_2d\": [0.0, 173.07, 51.71, 241.69], \"box_3d\": [1.78, 1.6, 4.29, -16.66, 1.79, 18.22], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 2.25, \"box_2d\": [45.81, 181.81, 173.67, 232.37], \"box_3d\": [1.5, 1.47, 3.62, -16.22, 1.81, 23.55], \"ry\": 1.65}]\n```" - } - ] - }, - { - "id": 771, - "data_source": "Mono3DRefer", - "scene_name": "002916", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.53, \"box_2d\": [545.29, 166.7, 569.52, 190.13], \"box_3d\": [1.57, 1.55, 3.32, -3.67, 1.18, 50.47], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [451.14, 179.86, 540.92, 240.43], \"box_3d\": [1.5, 1.7, 3.77, -3.08, 1.71, 19.97], \"ry\": 1.62}]\n```" - } - ] - }, - { - "id": 772, - "data_source": "Mono3DRefer", - "scene_name": "002928", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.09, \"box_2d\": [797.36, 164.62, 1241.0, 374.0], \"box_3d\": [1.59, 1.59, 3.89, 2.55, 1.58, 4.66], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [635.47, 172.07, 672.47, 203.15], \"box_3d\": [1.38, 1.64, 3.51, 2.05, 1.37, 34.05], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": -0.9, \"box_2d\": [0.0, 213.72, 325.96, 374.0], \"box_3d\": [1.33, 1.71, 4.09, -2.93, 1.62, 3.22], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [669.18, 170.16, 768.83, 243.39], \"box_3d\": [1.49, 1.56, 4.34, 2.38, 1.47, 17.03], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.36, \"box_2d\": [336.24, 181.41, 488.22, 285.87], \"box_3d\": [1.42, 1.53, 4.12, -3.14, 1.58, 12.06], \"ry\": -1.61}]\n```" - } - ] - }, - { - "id": 773, - "data_source": "Mono3DRefer", - "scene_name": "002932", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.96, \"box_2d\": [108.45, 198.69, 431.64, 374.0], \"box_3d\": [1.44, 1.72, 4.14, -3.45, 1.8, 8.19], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 774, - "data_source": "Mono3DRefer", - "scene_name": "002942", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [110.52, 190.59, 406.78, 368.47], \"box_3d\": [1.45, 1.64, 4.21, -3.67, 1.69, 8.34], \"ry\": 1.5}, {\"category\": \"Car\", \"angle\": 1.4, \"box_2d\": [656.62, 175.49, 744.23, 245.06], \"box_3d\": [1.62, 1.58, 3.89, 2.21, 1.71, 18.9], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [465.18, 182.62, 531.3, 228.64], \"box_3d\": [1.45, 1.67, 4.43, -3.88, 1.81, 25.44], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [637.68, 183.28, 699.03, 228.56], \"box_3d\": [1.47, 1.59, 3.62, 1.97, 1.85, 25.7], \"ry\": -1.66}]\n```" - } - ] - }, - { - "id": 775, - "data_source": "Mono3DRefer", - "scene_name": "002944", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [563.58, 167.5, 598.66, 194.93], \"box_3d\": [1.57, 1.67, 4.14, -1.17, 0.76, 43.61], \"ry\": 1.65}]\n```" - } - ] - }, - { - "id": 776, - "data_source": "Mono3DRefer", - "scene_name": "002953", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.76, \"box_2d\": [852.98, 168.11, 977.12, 213.19], \"box_3d\": [1.38, 1.43, 3.4, 9.8, 1.26, 23.4], \"ry\": -3.14}, {\"category\": \"Car\", \"angle\": 2.53, \"box_2d\": [993.78, 157.87, 1172.75, 209.67], \"box_3d\": [1.58, 1.79, 4.7, 15.31, 1.14, 23.53], \"ry\": 3.1}, {\"category\": \"Car\", \"angle\": 1.19, \"box_2d\": [252.65, 187.13, 324.39, 223.89], \"box_3d\": [1.52, 1.55, 3.91, -14.38, 2.16, 32.1], \"ry\": 0.77}]\n```" - } - ] - }, - { - "id": 777, - "data_source": "Mono3DRefer", - "scene_name": "002967", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [0.0, 206.9, 423.67, 374.0], \"box_3d\": [1.44, 1.63, 3.32, -2.6, 1.76, 5.11], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -2.13, \"box_2d\": [828.25, 197.05, 1241.0, 374.0], \"box_3d\": [1.55, 1.55, 3.27, 2.76, 1.75, 4.74], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [714.21, 192.19, 911.15, 334.22], \"box_3d\": [1.48, 1.6, 3.79, 2.46, 1.79, 9.82], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [481.06, 187.52, 556.88, 249.52], \"box_3d\": [1.56, 1.55, 3.64, -2.54, 2.0, 20.52], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [527.13, 174.84, 584.06, 232.94], \"box_3d\": [1.91, 1.66, 4.22, -1.91, 2.0, 25.98], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [561.1, 178.44, 610.32, 222.93], \"box_3d\": [1.83, 1.69, 4.28, -1.04, 2.09, 31.97], \"ry\": 1.65}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [618.74, 184.55, 658.11, 209.21], \"box_3d\": [1.51, 1.63, 4.03, 1.89, 2.3, 47.39], \"ry\": 1.86}]\n```" - } - ] - }, - { - "id": 778, - "data_source": "Mono3DRefer", - "scene_name": "002972", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.05, \"box_2d\": [837.69, 178.46, 1241.0, 374.0], \"box_3d\": [1.58, 1.67, 3.43, 2.85, 1.63, 4.86], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -2.65, \"box_2d\": [1218.94, 133.92, 1241.0, 225.41], \"box_3d\": [1.84, 1.94, 3.68, 13.96, 1.03, 13.35], \"ry\": -1.86}, {\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [361.38, 180.23, 487.65, 267.71], \"box_3d\": [1.51, 1.61, 3.57, -3.58, 1.67, 14.37], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [433.67, 178.56, 529.92, 240.95], \"box_3d\": [1.5, 1.6, 4.17, -3.38, 1.66, 19.52], \"ry\": 1.64}, {\"category\": \"Car\", \"angle\": 2.04, \"box_2d\": [915.63, 158.11, 1020.21, 214.81], \"box_3d\": [1.71, 1.62, 3.56, 11.66, 1.28, 23.5], \"ry\": 2.49}, {\"category\": \"Car\", \"angle\": 2.14, \"box_2d\": [841.8, 164.28, 938.19, 211.42], \"box_3d\": [1.52, 1.58, 3.28, 9.64, 1.25, 24.86], \"ry\": 2.5}, {\"category\": \"Car\", \"angle\": 2.21, \"box_2d\": [810.7, 163.58, 892.99, 206.65], \"box_3d\": [1.55, 1.47, 2.97, 9.2, 1.23, 27.5], \"ry\": 2.53}, {\"category\": \"Car\", \"angle\": 2.32, \"box_2d\": [793.38, 167.91, 877.67, 206.13], \"box_3d\": [1.55, 1.43, 3.49, 9.59, 1.35, 30.74], \"ry\": 2.61}, {\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [677.87, 175.77, 739.29, 223.55], \"box_3d\": [1.58, 1.68, 3.93, 3.44, 1.7, 26.02], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [513.68, 173.73, 569.18, 213.63], \"box_3d\": [1.58, 1.54, 3.98, -2.88, 1.64, 30.69], \"ry\": 1.69}, {\"category\": \"Car\", \"angle\": 3.0, \"box_2d\": [726.2, 163.23, 790.27, 191.65], \"box_3d\": [1.71, 1.67, 3.7, 9.21, 1.14, 45.13], \"ry\": -3.08}]\n```" - } - ] - }, - { - "id": 779, - "data_source": "Mono3DRefer", - "scene_name": "002996", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [127.31, 201.27, 457.06, 374.0], \"box_3d\": [1.44, 1.63, 3.32, -2.63, 1.77, 6.71], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -2.02, \"box_2d\": [782.82, 192.9, 1164.62, 374.0], \"box_3d\": [1.55, 1.55, 3.27, 2.74, 1.76, 6.35], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [699.74, 189.17, 859.58, 306.25], \"box_3d\": [1.48, 1.6, 3.79, 2.43, 1.77, 11.42], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [489.94, 185.95, 559.05, 242.87], \"box_3d\": [1.56, 1.55, 3.64, -2.57, 1.98, 22.1], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [530.43, 173.66, 584.13, 228.09], \"box_3d\": [1.91, 1.66, 4.22, -1.97, 1.96, 27.59], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [562.54, 177.04, 609.24, 219.3], \"box_3d\": [1.83, 1.69, 4.28, -1.08, 2.04, 33.55], \"ry\": 1.65}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [617.19, 184.77, 655.33, 208.61], \"box_3d\": [1.51, 1.63, 4.03, 1.81, 2.35, 48.98], \"ry\": 1.86}]\n```" - } - ] - }, - { - "id": 780, - "data_source": "Mono3DRefer", - "scene_name": "003032", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.09, \"box_2d\": [0.0, 230.36, 343.0, 374.0], \"box_3d\": [1.27, 1.54, 3.58, -3.74, 1.89, 6.13], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [295.11, 182.01, 461.7, 297.01], \"box_3d\": [1.63, 1.65, 3.91, -3.79, 1.8, 12.37], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [616.34, 176.04, 660.05, 216.25], \"box_3d\": [1.46, 1.53, 3.17, 1.01, 1.6, 28.07], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 781, - "data_source": "Mono3DRefer", - "scene_name": "003044", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.95, \"box_2d\": [0.0, 197.45, 96.07, 243.26], \"box_3d\": [1.52, 1.71, 4.11, -21.99, 2.52, 27.66], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": -1.06, \"box_2d\": [49.25, 199.0, 160.92, 242.31], \"box_3d\": [1.42, 1.7, 4.14, -19.37, 2.49, 27.86], \"ry\": -1.66}]\n```" - } - ] - }, - { - "id": 782, - "data_source": "Mono3DRefer", - "scene_name": "003045", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [510.7, 189.02, 561.67, 228.0], \"box_3d\": [1.42, 1.66, 4.11, -2.99, 2.12, 29.59], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 783, - "data_source": "Mono3DRefer", - "scene_name": "003047", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [551.06, 167.85, 582.87, 192.57], \"box_3d\": [1.57, 1.67, 4.14, -2.25, 0.7, 48.17], \"ry\": 1.63}]\n```" - } - ] - }, - { - "id": 784, - "data_source": "Mono3DRefer", - "scene_name": "003053", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.93, \"box_2d\": [307.13, 178.74, 451.93, 227.73], \"box_3d\": [1.47, 1.53, 4.21, -7.39, 1.69, 23.3], \"ry\": 3.05}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [645.42, 169.93, 726.89, 235.13], \"box_3d\": [1.64, 1.62, 4.2, 2.01, 1.6, 20.5], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [429.75, 184.08, 520.89, 251.21], \"box_3d\": [1.35, 1.47, 3.43, -3.04, 1.63, 16.63], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -2.55, \"box_2d\": [0.0, 188.22, 220.95, 344.02], \"box_3d\": [1.48, 1.61, 4.24, -6.81, 1.68, 7.74], \"ry\": 3.03}]\n```" - } - ] - }, - { - "id": 785, - "data_source": "Mono3DRefer", - "scene_name": "003068", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.03, \"box_2d\": [17.19, 184.7, 368.53, 374.0], \"box_3d\": [1.59, 1.63, 3.83, -4.12, 1.74, 7.84], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [658.48, 165.88, 696.48, 201.6], \"box_3d\": [1.62, 1.7, 3.29, 3.18, 1.33, 34.81], \"ry\": -1.51}]\n```" - } - ] - }, - { - "id": 786, - "data_source": "Mono3DRefer", - "scene_name": "003069", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [486.41, 174.39, 536.81, 205.38], \"box_3d\": [1.42, 1.78, 3.62, -4.34, 1.1, 35.17], \"ry\": 1.63}]\n```" - } - ] - }, - { - "id": 787, - "data_source": "Mono3DRefer", - "scene_name": "003081", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [259.4, 181.19, 507.03, 352.54], \"box_3d\": [1.57, 1.54, 4.01, -2.44, 1.68, 8.67], \"ry\": 1.62}]\n```" - } - ] - }, - { - "id": 788, - "data_source": "Mono3DRefer", - "scene_name": "003087", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.86, \"box_2d\": [0.0, 193.98, 299.45, 374.0], \"box_3d\": [1.47, 1.6, 3.66, -2.94, 1.6, 3.06], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [472.34, 169.56, 548.58, 229.35], \"box_3d\": [1.7, 1.63, 4.08, -3.07, 1.64, 22.74], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [707.14, 173.61, 858.8, 299.0], \"box_3d\": [1.59, 1.59, 2.47, 2.34, 1.62, 10.47], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [686.2, 184.52, 777.2, 258.48], \"box_3d\": [1.37, 1.59, 3.22, 2.43, 1.64, 15.28], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": -1.43, \"box_2d\": [514.76, 176.66, 566.48, 212.33], \"box_3d\": [1.4, 1.67, 3.72, -2.91, 1.58, 30.54], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": 1.52, \"box_2d\": [632.75, 173.59, 660.18, 197.32], \"box_3d\": [1.49, 1.61, 3.74, 2.34, 1.56, 47.54], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -2.73, \"box_2d\": [177.94, 166.1, 403.54, 241.49], \"box_3d\": [1.59, 1.63, 4.43, -7.22, 1.48, 16.49], \"ry\": -3.14}]\n```" - } - ] - }, - { - "id": 789, - "data_source": "Mono3DRefer", - "scene_name": "003102", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [622.66, 174.57, 726.06, 267.98], \"box_3d\": [1.46, 1.53, 3.17, 1.0, 1.51, 12.99], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 790, - "data_source": "Mono3DRefer", - "scene_name": "003104", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.03, \"box_2d\": [0.0, 148.83, 368.71, 374.0], \"box_3d\": [1.91, 1.66, 4.22, -3.18, 1.83, 5.17], \"ry\": 1.5}, {\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [311.13, 182.88, 495.71, 331.88], \"box_3d\": [1.83, 1.69, 4.28, -2.96, 2.0, 11.15], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [524.03, 194.51, 596.89, 241.22], \"box_3d\": [1.51, 1.63, 4.03, -1.79, 2.36, 26.81], \"ry\": 1.78}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [575.22, 192.66, 628.67, 230.7], \"box_3d\": [1.57, 1.44, 3.6, -0.34, 2.52, 33.08], \"ry\": 1.86}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [622.68, 188.98, 669.13, 219.03], \"box_3d\": [1.47, 1.51, 3.5, 1.92, 2.35, 38.34], \"ry\": 1.91}, {\"category\": \"Car\", \"angle\": -1.24, \"box_2d\": [786.08, 179.09, 823.5, 200.17], \"box_3d\": [1.77, 1.75, 4.67, 17.21, 2.34, 63.64], \"ry\": -0.98}]\n```" - } - ] - }, - { - "id": 791, - "data_source": "Mono3DRefer", - "scene_name": "003123", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.38, \"box_2d\": [307.62, 185.67, 444.03, 224.99], \"box_3d\": [1.36, 1.68, 4.49, -8.64, 1.86, 26.71], \"ry\": 0.07}]\n```" - } - ] - }, - { - "id": 792, - "data_source": "Mono3DRefer", - "scene_name": "003145", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.59, \"box_2d\": [153.0, 196.31, 188.73, 225.51], \"box_3d\": [1.45, 1.62, 3.83, -24.34, 2.79, 40.01], \"ry\": 1.04}]\n```" - } - ] - }, - { - "id": 793, - "data_source": "Mono3DRefer", - "scene_name": "003152", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [394.33, 188.38, 511.88, 265.4], \"box_3d\": [1.45, 1.77, 4.04, -3.41, 1.83, 16.21], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 794, - "data_source": "Mono3DRefer", - "scene_name": "003159", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.45, \"box_2d\": [999.43, 196.82, 1241.0, 374.0], \"box_3d\": [1.39, 1.45, 3.44, 2.96, 1.52, 2.37], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -2.47, \"box_2d\": [0.0, 205.51, 123.41, 244.43], \"box_3d\": [1.3, 1.7, 3.97, -20.6, 2.56, 26.7], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": 2.81, \"box_2d\": [826.05, 167.28, 930.55, 205.79], \"box_3d\": [1.38, 1.43, 3.4, 10.05, 1.2, 27.27], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": 2.6, \"box_2d\": [944.65, 163.68, 1095.78, 208.17], \"box_3d\": [1.58, 1.79, 4.7, 15.38, 1.27, 27.26], \"ry\": 3.11}]\n```" - } - ] - }, - { - "id": 795, - "data_source": "Mono3DRefer", - "scene_name": "003161", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [508.92, 197.24, 557.71, 239.89], \"box_3d\": [1.5, 1.62, 4.19, -3.09, 2.56, 29.46], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 796, - "data_source": "Mono3DRefer", - "scene_name": "003165", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.03, \"box_2d\": [201.39, 191.92, 290.89, 234.37], \"box_3d\": [1.45, 1.64, 3.61, -13.93, 2.22, 27.77], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [388.2, 182.81, 431.66, 206.63], \"box_3d\": [1.57, 1.9, 4.16, -14.05, 2.29, 50.81], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 797, - "data_source": "Mono3DRefer", - "scene_name": "003166", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.67, \"box_2d\": [1038.32, 168.53, 1241.0, 287.41], \"box_3d\": [1.42, 1.41, 4.3, 8.1, 1.37, 9.34], \"ry\": 0.03}]\n```" - } - ] - }, - { - "id": 798, - "data_source": "Mono3DRefer", - "scene_name": "003172", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.11, \"box_2d\": [0.0, 189.43, 274.39, 347.91], \"box_3d\": [1.49, 1.63, 3.68, -5.84, 1.73, 8.9], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": -2.1, \"box_2d\": [827.69, 184.91, 1241.0, 374.0], \"box_3d\": [1.49, 1.61, 3.55, 3.06, 1.6, 5.56], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": 1.26, \"box_2d\": [464.88, 178.2, 519.18, 209.27], \"box_3d\": [1.41, 1.57, 3.53, -5.81, 1.69, 35.05], \"ry\": 1.09}, {\"category\": \"Car\", \"angle\": 2.62, \"box_2d\": [544.07, 173.59, 607.48, 195.16], \"box_3d\": [1.47, 1.72, 4.22, -2.45, 1.54, 51.6], \"ry\": 2.58}, {\"category\": \"Car\", \"angle\": 0.75, \"box_2d\": [341.95, 179.52, 418.58, 205.38], \"box_3d\": [1.42, 1.71, 4.14, -13.26, 1.83, 41.53], \"ry\": 0.44}, {\"category\": \"Car\", \"angle\": 2.74, \"box_2d\": [649.14, 172.8, 716.3, 193.21], \"box_3d\": [1.48, 1.87, 4.74, 5.53, 1.51, 55.13], \"ry\": 2.84}]\n```" - } - ] - }, - { - "id": 799, - "data_source": "Mono3DRefer", - "scene_name": "003191", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [400.8, 185.14, 511.67, 261.49], \"box_3d\": [1.44, 1.62, 3.67, -3.28, 1.73, 15.83], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [487.45, 179.9, 550.28, 228.49], \"box_3d\": [1.51, 1.62, 3.68, -3.06, 1.76, 24.57], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 800, - "data_source": "Mono3DRefer", - "scene_name": "003205", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [0.0, 173.41, 409.65, 374.0], \"box_3d\": [1.65, 1.75, 3.39, -2.57, 1.68, 4.71], \"ry\": 1.47}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [369.67, 180.82, 521.75, 293.95], \"box_3d\": [1.57, 1.54, 4.01, -2.6, 1.71, 12.12], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 801, - "data_source": "Mono3DRefer", - "scene_name": "003217", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.0, \"box_2d\": [243.42, 189.97, 323.24, 225.32], \"box_3d\": [1.56, 1.91, 4.44, -16.05, 2.44, 35.64], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 802, - "data_source": "Mono3DRefer", - "scene_name": "003222", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.16, \"box_2d\": [112.83, 187.73, 220.84, 231.81], \"box_3d\": [1.45, 1.74, 3.42, -16.07, 2.02, 26.3], \"ry\": 1.61}]\n```" - } - ] - }, - { - "id": 803, - "data_source": "Mono3DRefer", - "scene_name": "003223", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [509.71, 174.93, 565.71, 203.59], \"box_3d\": [1.37, 1.78, 4.65, -3.68, 1.5, 37.24], \"ry\": 1.72}]\n```" - } - ] - }, - { - "id": 804, - "data_source": "Mono3DRefer", - "scene_name": "003248", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [443.74, 184.29, 493.62, 229.14], \"box_3d\": [1.65, 1.75, 3.39, -5.62, 2.12, 28.83], \"ry\": 1.43}]\n```" - } - ] - }, - { - "id": 805, - "data_source": "Mono3DRefer", - "scene_name": "003276", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.12, \"box_2d\": [119.15, 195.55, 242.72, 244.0], \"box_3d\": [1.46, 1.86, 4.01, -14.86, 2.3, 25.21], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 806, - "data_source": "Mono3DRefer", - "scene_name": "003279", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -3.11, \"box_2d\": [391.56, 180.03, 661.98, 271.9], \"box_3d\": [1.47, 1.77, 4.49, -1.48, 1.62, 12.97], \"ry\": 3.06}, {\"category\": \"Car\", \"angle\": -1.61, \"box_2d\": [536.85, 184.41, 564.41, 208.85], \"box_3d\": [1.52, 1.62, 4.5, -4.03, 2.31, 48.19], \"ry\": -1.7}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [441.1, 188.1, 476.01, 212.74], \"box_3d\": [1.42, 1.66, 4.5, -9.57, 2.42, 45.73], \"ry\": 1.48}]\n```" - } - ] - }, - { - "id": 807, - "data_source": "Mono3DRefer", - "scene_name": "003288", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [697.78, 166.52, 853.84, 286.11], \"box_3d\": [1.76, 1.71, 4.35, 2.69, 1.7, 12.9], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -2.41, \"box_2d\": [0.0, 185.52, 212.03, 348.81], \"box_3d\": [1.52, 1.58, 3.61, -6.49, 1.68, 7.63], \"ry\": -3.1}, {\"category\": \"Car\", \"angle\": 0.53, \"box_2d\": [0.0, 184.8, 334.04, 309.31], \"box_3d\": [1.52, 1.71, 4.49, -6.44, 1.71, 9.98], \"ry\": -0.03}, {\"category\": \"Car\", \"angle\": -2.62, \"box_2d\": [98.2, 185.86, 369.88, 280.62], \"box_3d\": [1.49, 1.61, 3.87, -6.4, 1.74, 12.44], \"ry\": -3.09}, {\"category\": \"Car\", \"angle\": -2.71, \"box_2d\": [190.61, 181.71, 392.39, 259.98], \"box_3d\": [1.5, 1.58, 3.47, -6.54, 1.7, 14.93], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": 0.33, \"box_2d\": [273.39, 164.91, 463.75, 242.2], \"box_3d\": [1.95, 1.64, 4.57, -6.44, 1.78, 19.42], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": -2.78, \"box_2d\": [258.17, 176.29, 422.6, 251.0], \"box_3d\": [1.66, 1.58, 3.31, -6.37, 1.77, 17.16], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [677.52, 183.29, 770.37, 246.67], \"box_3d\": [1.44, 1.66, 3.25, 2.76, 1.72, 18.35], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": -2.85, \"box_2d\": [318.22, 166.59, 452.41, 232.42], \"box_3d\": [1.87, 1.61, 3.53, -6.73, 1.72, 21.72], \"ry\": 3.14}, {\"category\": \"Car\", \"angle\": -2.86, \"box_2d\": [357.86, 174.58, 487.87, 223.98], \"box_3d\": [1.72, 1.65, 4.33, -6.85, 1.81, 26.54], \"ry\": -3.11}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [641.59, 179.1, 681.01, 210.3], \"box_3d\": [1.45, 1.59, 3.59, 2.47, 1.78, 35.8], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [663.45, 178.46, 728.46, 229.71], \"box_3d\": [1.54, 1.67, 3.72, 2.71, 1.74, 23.77], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 808, - "data_source": "Mono3DRefer", - "scene_name": "003289", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [602.75, 177.49, 673.26, 234.8], \"box_3d\": [1.36, 1.69, 3.38, 0.63, 1.5, 19.1], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.48, \"box_2d\": [1044.38, 168.43, 1193.57, 210.72], \"box_3d\": [1.45, 1.67, 4.32, 18.46, 1.32, 26.32], \"ry\": 3.09}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [374.66, 177.84, 426.84, 208.43], \"box_3d\": [1.59, 1.72, 3.86, -11.51, 1.88, 39.82], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [583.69, 175.82, 627.1, 215.44], \"box_3d\": [1.52, 1.67, 3.61, -0.22, 1.66, 29.73], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 809, - "data_source": "Mono3DRefer", - "scene_name": "003293", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.52, \"box_2d\": [691.09, 177.39, 718.63, 201.62], \"box_3d\": [1.54, 1.63, 3.67, 6.34, 1.86, 48.18], \"ry\": -1.39}, {\"category\": \"Car\", \"angle\": -1.98, \"box_2d\": [799.44, 181.62, 1174.24, 374.0], \"box_3d\": [1.58, 1.67, 3.84, 3.18, 1.68, 7.12], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [519.17, 182.55, 597.49, 237.78], \"box_3d\": [1.53, 1.67, 3.78, -1.53, 1.84, 22.14], \"ry\": 1.7}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [763.68, 147.92, 914.32, 297.61], \"box_3d\": [2.05, 1.5, 3.83, 3.52, 1.74, 11.87], \"ry\": -1.49}]\n```" - } - ] - }, - { - "id": 810, - "data_source": "Mono3DRefer", - "scene_name": "003297", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.55, \"box_2d\": [655.44, 181.0, 784.23, 297.56], \"box_3d\": [1.57, 1.73, 4.15, 1.69, 1.72, 11.93], \"ry\": 1.69}]\n```" - } - ] - }, - { - "id": 811, - "data_source": "Mono3DRefer", - "scene_name": "003320", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [1184.31, 178.71, 1241.0, 331.02], \"box_3d\": [1.46, 1.49, 3.63, 7.71, 1.51, 7.71], \"ry\": 2.55}, {\"category\": \"Car\", \"angle\": 2.58, \"box_2d\": [886.11, 150.38, 1206.8, 279.9], \"box_3d\": [1.71, 1.67, 3.7, 6.2, 1.43, 10.59], \"ry\": 3.09}, {\"category\": \"Car\", \"angle\": 2.09, \"box_2d\": [0.0, 189.02, 272.92, 374.0], \"box_3d\": [1.64, 1.63, 4.51, -4.42, 1.8, 5.81], \"ry\": 1.46}, {\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [183.24, 185.16, 394.51, 313.5], \"box_3d\": [1.6, 1.66, 4.33, -4.82, 1.82, 11.44], \"ry\": 1.51}]\n```" - } - ] - }, - { - "id": 812, - "data_source": "Mono3DRefer", - "scene_name": "003332", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [547.12, 175.78, 580.33, 201.5], \"box_3d\": [1.42, 1.66, 4.11, -2.7, 1.61, 42.4], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [369.13, 180.26, 513.26, 281.66], \"box_3d\": [1.6, 1.68, 3.94, -2.99, 1.74, 13.43], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 813, - "data_source": "Mono3DRefer", - "scene_name": "003346", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [663.4, 176.74, 715.27, 219.44], \"box_3d\": [1.6, 1.53, 4.18, 3.12, 1.77, 29.34], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [485.35, 176.15, 558.63, 233.86], \"box_3d\": [1.55, 1.68, 4.42, -2.59, 1.67, 21.8], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 814, - "data_source": "Mono3DRefer", - "scene_name": "003351", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [398.61, 179.27, 437.37, 199.08], \"box_3d\": [1.56, 1.91, 4.44, -15.84, 2.1, 59.69], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 2.06, \"box_2d\": [182.21, 189.03, 284.2, 235.82], \"box_3d\": [1.48, 1.71, 3.54, -13.24, 2.08, 25.5], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 815, - "data_source": "Mono3DRefer", - "scene_name": "003358", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.13, \"box_2d\": [864.03, 211.46, 1241.0, 374.0], \"box_3d\": [1.37, 1.51, 3.39, 2.68, 1.66, 3.92], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [496.54, 175.04, 561.58, 227.74], \"box_3d\": [1.55, 1.68, 4.42, -2.59, 1.65, 23.67], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [660.01, 175.9, 707.99, 215.92], \"box_3d\": [1.6, 1.53, 4.18, 3.1, 1.75, 31.16], \"ry\": -1.59}]\n```" - } - ] - }, - { - "id": 816, - "data_source": "Mono3DRefer", - "scene_name": "003360", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.41, \"box_2d\": [609.07, 186.03, 658.34, 218.99], \"box_3d\": [1.44, 1.66, 4.53, 1.18, 2.1, 34.77], \"ry\": -1.38}]\n```" - } - ] - }, - { - "id": 817, - "data_source": "Mono3DRefer", - "scene_name": "003376", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [538.88, 175.16, 565.94, 200.78], \"box_3d\": [1.7, 1.72, 4.11, -4.01, 1.89, 50.45], \"ry\": 1.53}]\n```" - } - ] - }, - { - "id": 818, - "data_source": "Mono3DRefer", - "scene_name": "003409", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [501.92, 180.86, 540.94, 206.64], \"box_3d\": [1.43, 1.61, 4.13, -5.2, 1.91, 42.55], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [665.64, 171.65, 705.02, 202.59], \"box_3d\": [1.47, 1.6, 3.87, 3.74, 1.44, 36.62], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -2.36, \"box_2d\": [945.13, 177.69, 1241.0, 374.0], \"box_3d\": [1.55, 1.66, 4.15, 3.05, 1.57, 2.76], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -0.91, \"box_2d\": [0.0, 199.64, 230.34, 374.0], \"box_3d\": [1.63, 1.61, 3.81, -5.83, 1.97, 7.48], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [330.46, 186.13, 442.08, 254.34], \"box_3d\": [1.53, 1.68, 3.86, -5.67, 1.9, 18.63], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [400.85, 186.02, 477.97, 233.93], \"box_3d\": [1.42, 1.55, 3.85, -5.61, 1.88, 24.0], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [443.01, 183.76, 497.49, 222.72], \"box_3d\": [1.43, 1.48, 3.71, -5.56, 1.88, 28.9], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [471.31, 182.89, 518.51, 214.39], \"box_3d\": [1.4, 1.65, 4.02, -5.52, 1.9, 34.82], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 2.85, \"box_2d\": [782.77, 154.41, 889.68, 195.34], \"box_3d\": [1.64, 1.59, 3.99, 9.45, 0.91, 30.42], \"ry\": -3.14}, {\"category\": \"Car\", \"angle\": 2.87, \"box_2d\": [756.29, 156.1, 835.9, 191.9], \"box_3d\": [1.67, 1.67, 3.43, 9.03, 0.9, 35.29], \"ry\": 3.12}, {\"category\": \"Car\", \"angle\": -0.66, \"box_2d\": [1107.58, 133.23, 1241.0, 173.32], \"box_3d\": [1.53, 1.85, 4.27, 23.15, -0.01, 29.57], \"ry\": 0.0}, {\"category\": \"Car\", \"angle\": -0.61, \"box_2d\": [1065.54, 135.26, 1196.74, 175.01], \"box_3d\": [1.67, 1.85, 4.49, 23.2, 0.07, 32.27], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": -0.59, \"box_2d\": [1036.7, 140.17, 1164.17, 173.82], \"box_3d\": [1.53, 1.85, 4.94, 23.84, 0.01, 35.23], \"ry\": -0.0}, {\"category\": \"Car\", \"angle\": -0.58, \"box_2d\": [1011.22, 141.36, 1128.71, 173.61], \"box_3d\": [1.57, 1.85, 4.92, 23.8, 0.01, 37.48], \"ry\": -0.01}]\n```" - } - ] - }, - { - "id": 819, - "data_source": "Mono3DRefer", - "scene_name": "003410", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.18, \"box_2d\": [0.92, 197.79, 430.44, 374.0], \"box_3d\": [1.33, 1.71, 4.09, -2.97, 1.62, 6.47], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [654.5, 171.96, 734.69, 231.36], \"box_3d\": [1.49, 1.56, 4.34, 2.25, 1.5, 20.49], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": -1.4, \"box_2d\": [396.29, 179.49, 504.94, 257.91], \"box_3d\": [1.42, 1.53, 4.12, -3.28, 1.58, 15.34], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.88, \"box_2d\": [731.01, 167.06, 1005.9, 356.93], \"box_3d\": [1.59, 1.59, 3.89, 2.48, 1.57, 8.08], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [626.66, 173.05, 660.48, 201.29], \"box_3d\": [1.38, 1.64, 3.51, 1.75, 1.41, 37.29], \"ry\": -1.5}]\n```" - } - ] - }, - { - "id": 820, - "data_source": "Mono3DRefer", - "scene_name": "003416", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.48, \"box_2d\": [157.89, 193.75, 456.71, 285.39], \"box_3d\": [1.38, 1.66, 4.38, -5.03, 1.77, 12.17], \"ry\": 0.09}, {\"category\": \"Car\", \"angle\": -0.47, \"box_2d\": [831.68, 172.3, 1241.0, 333.3], \"box_3d\": [1.48, 1.86, 4.18, 4.66, 1.49, 7.62], \"ry\": 0.06}, {\"category\": \"Car\", \"angle\": -0.42, \"box_2d\": [767.72, 159.48, 1106.2, 287.31], \"box_3d\": [1.74, 1.66, 4.37, 4.74, 1.58, 10.94], \"ry\": -0.03}, {\"category\": \"Car\", \"angle\": -2.83, \"box_2d\": [368.31, 185.46, 518.33, 235.28], \"box_3d\": [1.47, 1.67, 4.28, -5.21, 1.89, 22.68], \"ry\": -3.05}, {\"category\": \"Car\", \"angle\": -3.05, \"box_2d\": [418.33, 183.49, 536.85, 221.76], \"box_3d\": [1.38, 1.67, 4.37, -5.09, 1.81, 27.96], \"ry\": 3.05}]\n```" - } - ] - }, - { - "id": 821, - "data_source": "Mono3DRefer", - "scene_name": "003445", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.58, \"box_2d\": [587.35, 183.5, 625.37, 217.88], \"box_3d\": [1.41, 1.59, 4.47, -0.19, 1.91, 32.67], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 822, - "data_source": "Mono3DRefer", - "scene_name": "003451", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [417.51, 176.26, 485.7, 241.28], \"box_3d\": [1.68, 1.63, 3.9, -4.53, 1.8, 20.89], \"ry\": 1.43}, {\"category\": \"Car\", \"angle\": 1.48, \"box_2d\": [439.4, 180.02, 483.18, 212.5], \"box_3d\": [1.47, 1.7, 4.39, -7.36, 1.83, 35.3], \"ry\": 1.28}]\n```" - } - ] - }, - { - "id": 823, - "data_source": "Mono3DRefer", - "scene_name": "003452", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.49, \"box_2d\": [350.72, 192.41, 387.53, 223.18], \"box_3d\": [1.65, 1.75, 3.39, -14.11, 2.82, 42.09], \"ry\": 1.17}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [242.66, 194.15, 451.83, 336.97], \"box_3d\": [1.47, 1.7, 4.39, -3.43, 1.82, 10.17], \"ry\": 1.48}]\n```" - } - ] - }, - { - "id": 824, - "data_source": "Mono3DRefer", - "scene_name": "003466", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.62, \"box_2d\": [557.34, 157.01, 585.8, 180.84], \"box_3d\": [1.49, 1.68, 4.08, -1.91, -0.06, 47.44], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 825, - "data_source": "Mono3DRefer", - "scene_name": "003478", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [438.91, 171.93, 537.88, 252.06], \"box_3d\": [1.7, 1.63, 4.08, -2.85, 1.71, 17.5], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -1.98, \"box_2d\": [803.1, 184.61, 1220.06, 374.0], \"box_3d\": [1.59, 1.59, 2.47, 2.44, 1.69, 5.1], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [726.63, 191.29, 895.69, 315.47], \"box_3d\": [1.37, 1.59, 3.22, 2.53, 1.66, 9.91], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": -1.41, \"box_2d\": [502.82, 182.12, 567.59, 225.9], \"box_3d\": [1.4, 1.67, 3.72, -2.59, 1.74, 25.38], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": 1.51, \"box_2d\": [644.21, 177.53, 676.25, 204.33], \"box_3d\": [1.49, 1.61, 3.74, 2.88, 1.78, 42.33], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -2.58, \"box_2d\": [0.0, 170.29, 320.85, 283.07], \"box_3d\": [1.59, 1.63, 4.43, -7.1, 1.59, 11.25], \"ry\": -3.13}]\n```" - } - ] - }, - { - "id": 826, - "data_source": "Mono3DRefer", - "scene_name": "003483", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.04, \"box_2d\": [0.0, 157.56, 411.14, 374.0], \"box_3d\": [1.7, 1.63, 4.08, -3.01, 1.65, 5.73], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.33, \"box_2d\": [375.78, 188.05, 513.37, 276.76], \"box_3d\": [1.4, 1.67, 3.72, -3.01, 1.72, 13.71], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": 1.47, \"box_2d\": [634.67, 181.93, 682.42, 219.87], \"box_3d\": [1.49, 1.61, 3.74, 1.96, 1.88, 30.59], \"ry\": 1.53}]\n```" - } - ] - }, - { - "id": 827, - "data_source": "Mono3DRefer", - "scene_name": "003486", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.51, \"box_2d\": [694.16, 180.37, 722.37, 205.02], \"box_3d\": [1.54, 1.63, 3.67, 6.49, 2.05, 47.57], \"ry\": -1.38}, {\"category\": \"Car\", \"angle\": 2.43, \"box_2d\": [0.0, 152.96, 135.69, 361.68], \"box_3d\": [1.93, 1.59, 4.01, -6.52, 1.76, 6.43], \"ry\": 1.66}, {\"category\": \"Car\", \"angle\": -1.85, \"box_2d\": [758.17, 178.74, 959.72, 311.06], \"box_3d\": [1.58, 1.67, 3.84, 3.36, 1.68, 10.63], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [570.63, 182.12, 628.06, 221.85], \"box_3d\": [1.53, 1.67, 3.78, -0.41, 1.93, 30.02], \"ry\": 1.75}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [744.48, 157.93, 848.68, 268.8], \"box_3d\": [2.05, 1.5, 3.83, 3.79, 1.8, 15.34], \"ry\": -1.49}]\n```" - } - ] - }, - { - "id": 828, - "data_source": "Mono3DRefer", - "scene_name": "003494", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [681.42, 180.55, 785.6, 256.04], \"box_3d\": [1.5, 1.59, 3.79, 2.64, 1.69, 16.44], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [647.47, 178.91, 690.39, 215.39], \"box_3d\": [1.49, 1.57, 4.07, 2.5, 1.76, 31.74], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [667.47, 177.38, 738.4, 235.74], \"box_3d\": [1.57, 1.59, 3.75, 2.65, 1.73, 21.55], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 0.3, \"box_2d\": [322.53, 190.21, 410.74, 222.28], \"box_3d\": [1.52, 1.51, 3.94, -12.25, 2.42, 36.37], \"ry\": -0.02}, {\"category\": \"Car\", \"angle\": 0.46, \"box_2d\": [204.7, 191.63, 304.17, 225.1], \"box_3d\": [1.58, 1.56, 4.24, -17.9, 2.56, 36.36], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [635.58, 178.6, 662.77, 201.2], \"box_3d\": [1.49, 1.55, 3.93, 2.67, 1.91, 50.2], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [594.27, 178.41, 619.2, 202.13], \"box_3d\": [1.54, 1.63, 3.67, -0.25, 1.94, 49.35], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [143.78, 197.57, 449.39, 374.0], \"box_3d\": [1.45, 1.59, 3.59, -2.91, 1.75, 7.45], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -2.31, \"box_2d\": [904.66, 178.71, 1241.0, 374.0], \"box_3d\": [1.64, 1.64, 3.89, 2.72, 1.67, 2.74], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.98, \"box_2d\": [785.41, 170.22, 1074.24, 374.0], \"box_3d\": [1.69, 1.39, 2.65, 2.67, 1.7, 6.7], \"ry\": -1.62}]\n```" - } - ] - }, - { - "id": 829, - "data_source": "Mono3DRefer", - "scene_name": "003516", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.57, \"box_2d\": [0.0, 208.84, 238.54, 338.87], \"box_3d\": [1.42, 1.62, 3.5, -6.97, 1.94, 9.22], \"ry\": 3.07}, {\"category\": \"Car\", \"angle\": -2.86, \"box_2d\": [299.0, 190.28, 439.26, 242.43], \"box_3d\": [1.41, 1.54, 3.6, -7.02, 1.95, 21.13], \"ry\": 3.1}]\n```" - } - ] - }, - { - "id": 830, - "data_source": "Mono3DRefer", - "scene_name": "003517", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [402.2, 187.19, 504.27, 256.75], \"box_3d\": [1.39, 1.52, 3.56, -3.55, 1.75, 16.73], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [525.35, 182.16, 553.42, 203.82], \"box_3d\": [1.34, 1.64, 4.57, -4.67, 1.98, 47.96], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": -2.87, \"box_2d\": [0.0, 194.19, 127.2, 312.62], \"box_3d\": [1.53, 1.54, 4.35, -9.31, 1.86, 9.79], \"ry\": 2.67}, {\"category\": \"Car\", \"angle\": 2.95, \"box_2d\": [423.23, 181.17, 505.37, 210.82], \"box_3d\": [1.52, 1.41, 4.22, -8.02, 2.01, 39.6], \"ry\": 2.75}]\n```" - } - ] - }, - { - "id": 831, - "data_source": "Mono3DRefer", - "scene_name": "003518", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.08, \"box_2d\": [0.0, 185.83, 353.24, 346.25], \"box_3d\": [1.37, 1.78, 4.65, -4.9, 1.56, 8.72], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 2.17, \"box_2d\": [1165.02, 150.45, 1241.0, 202.55], \"box_3d\": [1.48, 1.51, 4.35, 19.13, 0.83, 21.72], \"ry\": 2.89}]\n```" - } - ] - }, - { - "id": 832, - "data_source": "Mono3DRefer", - "scene_name": "003524", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [285.12, 194.74, 448.54, 304.45], \"box_3d\": [1.39, 1.48, 3.76, -3.75, 1.79, 11.63], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [514.95, 178.87, 558.59, 211.13], \"box_3d\": [1.46, 1.75, 4.23, -3.54, 1.77, 35.23], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 833, - "data_source": "Mono3DRefer", - "scene_name": "003553", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -�� (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.27, \"box_2d\": [0.0, 198.63, 57.49, 255.6], \"box_3d\": [1.67, 1.82, 4.42, -19.52, 2.53, 22.06], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 2.05, \"box_2d\": [186.84, 191.84, 273.15, 233.95], \"box_3d\": [1.63, 1.87, 3.47, -16.25, 2.49, 30.99], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [389.57, 187.13, 424.47, 204.87], \"box_3d\": [1.33, 1.72, 3.63, -16.27, 2.5, 57.92], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 834, - "data_source": "Mono3DRefer", - "scene_name": "003565", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.05, \"box_2d\": [791.58, 199.04, 1241.0, 374.0], \"box_3d\": [1.54, 1.6, 3.52, 2.39, 1.76, 4.56], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.24, \"box_2d\": [122.28, 194.05, 451.58, 374.0], \"box_3d\": [1.54, 1.5, 3.48, -2.55, 1.77, 6.48], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [356.34, 191.77, 509.0, 303.13], \"box_3d\": [1.44, 1.63, 3.32, -2.69, 1.78, 11.47], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.88, \"box_2d\": [711.61, 185.4, 882.04, 305.9], \"box_3d\": [1.55, 1.55, 3.27, 2.67, 1.76, 11.19], \"ry\": -1.65}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [670.16, 185.83, 770.13, 262.68], \"box_3d\": [1.48, 1.6, 3.79, 2.31, 1.8, 16.25], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [507.8, 184.64, 563.14, 230.56], \"box_3d\": [1.56, 1.55, 3.64, -2.74, 2.02, 26.86], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [539.23, 173.54, 585.23, 219.37], \"box_3d\": [1.91, 1.66, 4.22, -2.11, 1.97, 32.37], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [564.21, 176.37, 604.75, 213.01], \"box_3d\": [1.83, 1.69, 4.28, -1.33, 2.03, 38.38], \"ry\": 1.64}]\n```" - } - ] - }, - { - "id": 835, - "data_source": "Mono3DRefer", - "scene_name": "003569", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.19, \"box_2d\": [0.0, 145.42, 301.83, 374.0], \"box_3d\": [1.91, 1.66, 4.22, -3.18, 1.82, 3.57], \"ry\": 1.5}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [240.73, 167.59, 477.4, 349.02], \"box_3d\": [1.83, 1.69, 4.28, -2.99, 1.8, 9.48], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [516.83, 182.43, 594.35, 230.45], \"box_3d\": [1.51, 1.63, 4.03, -1.84, 1.86, 25.14], \"ry\": 1.78}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [569.04, 180.59, 626.93, 219.46], \"box_3d\": [1.57, 1.44, 3.6, -0.49, 1.92, 31.34], \"ry\": 1.87}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [620.27, 182.01, 668.66, 212.8], \"box_3d\": [1.47, 1.51, 3.5, 1.77, 1.95, 36.71], \"ry\": 1.91}, {\"category\": \"Car\", \"angle\": -1.25, \"box_2d\": [788.94, 165.74, 827.12, 187.24], \"box_3d\": [1.77, 1.75, 4.67, 17.03, 1.2, 61.96], \"ry\": -0.98}]\n```" - } - ] - }, - { - "id": 836, - "data_source": "Mono3DRefer", - "scene_name": "003573", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [819.48, 120.53, 858.01, 148.47], \"box_3d\": [1.5, 1.62, 4.19, 14.58, -1.65, 45.93], \"ry\": 2.05}]\n```" - } - ] - }, - { - "id": 837, - "data_source": "Mono3DRefer", - "scene_name": "003575", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.59, \"box_2d\": [237.07, 183.53, 457.92, 273.33], \"box_3d\": [1.41, 1.62, 3.6, -4.84, 1.64, 13.57], \"ry\": -0.93}, {\"category\": \"Car\", \"angle\": -0.54, \"box_2d\": [130.3, 168.19, 384.05, 268.22], \"box_3d\": [1.77, 1.96, 4.39, -7.24, 1.71, 15.16], \"ry\": -0.98}]\n```" - } - ] - }, - { - "id": 838, - "data_source": "Mono3DRefer", - "scene_name": "003577", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.23, \"box_2d\": [810.84, 163.69, 889.6, 204.4], \"box_3d\": [1.55, 1.47, 2.97, 9.63, 1.21, 28.98], \"ry\": 2.55}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [687.28, 173.35, 745.64, 218.25], \"box_3d\": [1.58, 1.68, 3.93, 3.95, 1.62, 27.56], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [531.53, 172.54, 583.56, 210.25], \"box_3d\": [1.58, 1.54, 3.98, -2.32, 1.59, 32.36], \"ry\": 1.7}, {\"category\": \"Car\", \"angle\": 3.01, \"box_2d\": [734.1, 162.5, 796.12, 189.99], \"box_3d\": [1.71, 1.67, 3.7, 9.96, 1.08, 46.64], \"ry\": -3.07}, {\"category\": \"Car\", \"angle\": 2.05, \"box_2d\": [99.33, 185.65, 496.14, 374.0], \"box_3d\": [1.54, 1.64, 4.08, -2.56, 1.69, 6.99], \"ry\": 1.71}, {\"category\": \"Car\", \"angle\": -1.95, \"box_2d\": [803.55, 175.69, 1185.25, 374.0], \"box_3d\": [1.58, 1.67, 3.43, 2.94, 1.63, 6.47], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -2.61, \"box_2d\": [1185.36, 138.75, 1241.0, 226.45], \"box_3d\": [1.84, 1.94, 3.68, 14.31, 1.11, 14.65], \"ry\": -1.85}, {\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [399.12, 179.48, 514.28, 256.28], \"box_3d\": [1.51, 1.61, 3.57, -3.32, 1.67, 16.06], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [460.35, 178.62, 548.89, 235.61], \"box_3d\": [1.5, 1.6, 4.17, -3.01, 1.68, 21.15], \"ry\": 1.67}, {\"category\": \"Car\", \"angle\": 2.06, \"box_2d\": [909.33, 160.17, 1009.71, 213.46], \"box_3d\": [1.71, 1.62, 3.56, 12.05, 1.31, 24.88], \"ry\": 2.51}, {\"category\": \"Car\", \"angle\": 2.16, \"box_2d\": [841.7, 164.6, 933.88, 208.92], \"box_3d\": [1.52, 1.58, 3.28, 10.12, 1.24, 26.33], \"ry\": 2.52}]\n```" - } - ] - }, - { - "id": 839, - "data_source": "Mono3DRefer", - "scene_name": "003578", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [474.39, 187.54, 520.84, 227.84], \"box_3d\": [1.57, 1.54, 4.01, -4.81, 2.23, 31.11], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [453.96, 187.34, 512.62, 243.1], \"box_3d\": [1.65, 1.75, 3.39, -4.14, 2.15, 23.74], \"ry\": 1.44}]\n```" - } - ] - }, - { - "id": 840, - "data_source": "Mono3DRefer", - "scene_name": "003584", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [412.35, 195.22, 525.22, 277.11], \"box_3d\": [1.46, 1.75, 4.23, -3.0, 2.01, 15.94], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 841, - "data_source": "Mono3DRefer", - "scene_name": "003585", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [531.68, 181.52, 554.68, 202.15], \"box_3d\": [1.47, 1.43, 3.11, -4.96, 2.12, 53.64], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 2.02, \"box_2d\": [231.31, 178.4, 287.19, 207.97], \"box_3d\": [1.71, 1.44, 4.06, -21.45, 2.07, 44.23], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.99, \"box_2d\": [276.94, 181.61, 325.0, 204.62], \"box_3d\": [1.53, 1.67, 3.88, -21.67, 2.16, 50.69], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 2.56, \"box_2d\": [988.38, 194.13, 1081.47, 227.29], \"box_3d\": [1.68, 1.39, 4.26, 22.95, 2.87, 39.07], \"ry\": 3.09}, {\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [317.97, 181.95, 357.21, 202.85], \"box_3d\": [1.53, 1.47, 4.05, -21.01, 2.25, 55.74], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [548.46, 179.73, 569.1, 196.37], \"box_3d\": [1.41, 1.48, 4.93, -4.53, 2.03, 64.08], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [564.93, 178.81, 617.28, 230.97], \"box_3d\": [1.62, 1.63, 4.5, -0.63, 1.84, 24.92], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.14, \"box_2d\": [163.51, 182.38, 236.92, 212.39], \"box_3d\": [1.46, 1.31, 4.15, -21.41, 1.97, 37.82], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": -0.69, \"box_2d\": [1077.9, 196.28, 1154.69, 226.85], \"box_3d\": [1.57, 1.8, 2.99, 27.91, 2.9, 39.86], \"ry\": -0.08}]\n```" - } - ] - }, - { - "id": 842, - "data_source": "Mono3DRefer", - "scene_name": "003610", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.6, \"box_2d\": [454.25, 186.21, 516.58, 245.31], \"box_3d\": [1.65, 1.75, 3.39, -3.85, 2.08, 22.43], \"ry\": 1.43}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [476.63, 188.27, 525.76, 230.58], \"box_3d\": [1.57, 1.54, 4.01, -4.46, 2.23, 29.79], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 843, - "data_source": "Mono3DRefer", - "scene_name": "003623", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [518.74, 200.28, 564.2, 241.29], \"box_3d\": [1.5, 1.62, 4.19, -2.9, 2.74, 30.86], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 844, - "data_source": "Mono3DRefer", - "scene_name": "003630", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [360.57, 185.95, 409.88, 212.85], \"box_3d\": [1.48, 1.71, 3.54, -13.26, 2.28, 42.67], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 845, - "data_source": "Mono3DRefer", - "scene_name": "003631", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.6, \"box_2d\": [237.55, 193.5, 272.32, 222.41], \"box_3d\": [1.45, 1.62, 3.83, -19.74, 2.64, 40.15], \"ry\": 1.14}]\n```" - } - ] - }, - { - "id": 846, - "data_source": "Mono3DRefer", - "scene_name": "003635", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.57, \"box_2d\": [583.65, 212.45, 614.38, 244.6], \"box_3d\": [1.5, 1.62, 4.19, -0.65, 3.82, 40.4], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 847, - "data_source": "Mono3DRefer", - "scene_name": "003671", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.19, \"box_2d\": [665.2, 178.52, 760.87, 211.04], \"box_3d\": [1.45, 1.6, 4.23, 4.79, 1.74, 34.06], \"ry\": -0.05}, {\"category\": \"Car\", \"angle\": 3.11, \"box_2d\": [650.04, 177.13, 702.57, 196.91], \"box_3d\": [1.48, 1.72, 3.99, 5.11, 1.83, 56.17], \"ry\": -3.09}]\n```" - } - ] - }, - { - "id": 848, - "data_source": "Mono3DRefer", - "scene_name": "003674", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.23, \"box_2d\": [853.14, 206.38, 1241.0, 374.0], \"box_3d\": [1.54, 1.6, 3.52, 2.41, 1.75, 2.98], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.14, \"box_2d\": [0.0, 197.09, 419.36, 374.0], \"box_3d\": [1.54, 1.5, 3.48, -2.52, 1.75, 4.9], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [311.94, 193.53, 498.09, 326.94], \"box_3d\": [1.44, 1.63, 3.32, -2.65, 1.77, 9.9], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.91, \"box_2d\": [730.41, 184.87, 939.29, 329.28], \"box_3d\": [1.55, 1.55, 3.27, 2.72, 1.73, 9.6], \"ry\": -1.64}, {\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [678.36, 186.99, 793.41, 273.88], \"box_3d\": [1.48, 1.6, 3.79, 2.36, 1.8, 14.64], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [503.5, 184.09, 562.59, 233.02], \"box_3d\": [1.56, 1.55, 3.64, -2.66, 1.97, 25.28], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [536.9, 171.95, 585.11, 220.29], \"box_3d\": [1.91, 1.66, 4.22, -2.05, 1.9, 30.8], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [564.61, 175.51, 606.97, 213.84], \"box_3d\": [1.83, 1.69, 4.28, -1.2, 1.98, 36.79], \"ry\": 1.64}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [613.94, 180.83, 649.84, 202.85], \"box_3d\": [1.51, 1.63, 4.03, 1.61, 2.1, 52.24], \"ry\": 1.85}]\n```" - } - ] - }, - { - "id": 849, - "data_source": "Mono3DRefer", - "scene_name": "003701", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.97, \"box_2d\": [76.8, 169.71, 416.69, 370.62], \"box_3d\": [1.66, 1.67, 4.49, -3.69, 1.67, 8.28], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.04, \"box_2d\": [145.73, 180.33, 220.66, 215.14], \"box_3d\": [1.56, 1.72, 3.09, -20.21, 1.92, 34.24], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [474.8, 177.86, 532.53, 218.99], \"box_3d\": [1.46, 1.59, 4.12, -4.07, 1.67, 27.94], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.17, \"box_2d\": [309.8, 176.0, 350.44, 198.64], \"box_3d\": [1.57, 1.57, 3.32, -20.28, 1.82, 52.34], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [337.03, 175.11, 377.6, 196.12], \"box_3d\": [1.57, 1.2, 5.24, -19.92, 1.77, 57.02], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [529.81, 173.16, 560.99, 201.39], \"box_3d\": [1.72, 1.66, 4.32, -4.15, 1.77, 46.63], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [584.48, 174.58, 622.53, 211.74], \"box_3d\": [1.62, 1.63, 4.5, -0.28, 1.73, 34.05], \"ry\": -1.55}]\n```" - } - ] - }, - { - "id": 850, - "data_source": "Mono3DRefer", - "scene_name": "003705", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [313.64, 193.37, 458.81, 292.73], \"box_3d\": [1.39, 1.48, 3.76, -3.76, 1.79, 12.59], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [517.38, 178.17, 559.19, 209.42], \"box_3d\": [1.46, 1.75, 4.23, -3.58, 1.75, 36.3], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 851, - "data_source": "Mono3DRefer", - "scene_name": "003710", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.82, \"box_2d\": [705.53, 185.69, 840.11, 270.76], \"box_3d\": [1.38, 1.51, 4.1, 2.97, 1.66, 14.18], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": -2.31, \"box_2d\": [975.46, 161.12, 1241.0, 374.0], \"box_3d\": [1.6, 1.66, 3.55, 3.76, 1.54, 4.01], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 2.28, \"box_2d\": [0.0, 188.03, 41.86, 230.24], \"box_3d\": [1.32, 1.51, 3.55, -20.08, 1.82, 22.71], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 852, - "data_source": "Mono3DRefer", - "scene_name": "003724", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [518.09, 176.84, 552.09, 207.12], \"box_3d\": [1.7, 1.72, 4.11, -4.44, 1.96, 43.02], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 853, - "data_source": "Mono3DRefer", - "scene_name": "003727", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [517.03, 180.88, 582.84, 223.44], \"box_3d\": [1.44, 1.68, 4.39, -2.18, 1.75, 26.88], \"ry\": 1.67}]\n```" - } - ] - }, - { - "id": 854, - "data_source": "Mono3DRefer", - "scene_name": "003733", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [468.63, 187.03, 543.37, 242.67], \"box_3d\": [1.42, 1.66, 4.11, -2.99, 1.87, 21.21], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 855, - "data_source": "Mono3DRefer", - "scene_name": "003738", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [280.63, 193.0, 490.63, 331.16], \"box_3d\": [1.46, 1.75, 4.23, -2.93, 1.79, 10.25], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 856, - "data_source": "Mono3DRefer", - "scene_name": "003745", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [266.8, 176.69, 324.76, 203.4], \"box_3d\": [1.57, 1.2, 5.24, -19.7, 1.83, 45.4], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [535.29, 178.15, 565.5, 200.8], \"box_3d\": [1.41, 1.69, 4.36, -3.92, 1.78, 47.66], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [462.46, 171.45, 529.9, 227.13], \"box_3d\": [1.72, 1.66, 4.32, -3.83, 1.71, 24.71], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [587.75, 174.28, 625.58, 211.96], \"box_3d\": [1.62, 1.63, 4.5, -0.16, 1.72, 33.63], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 2.1, \"box_2d\": [0.0, 194.12, 355.49, 374.0], \"box_3d\": [1.46, 1.59, 4.12, -3.88, 1.7, 6.5], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.08, \"box_2d\": [218.56, 178.12, 277.81, 207.84], \"box_3d\": [1.57, 1.57, 3.32, -20.15, 1.88, 40.24], \"ry\": -1.55}]\n```" - } - ] - }, - { - "id": 857, - "data_source": "Mono3DRefer", - "scene_name": "003752", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [539.11, 180.22, 596.33, 217.1], \"box_3d\": [1.44, 1.68, 4.39, -1.75, 1.76, 30.61], \"ry\": 1.69}]\n```" - } - ] - }, - { - "id": 858, - "data_source": "Mono3DRefer", - "scene_name": "003756", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 3.05, \"box_2d\": [986.66, 176.62, 1200.87, 246.26], \"box_3d\": [1.54, 1.68, 4.16, 11.53, 1.65, 17.63], \"ry\": -2.67}, {\"category\": \"Car\", \"angle\": -0.79, \"box_2d\": [702.69, 198.98, 790.69, 232.42], \"box_3d\": [1.37, 1.56, 4.19, 6.44, 2.66, 33.97], \"ry\": -0.61}]\n```" - } - ] - }, - { - "id": 859, - "data_source": "Mono3DRefer", - "scene_name": "003763", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.96, \"box_2d\": [0.0, 183.43, 322.7, 366.13], \"box_3d\": [1.57, 1.67, 4.14, -4.76, 1.61, 8.46], \"ry\": 1.46}]\n```" - } - ] - }, - { - "id": 860, - "data_source": "Mono3DRefer", - "scene_name": "003768", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.62, \"box_2d\": [450.88, 187.63, 501.99, 235.01], \"box_3d\": [1.65, 1.75, 3.39, -5.09, 2.24, 27.64], \"ry\": 1.44}]\n```" - } - ] - }, - { - "id": 861, - "data_source": "Mono3DRefer", - "scene_name": "003776", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [561.73, 174.24, 619.59, 230.15], \"box_3d\": [1.6, 1.66, 3.55, -0.7, 1.67, 22.66], \"ry\": -1.63}]\n```" - } - ] - }, - { - "id": 862, - "data_source": "Mono3DRefer", - "scene_name": "003789", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [451.48, 182.98, 514.64, 232.79], \"box_3d\": [1.7, 1.72, 4.11, -4.73, 2.1, 27.16], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 863, - "data_source": "Mono3DRefer", - "scene_name": "003794", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [292.24, 184.81, 366.45, 219.38], \"box_3d\": [1.55, 1.88, 4.59, -13.75, 2.16, 35.57], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 864, - "data_source": "Mono3DRefer", - "scene_name": "003803", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.59, \"box_2d\": [939.59, 134.64, 1241.0, 252.16], \"box_3d\": [1.47, 1.52, 4.43, 7.08, 1.0, 9.94], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [565.01, 164.78, 611.12, 203.62], \"box_3d\": [1.6, 1.68, 3.94, -0.94, 1.29, 31.91], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [618.2, 163.32, 640.86, 181.01], \"box_3d\": [1.42, 1.66, 4.11, 1.66, 0.68, 60.75], \"ry\": 1.66}, {\"category\": \"Car\", \"angle\": -1.37, \"box_2d\": [491.38, 165.26, 583.26, 233.84], \"box_3d\": [1.77, 1.75, 4.67, -2.02, 1.6, 21.06], \"ry\": -1.47}, {\"category\": \"Car\", \"angle\": -1.92, \"box_2d\": [0.0, 241.35, 52.27, 336.99], \"box_3d\": [1.72, 1.39, 3.07, -13.21, 3.23, 14.99], \"ry\": -2.64}, {\"category\": \"Car\", \"angle\": -0.87, \"box_2d\": [1220.47, 136.02, 1241.0, 281.11], \"box_3d\": [1.48, 1.63, 3.98, 8.38, 1.08, 6.76], \"ry\": 0.0}]\n```" - } - ] - }, - { - "id": 865, - "data_source": "Mono3DRefer", - "scene_name": "003817", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.54, \"box_2d\": [533.37, 184.15, 566.85, 218.3], \"box_3d\": [1.68, 1.63, 3.9, -3.25, 2.3, 38.29], \"ry\": 1.46}, {\"category\": \"Car\", \"angle\": 1.43, \"box_2d\": [502.87, 185.4, 530.63, 204.14], \"box_3d\": [1.47, 1.7, 4.39, -7.84, 2.53, 60.27], \"ry\": 1.3}]\n```" - } - ] - }, - { - "id": 866, - "data_source": "Mono3DRefer", - "scene_name": "003824", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [688.15, 172.02, 758.89, 237.73], \"box_3d\": [1.56, 1.65, 3.69, 2.9, 1.57, 19.05], \"ry\": -1.44}, {\"category\": \"Car\", \"angle\": 1.97, \"box_2d\": [381.35, 172.3, 511.07, 258.35], \"box_3d\": [1.76, 1.58, 3.68, -3.68, 1.77, 16.69], \"ry\": 1.76}, {\"category\": \"Car\", \"angle\": 2.33, \"box_2d\": [0.0, 188.76, 253.33, 298.64], \"box_3d\": [1.56, 1.62, 4.19, -8.53, 1.88, 12.84], \"ry\": 1.76}, {\"category\": \"Car\", \"angle\": 2.09, \"box_2d\": [395.33, 184.21, 503.23, 229.81], \"box_3d\": [1.42, 1.67, 4.53, -5.57, 1.85, 25.48], \"ry\": 1.88}, {\"category\": \"Car\", \"angle\": -0.7, \"box_2d\": [805.64, 165.8, 877.9, 193.9], \"box_3d\": [1.54, 1.63, 3.74, 13.2, 1.16, 41.18], \"ry\": -0.4}]\n```" - } - ] - }, - { - "id": 867, - "data_source": "Mono3DRefer", - "scene_name": "003825", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.62, \"box_2d\": [421.12, 180.3, 489.39, 239.51], \"box_3d\": [1.41, 1.59, 4.47, -4.17, 1.63, 19.76], \"ry\": 1.41}]\n```" - } - ] - }, - { - "id": 868, - "data_source": "Mono3DRefer", - "scene_name": "003832", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.08, \"box_2d\": [177.41, 174.29, 249.34, 213.43], \"box_3d\": [1.92, 1.58, 3.93, -20.69, 2.03, 37.75], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [584.98, 176.81, 624.06, 215.69], \"box_3d\": [1.62, 1.63, 4.5, -0.22, 1.82, 32.65], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [423.87, 181.29, 513.45, 248.37], \"box_3d\": [1.55, 1.51, 3.75, -3.6, 1.78, 18.73], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 869, - "data_source": "Mono3DRefer", - "scene_name": "003841", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.51, \"box_2d\": [529.2, 175.86, 551.0, 195.24], \"box_3d\": [1.66, 1.73, 3.05, -6.2, 1.95, 64.12], \"ry\": -1.61}]\n```" - } - ] - }, - { - "id": 870, - "data_source": "Mono3DRefer", - "scene_name": "003847", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.13, \"box_2d\": [0.0, 196.14, 339.3, 374.0], \"box_3d\": [1.45, 1.64, 4.21, -3.53, 1.67, 5.11], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.37, \"box_2d\": [682.32, 176.01, 795.56, 263.48], \"box_3d\": [1.62, 1.58, 3.89, 2.57, 1.7, 15.43], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [455.81, 182.98, 533.77, 237.23], \"box_3d\": [1.45, 1.67, 4.43, -3.43, 1.78, 21.95], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [655.08, 182.49, 727.37, 235.44], \"box_3d\": [1.47, 1.59, 3.62, 2.37, 1.78, 22.18], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [519.74, 181.75, 554.68, 211.95], \"box_3d\": [1.46, 1.51, 3.24, -3.72, 1.93, 37.0], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 871, - "data_source": "Mono3DRefer", - "scene_name": "003850", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.97, \"box_2d\": [782.25, 183.44, 1098.27, 353.48], \"box_3d\": [1.38, 1.51, 4.1, 3.21, 1.52, 8.07], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": 2.02, \"box_2d\": [192.89, 177.63, 254.34, 205.33], \"box_3d\": [1.41, 1.48, 3.63, -20.84, 1.69, 39.01], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.99, \"box_2d\": [230.73, 175.74, 283.81, 199.0], \"box_3d\": [1.32, 1.51, 3.55, -21.03, 1.51, 43.1], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 872, - "data_source": "Mono3DRefer", - "scene_name": "003852", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.08, \"box_2d\": [846.52, 200.62, 1241.0, 374.0], \"box_3d\": [1.51, 1.56, 3.81, 3.06, 1.77, 5.17], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.87, \"box_2d\": [739.45, 184.32, 948.6, 322.03], \"box_3d\": [1.57, 1.68, 3.83, 3.05, 1.76, 10.37], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [477.58, 176.15, 535.12, 229.59], \"box_3d\": [1.52, 1.5, 3.85, -3.21, 1.64, 22.65], \"ry\": 1.5}, {\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [663.92, 179.86, 715.32, 218.63], \"box_3d\": [1.43, 1.59, 3.68, 3.07, 1.71, 28.69], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [520.69, 179.58, 558.66, 207.83], \"box_3d\": [1.29, 1.59, 4.12, -3.44, 1.64, 35.56], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [541.39, 172.47, 570.47, 199.01], \"box_3d\": [1.52, 1.61, 3.59, -3.25, 1.52, 43.49], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.49, \"box_2d\": [643.29, 175.87, 676.12, 201.92], \"box_3d\": [1.51, 1.65, 4.28, 2.99, 1.72, 44.4], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 873, - "data_source": "Mono3DRefer", - "scene_name": "003859", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.95, \"box_2d\": [0.0, 195.15, 374.97, 373.0], \"box_3d\": [1.46, 1.69, 4.52, -2.73, 1.59, 3.54], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.1, \"box_2d\": [491.29, 168.65, 563.95, 201.26], \"box_3d\": [1.43, 1.66, 4.15, -3.39, 0.86, 33.72], \"ry\": -1.2}, {\"category\": \"Car\", \"angle\": -1.44, \"box_2d\": [565.3, 165.93, 600.61, 191.41], \"box_3d\": [1.45, 1.65, 3.52, -1.05, 0.56, 42.98], \"ry\": -1.47}, {\"category\": \"Car\", \"angle\": 0.33, \"box_2d\": [0.0, 179.46, 106.65, 231.41], \"box_3d\": [1.48, 1.56, 3.62, -17.26, 1.44, 21.57], \"ry\": -0.35}, {\"category\": \"Car\", \"angle\": 2.03, \"box_2d\": [356.97, 163.05, 418.32, 193.73], \"box_3d\": [1.68, 1.67, 4.29, -12.37, 0.66, 41.87], \"ry\": 1.74}]\n```" - } - ] - }, - { - "id": 874, - "data_source": "Mono3DRefer", - "scene_name": "003864", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [657.73, 179.18, 698.16, 209.45], \"box_3d\": [1.41, 1.65, 4.28, 3.32, 1.75, 36.26], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 0.18, \"box_2d\": [424.86, 178.94, 537.36, 215.04], \"box_3d\": [1.44, 1.8, 4.44, -5.43, 1.72, 30.61], \"ry\": 0.0}, {\"category\": \"Car\", \"angle\": 0.33, \"box_2d\": [358.88, 178.89, 506.01, 229.36], \"box_3d\": [1.53, 1.8, 4.22, -5.67, 1.74, 23.17], \"ry\": 0.09}, {\"category\": \"Car\", \"angle\": 0.36, \"box_2d\": [340.17, 181.55, 492.06, 236.66], \"box_3d\": [1.5, 1.46, 3.95, -5.56, 1.77, 20.75], \"ry\": 0.1}, {\"category\": \"Car\", \"angle\": 0.31, \"box_2d\": [297.96, 180.74, 460.97, 245.56], \"box_3d\": [1.53, 1.64, 3.6, -5.81, 1.75, 18.32], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": -2.68, \"box_2d\": [131.96, 182.92, 359.89, 274.33], \"box_3d\": [1.53, 1.51, 3.38, -6.55, 1.73, 13.11], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": -2.65, \"box_2d\": [25.55, 190.91, 344.38, 299.35], \"box_3d\": [1.41, 1.63, 3.75, -6.16, 1.71, 10.68], \"ry\": 3.12}, {\"category\": \"Car\", \"angle\": -2.36, \"box_2d\": [0.0, 197.32, 142.89, 374.0], \"box_3d\": [1.47, 1.53, 4.21, -6.51, 1.72, 5.92], \"ry\": 3.11}]\n```" - } - ] - }, - { - "id": 875, - "data_source": "Mono3DRefer", - "scene_name": "003867", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.46, \"box_2d\": [445.97, 175.04, 485.25, 202.84], \"box_3d\": [1.47, 1.7, 4.39, -8.23, 1.62, 40.75], \"ry\": 1.26}, {\"category\": \"Car\", \"angle\": 1.59, \"box_2d\": [450.32, 171.33, 499.41, 221.51], \"box_3d\": [1.68, 1.63, 3.9, -4.95, 1.66, 26.45], \"ry\": 1.4}]\n```" - } - ] - }, - { - "id": 876, - "data_source": "Mono3DRefer", - "scene_name": "003874", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.75, \"box_2d\": [687.29, 207.45, 760.99, 235.63], \"box_3d\": [1.37, 1.56, 4.19, 6.5, 3.42, 41.06], \"ry\": -0.6}]\n```" - } - ] - }, - { - "id": 877, - "data_source": "Mono3DRefer", - "scene_name": "003880", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.14, \"box_2d\": [0.0, 197.34, 400.19, 373.0], \"box_3d\": [1.57, 1.67, 4.14, -3.24, 1.76, 5.85], \"ry\": 1.65}]\n```" - } - ] - }, - { - "id": 878, - "data_source": "Mono3DRefer", - "scene_name": "003888", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [485.34, 178.46, 531.98, 216.58], \"box_3d\": [1.7, 1.72, 4.11, -4.82, 1.98, 34.59], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 879, - "data_source": "Mono3DRefer", - "scene_name": "003893", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [409.0, 198.93, 520.71, 281.38], \"box_3d\": [1.42, 1.66, 4.11, -3.02, 2.05, 15.59], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.6, \"box_2d\": [560.43, 174.92, 584.25, 199.3], \"box_3d\": [1.97, 1.87, 4.53, -3.17, 2.17, 61.2], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 880, - "data_source": "Mono3DRefer", - "scene_name": "003906", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [404.12, 182.33, 482.86, 229.78], \"box_3d\": [1.51, 1.81, 4.07, -5.82, 1.86, 25.51], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 881, - "data_source": "Mono3DRefer", - "scene_name": "003909", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [528.17, 175.92, 592.53, 224.77], \"box_3d\": [1.45, 1.64, 4.21, -1.58, 1.57, 23.75], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.5, \"box_2d\": [710.65, 177.05, 751.36, 214.15], \"box_3d\": [1.62, 1.58, 3.89, 5.55, 1.83, 33.61], \"ry\": 1.67}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [592.48, 176.85, 631.1, 204.25], \"box_3d\": [1.45, 1.67, 4.43, 0.13, 1.69, 40.78], \"ry\": 1.69}, {\"category\": \"Car\", \"angle\": -2.05, \"box_2d\": [907.55, 186.89, 1241.0, 374.0], \"box_3d\": [1.66, 1.56, 3.42, 2.93, 1.76, 4.01], \"ry\": -1.45}]\n```" - } - ] - }, - { - "id": 882, - "data_source": "Mono3DRefer", - "scene_name": "003913", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.54, \"box_2d\": [541.15, 186.73, 570.18, 216.14], \"box_3d\": [1.68, 1.63, 3.9, -3.41, 2.56, 44.5], \"ry\": 1.46}]\n```" - } - ] - }, - { - "id": 883, - "data_source": "Mono3DRefer", - "scene_name": "003932", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.31, \"box_2d\": [0.0, 201.34, 275.74, 374.0], \"box_3d\": [1.45, 1.64, 4.21, -3.46, 1.66, 3.49], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.37, \"box_2d\": [698.46, 176.37, 829.63, 277.07], \"box_3d\": [1.62, 1.58, 3.89, 2.7, 1.7, 13.65], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [447.76, 184.03, 534.29, 243.79], \"box_3d\": [1.45, 1.67, 4.43, -3.24, 1.78, 20.21], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [665.85, 182.0, 745.55, 239.85], \"box_3d\": [1.47, 1.59, 3.62, 2.57, 1.74, 20.43], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [522.03, 181.35, 558.42, 213.11], \"box_3d\": [1.46, 1.51, 3.24, -3.4, 1.88, 35.22], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 884, - "data_source": "Mono3DRefer", - "scene_name": "003933", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.72, \"box_2d\": [210.03, 186.43, 366.17, 242.54], \"box_3d\": [1.35, 1.6, 3.86, -8.85, 1.76, 20.08], \"ry\": -1.14}, {\"category\": \"Car\", \"angle\": -0.96, \"box_2d\": [449.48, 177.62, 537.04, 217.79], \"box_3d\": [1.49, 1.53, 3.8, -4.63, 1.69, 28.9], \"ry\": -1.11}, {\"category\": \"Car\", \"angle\": 2.54, \"box_2d\": [37.4, 185.09, 206.48, 240.89], \"box_3d\": [1.49, 1.34, 4.25, -14.78, 1.89, 22.07], \"ry\": 1.96}, {\"category\": \"Car\", \"angle\": 2.14, \"box_2d\": [398.01, 172.87, 474.28, 209.17], \"box_3d\": [1.72, 1.53, 4.55, -8.76, 1.75, 36.65], \"ry\": 1.91}, {\"category\": \"Car\", \"angle\": -1.15, \"box_2d\": [578.23, 174.38, 631.24, 202.7], \"box_3d\": [1.49, 1.56, 3.65, -0.27, 1.59, 39.99], \"ry\": -1.15}, {\"category\": \"Car\", \"angle\": -1.32, \"box_2d\": [704.49, 171.7, 729.91, 188.58], \"box_3d\": [1.44, 1.46, 3.16, 9.48, 1.36, 63.76], \"ry\": -1.17}]\n```" - } - ] - }, - { - "id": 885, - "data_source": "Mono3DRefer", - "scene_name": "003952", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [575.49, 177.67, 619.2, 219.81], \"box_3d\": [1.6, 1.66, 3.55, -0.53, 1.81, 29.43], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.1, \"box_2d\": [149.22, 178.77, 221.49, 206.19], \"box_3d\": [1.43, 1.73, 3.87, -23.47, 1.77, 39.99], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [259.93, 178.66, 307.91, 199.99], \"box_3d\": [1.5, 1.79, 4.5, -24.2, 1.95, 53.66], \"ry\": 1.5}, {\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [214.09, 178.65, 269.98, 203.37], \"box_3d\": [1.5, 1.62, 4.05, -23.58, 1.89, 46.33], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 886, - "data_source": "Mono3DRefer", - "scene_name": "003953", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.91, \"box_2d\": [380.05, 177.67, 470.78, 212.3], \"box_3d\": [1.49, 1.64, 3.7, -8.36, 1.73, 32.74], \"ry\": 3.13}]\n```" - } - ] - }, - { - "id": 887, - "data_source": "Mono3DRefer", - "scene_name": "003961", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [292.09, 170.14, 477.41, 288.37], \"box_3d\": [1.7, 1.82, 4.45, -3.72, 1.69, 12.68], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.41, \"box_2d\": [654.38, 170.09, 695.34, 204.08], \"box_3d\": [1.57, 1.56, 2.62, 3.14, 1.46, 34.9], \"ry\": -1.32}, {\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [619.09, 168.42, 653.68, 188.87], \"box_3d\": [1.43, 1.68, 4.19, 1.96, 1.14, 52.97], \"ry\": 1.82}]\n```" - } - ] - }, - { - "id": 888, - "data_source": "Mono3DRefer", - "scene_name": "003965", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [615.97, 165.51, 686.94, 230.15], \"box_3d\": [1.46, 1.53, 3.17, 0.92, 1.32, 18.06], \"ry\": -1.59}]\n```" - } - ] - }, - { - "id": 889, - "data_source": "Mono3DRefer", - "scene_name": "003966", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [426.51, 180.93, 493.87, 222.76], \"box_3d\": [1.51, 1.81, 4.07, -5.87, 1.85, 28.53], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 890, - "data_source": "Mono3DRefer", - "scene_name": "003990", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [586.97, 177.48, 673.77, 254.18], \"box_3d\": [1.37, 1.56, 3.54, 0.32, 1.48, 14.83], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.43, \"box_2d\": [481.17, 170.47, 564.97, 235.87], \"box_3d\": [1.61, 1.69, 4.26, -2.34, 1.58, 20.07], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [610.21, 170.56, 665.68, 222.45], \"box_3d\": [1.53, 1.63, 4.5, 0.9, 1.49, 23.73], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": -0.93, \"box_2d\": [1085.36, 135.69, 1241.0, 201.47], \"box_3d\": [1.55, 1.61, 3.64, 14.15, 0.67, 18.23], \"ry\": -0.27}]\n```" - } - ] - }, - { - "id": 891, - "data_source": "Mono3DRefer", - "scene_name": "004005", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.62, \"box_2d\": [546.66, 176.89, 566.39, 193.94], \"box_3d\": [1.45, 1.53, 3.98, -4.75, 1.83, 64.21], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [513.76, 176.53, 549.87, 207.66], \"box_3d\": [1.65, 1.56, 3.98, -4.37, 1.87, 40.52], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 892, - "data_source": "Mono3DRefer", - "scene_name": "004038", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.78, \"box_2d\": [937.2, 111.11, 1241.0, 374.0], \"box_3d\": [1.79, 1.72, 4.35, 3.67, 1.5, 3.84], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 2.17, \"box_2d\": [0.0, 214.26, 290.9, 374.0], \"box_3d\": [1.45, 1.49, 4.07, -3.93, 1.85, 5.14], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [323.57, 200.48, 459.44, 287.48], \"box_3d\": [1.27, 1.54, 3.58, -3.91, 1.84, 13.3], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [412.23, 181.12, 500.77, 249.07], \"box_3d\": [1.63, 1.65, 3.91, -4.08, 1.87, 19.53], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [606.21, 177.5, 641.44, 209.28], \"box_3d\": [1.46, 1.53, 3.17, 0.6, 1.7, 35.11], \"ry\": -1.61}]\n```" - } - ] - }, - { - "id": 893, - "data_source": "Mono3DRefer", - "scene_name": "004046", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.03, \"box_2d\": [220.94, 177.53, 276.67, 200.53], \"box_3d\": [1.43, 1.73, 3.87, -23.63, 1.75, 47.3], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [300.08, 177.07, 339.39, 195.75], \"box_3d\": [1.5, 1.79, 4.5, -24.48, 1.88, 60.96], \"ry\": 1.5}, {\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [267.31, 178.08, 311.77, 199.28], \"box_3d\": [1.5, 1.62, 4.05, -23.8, 1.9, 53.67], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [571.07, 178.28, 618.25, 223.78], \"box_3d\": [1.6, 1.66, 3.55, -0.59, 1.82, 27.4], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 894, - "data_source": "Mono3DRefer", - "scene_name": "004050", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.93, \"box_2d\": [406.33, 177.94, 486.72, 208.89], \"box_3d\": [1.49, 1.64, 3.7, -8.26, 1.77, 36.53], \"ry\": 3.13}]\n```" - } - ] - }, - { - "id": 895, - "data_source": "Mono3DRefer", - "scene_name": "004135", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.13, \"box_2d\": [182.71, 193.4, 367.25, 287.68], \"box_3d\": [1.56, 1.59, 3.69, -6.52, 2.01, 14.4], \"ry\": 1.71}, {\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [320.0, 176.11, 445.98, 258.25], \"box_3d\": [1.93, 1.59, 4.01, -5.88, 2.04, 19.08], \"ry\": 1.71}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [713.24, 180.33, 786.08, 235.57], \"box_3d\": [1.58, 1.67, 3.84, 4.28, 1.83, 22.8], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [659.18, 181.59, 689.9, 201.74], \"box_3d\": [1.53, 1.67, 3.78, 5.15, 2.24, 57.42], \"ry\": 1.87}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [719.93, 167.51, 763.52, 225.66], \"box_3d\": [2.05, 1.5, 3.83, 4.96, 1.89, 27.5], \"ry\": -1.4}, {\"category\": \"Car\", \"angle\": -1.48, \"box_2d\": [686.51, 181.6, 717.1, 206.96], \"box_3d\": [1.54, 1.63, 3.67, 5.92, 2.12, 46.42], \"ry\": -1.36}]\n```" - } - ] - }, - { - "id": 896, - "data_source": "Mono3DRefer", - "scene_name": "004151", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [552.87, 181.15, 576.9, 200.28], \"box_3d\": [1.52, 1.73, 4.28, -3.76, 2.23, 60.38], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 897, - "data_source": "Mono3DRefer", - "scene_name": "004158", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.08, \"box_2d\": [0.0, 190.3, 388.88, 374.0], \"box_3d\": [1.47, 1.6, 3.66, -2.96, 1.62, 5.1], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [481.73, 169.08, 553.51, 223.29], \"box_3d\": [1.7, 1.63, 4.08, -3.12, 1.61, 24.87], \"ry\": 1.65}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [693.3, 173.49, 811.02, 275.24], \"box_3d\": [1.59, 1.59, 2.47, 2.34, 1.62, 12.6], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [676.39, 182.07, 752.58, 245.55], \"box_3d\": [1.37, 1.59, 3.22, 2.39, 1.61, 17.43], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": -1.44, \"box_2d\": [521.85, 176.92, 569.67, 210.1], \"box_3d\": [1.4, 1.67, 3.72, -2.88, 1.6, 32.69], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": 1.54, \"box_2d\": [632.14, 173.35, 657.55, 196.01], \"box_3d\": [1.49, 1.61, 3.74, 2.35, 1.55, 49.71], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -2.77, \"box_2d\": [228.43, 168.31, 425.2, 234.71], \"box_3d\": [1.59, 1.63, 4.43, -7.24, 1.51, 18.61], \"ry\": -3.14}]\n```" - } - ] - }, - { - "id": 898, - "data_source": "Mono3DRefer", - "scene_name": "004159", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.52, \"box_2d\": [370.48, 194.0, 409.03, 228.63], \"box_3d\": [1.65, 1.75, 3.39, -11.61, 2.79, 37.81], \"ry\": 1.22}, {\"category\": \"Car\", \"angle\": 2.03, \"box_2d\": [0.0, 204.64, 392.1, 374.0], \"box_3d\": [1.47, 1.7, 4.39, -2.9, 1.76, 4.8], \"ry\": 1.51}]\n```" - } - ] - }, - { - "id": 899, - "data_source": "Mono3DRefer", - "scene_name": "004174", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [433.8, 178.02, 541.95, 258.83], \"box_3d\": [1.55, 1.68, 4.42, -2.62, 1.68, 16.2], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [671.22, 176.1, 738.45, 229.72], \"box_3d\": [1.6, 1.53, 4.18, 3.0, 1.72, 23.8], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [658.16, 174.6, 711.45, 216.13], \"box_3d\": [1.56, 1.57, 3.98, 2.95, 1.66, 29.47], \"ry\": -1.62}]\n```" - } - ] - }, - { - "id": 900, - "data_source": "Mono3DRefer", - "scene_name": "004176", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [513.56, 182.58, 552.82, 214.39], \"box_3d\": [1.45, 1.53, 3.98, -3.77, 1.95, 35.61], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [293.91, 183.15, 463.17, 299.47], \"box_3d\": [1.65, 1.56, 3.98, -3.79, 1.84, 12.38], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 901, - "data_source": "Mono3DRefer", - "scene_name": "004184", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.96, \"box_2d\": [434.24, 164.98, 530.09, 200.84], \"box_3d\": [1.74, 1.75, 4.57, -6.51, 1.38, 36.87], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": -2.85, \"box_2d\": [339.49, 165.49, 476.14, 229.98], \"box_3d\": [1.86, 1.69, 3.69, -6.16, 1.68, 22.11], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": -2.9, \"box_2d\": [376.78, 170.11, 498.54, 218.25], \"box_3d\": [1.68, 1.69, 4.07, -6.33, 1.61, 26.63], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [784.83, 150.27, 840.89, 189.51], \"box_3d\": [2.04, 1.99, 4.82, 11.2, 0.89, 40.26], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": -2.94, \"box_2d\": [412.89, 172.94, 514.1, 208.41], \"box_3d\": [1.48, 1.56, 4.12, -6.42, 1.51, 31.71], \"ry\": -3.14}]\n```" - } - ] - }, - { - "id": 902, - "data_source": "Mono3DRefer", - "scene_name": "004201", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [294.97, 192.86, 476.61, 305.16], \"box_3d\": [1.44, 1.72, 4.14, -3.49, 1.82, 11.92], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 903, - "data_source": "Mono3DRefer", - "scene_name": "004203", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [427.52, 188.08, 512.98, 269.27], \"box_3d\": [1.65, 1.75, 3.39, -3.21, 2.03, 16.9], \"ry\": 1.44}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [471.21, 190.36, 534.34, 243.35], \"box_3d\": [1.57, 1.54, 4.01, -3.57, 2.2, 24.35], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 904, - "data_source": "Mono3DRefer", - "scene_name": "004208", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.25, \"box_2d\": [0.0, 201.94, 275.46, 374.0], \"box_3d\": [1.7, 1.43, 3.2, -3.22, 1.9, 3.71], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.53, \"box_2d\": [521.11, 180.48, 547.5, 206.01], \"box_3d\": [1.68, 1.63, 3.9, -5.31, 2.22, 50.08], \"ry\": 1.42}]\n```" - } - ] - }, - { - "id": 905, - "data_source": "Mono3DRefer", - "scene_name": "004225", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.0, \"box_2d\": [0.0, 194.05, 442.7, 374.0], \"box_3d\": [1.47, 1.59, 3.96, -2.71, 1.69, 5.93], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.58, \"box_2d\": [580.36, 184.76, 678.78, 275.78], \"box_3d\": [1.57, 1.73, 4.15, 0.36, 1.84, 14.85], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 906, - "data_source": "Mono3DRefer", - "scene_name": "004230", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [218.41, 195.66, 455.59, 345.71], \"box_3d\": [1.45, 1.53, 3.98, -3.3, 1.8, 9.45], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 907, - "data_source": "Mono3DRefer", - "scene_name": "004231", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.34, \"box_2d\": [530.42, 186.1, 927.71, 329.15], \"box_3d\": [1.45, 1.68, 4.53, 1.32, 1.65, 8.92], \"ry\": -0.2}]\n```" - } - ] - }, - { - "id": 908, - "data_source": "Mono3DRefer", - "scene_name": "004238", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [257.77, 182.97, 493.02, 353.5], \"box_3d\": [1.63, 1.6, 3.73, -2.63, 1.77, 8.86], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -1.95, \"box_2d\": [759.16, 187.96, 1072.01, 374.0], \"box_3d\": [1.59, 1.49, 3.34, 2.56, 1.76, 6.98], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [681.51, 189.86, 823.88, 299.0], \"box_3d\": [1.54, 1.6, 3.52, 2.23, 1.86, 12.36], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.42, \"box_2d\": [409.61, 182.68, 515.55, 273.73], \"box_3d\": [1.54, 1.5, 3.48, -2.81, 1.75, 14.17], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [453.43, 185.87, 534.46, 246.61], \"box_3d\": [1.44, 1.63, 3.32, -3.06, 1.81, 19.32], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [660.65, 182.87, 745.62, 248.16], \"box_3d\": [1.55, 1.55, 3.27, 2.33, 1.83, 19.08], \"ry\": -1.66}, {\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [638.58, 182.83, 700.65, 232.15], \"box_3d\": [1.48, 1.6, 3.79, 1.87, 1.83, 24.03], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [519.14, 181.66, 560.96, 216.57], \"box_3d\": [1.56, 1.55, 3.64, -3.32, 1.99, 34.46], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [539.84, 172.8, 576.4, 209.33], \"box_3d\": [1.91, 1.66, 4.22, -2.86, 1.93, 40.09], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [556.33, 176.23, 589.86, 206.51], \"box_3d\": [1.83, 1.69, 4.28, -2.33, 2.06, 46.03], \"ry\": 1.62}]\n```" - } - ] - }, - { - "id": 909, - "data_source": "Mono3DRefer", - "scene_name": "004266", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [519.87, 183.06, 560.72, 223.19], \"box_3d\": [1.97, 1.87, 4.53, -3.67, 2.53, 38.37], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [554.26, 186.14, 575.67, 206.99], \"box_3d\": [1.7, 1.53, 3.82, -3.88, 2.88, 62.46], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 910, - "data_source": "Mono3DRefer", - "scene_name": "004294", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.38, \"box_2d\": [701.83, 189.81, 845.31, 288.01], \"box_3d\": [1.44, 1.64, 4.5, 2.77, 1.8, 13.4], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [478.56, 176.52, 540.77, 227.24], \"box_3d\": [1.46, 1.51, 3.24, -3.11, 1.59, 22.6], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.47, \"box_2d\": [671.08, 181.83, 725.28, 221.68], \"box_3d\": [1.38, 1.59, 4.19, 3.24, 1.73, 27.47], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [651.89, 176.39, 678.87, 200.49], \"box_3d\": [1.51, 1.56, 3.81, 3.59, 1.76, 47.53], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": 1.96, \"box_2d\": [128.39, 187.11, 470.85, 374.0], \"box_3d\": [1.45, 1.67, 4.43, -2.92, 1.64, 7.85], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": -1.93, \"box_2d\": [762.51, 194.43, 1050.65, 374.0], \"box_3d\": [1.47, 1.59, 3.62, 2.86, 1.75, 7.93], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [365.78, 175.93, 502.65, 269.03], \"box_3d\": [1.56, 1.68, 3.69, -3.29, 1.64, 14.04], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 911, - "data_source": "Mono3DRefer", - "scene_name": "004311", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.92, \"box_2d\": [774.65, 165.16, 1178.03, 374.0], \"box_3d\": [1.49, 1.64, 4.17, 2.7, 1.47, 6.57], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -2.41, \"box_2d\": [0.0, 179.65, 192.11, 288.12], \"box_3d\": [1.47, 1.58, 3.55, -8.44, 1.59, 10.66], \"ry\": -3.07}, {\"category\": \"Car\", \"angle\": -1.5, \"box_2d\": [494.14, 166.38, 566.79, 229.9], \"box_3d\": [1.6, 1.75, 3.89, -2.18, 1.46, 20.32], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [685.74, 167.42, 783.09, 235.94], \"box_3d\": [1.59, 1.64, 4.03, 3.1, 1.49, 18.96], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [676.2, 167.34, 740.21, 214.63], \"box_3d\": [1.5, 1.8, 4.5, 3.32, 1.36, 25.4], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": -1.45, \"box_2d\": [552.73, 168.59, 594.09, 201.55], \"box_3d\": [1.46, 1.56, 3.21, -1.71, 1.3, 33.96], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [662.14, 159.9, 704.46, 199.4], \"box_3d\": [1.83, 1.75, 4.52, 3.55, 1.25, 35.92], \"ry\": -1.54}]\n```" - } - ] - }, - { - "id": 912, - "data_source": "Mono3DRefer", - "scene_name": "004316", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.6, \"box_2d\": [538.5, 179.12, 585.35, 220.93], \"box_3d\": [1.41, 1.59, 4.47, -1.77, 1.66, 26.96], \"ry\": 1.53}]\n```" - } - ] - }, - { - "id": 913, - "data_source": "Mono3DRefer", - "scene_name": "004325", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.51, \"box_2d\": [0.0, 213.49, 187.08, 365.75], \"box_3d\": [1.42, 1.62, 3.5, -6.93, 1.93, 7.92], \"ry\": 3.07}, {\"category\": \"Car\", \"angle\": -2.85, \"box_2d\": [279.58, 190.55, 430.22, 246.29], \"box_3d\": [1.41, 1.54, 3.6, -6.97, 1.92, 19.83], \"ry\": 3.1}]\n```" - } - ] - }, - { - "id": 914, - "data_source": "Mono3DRefer", - "scene_name": "004342", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.86, 718.86), (cx, cy) = (607.19, 185.22). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [550.33, 176.78, 596.93, 211.85], \"box_3d\": [1.6, 1.76, 3.84, -1.75, 1.21, 35.15], \"ry\": -1.75}, {\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [504.57, 182.06, 538.4, 203.37], \"box_3d\": [1.39, 1.56, 3.45, -5.95, 1.18, 49.18], \"ry\": -1.91}]\n```" - } - ] - }, - { - "id": 915, - "data_source": "Mono3DRefer", - "scene_name": "004345", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [559.18, 183.63, 603.73, 214.44], \"box_3d\": [1.5, 1.7, 3.77, -1.47, 2.09, 37.89], \"ry\": 1.71}]\n```" - } - ] - }, - { - "id": 916, - "data_source": "Mono3DRefer", - "scene_name": "004349", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.06, \"box_2d\": [351.66, 187.77, 640.83, 316.96], \"box_3d\": [1.46, 1.53, 3.5, -1.42, 1.58, 9.56], \"ry\": -0.21}]\n```" - } - ] - }, - { - "id": 917, - "data_source": "Mono3DRefer", - "scene_name": "004352", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [652.73, 177.07, 677.69, 199.43], \"box_3d\": [1.51, 1.56, 3.81, 3.85, 1.82, 51.07], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": 1.05, \"box_2d\": [849.37, 184.13, 1241.0, 374.0], \"box_3d\": [1.62, 1.58, 3.89, 2.93, 1.71, 4.68], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [315.35, 185.05, 507.83, 302.63], \"box_3d\": [1.45, 1.67, 4.43, -2.89, 1.67, 11.37], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": -1.84, \"box_2d\": [723.58, 189.5, 891.99, 303.36], \"box_3d\": [1.47, 1.59, 3.62, 2.91, 1.76, 11.52], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [424.02, 178.64, 526.0, 251.01], \"box_3d\": [1.56, 1.68, 3.69, -3.19, 1.71, 17.53], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.42, \"box_2d\": [686.97, 185.58, 790.42, 258.81], \"box_3d\": [1.44, 1.64, 4.5, 2.82, 1.77, 16.95], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [503.47, 178.49, 554.5, 221.87], \"box_3d\": [1.46, 1.51, 3.24, -2.92, 1.68, 26.16], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.48, \"box_2d\": [668.88, 181.52, 715.45, 216.33], \"box_3d\": [1.38, 1.59, 4.19, 3.44, 1.76, 31.11], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 918, - "data_source": "Mono3DRefer", - "scene_name": "004354", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [555.36, 178.31, 578.49, 195.76], \"box_3d\": [1.45, 1.77, 4.04, -3.73, 1.94, 62.78], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 919, - "data_source": "Mono3DRefer", - "scene_name": "004355", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [616.8, 177.5, 685.37, 237.87], \"box_3d\": [1.37, 1.56, 3.54, 0.96, 1.51, 18.33], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.44, \"box_2d\": [529.82, 167.55, 597.42, 221.13], \"box_3d\": [1.61, 1.69, 4.26, -1.48, 1.48, 24.01], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [635.85, 167.11, 683.29, 211.42], \"box_3d\": [1.53, 1.63, 4.5, 1.82, 1.36, 27.42], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": -0.94, \"box_2d\": [1129.02, 135.24, 1241.0, 198.09], \"box_3d\": [1.48, 1.6, 3.84, 15.33, 0.59, 18.17], \"ry\": -0.24}, {\"category\": \"Car\", \"angle\": -0.84, \"box_2d\": [1044.14, 137.89, 1194.51, 193.27], \"box_3d\": [1.55, 1.61, 3.64, 15.01, 0.57, 21.36], \"ry\": -0.23}]\n```" - } - ] - }, - { - "id": 920, - "data_source": "Mono3DRefer", - "scene_name": "004356", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [348.02, 181.41, 494.37, 306.97], \"box_3d\": [1.65, 1.75, 3.39, -2.84, 1.79, 11.36], \"ry\": 1.46}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [446.87, 183.44, 532.45, 252.16], \"box_3d\": [1.57, 1.54, 4.01, -3.06, 1.86, 18.82], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 921, - "data_source": "Mono3DRefer", - "scene_name": "004357", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [522.72, 176.2, 554.99, 205.42], \"box_3d\": [1.7, 1.72, 4.11, -4.37, 1.93, 44.5], \"ry\": 1.53}]\n```" - } - ] - }, - { - "id": 922, - "data_source": "Mono3DRefer", - "scene_name": "004371", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.37, \"box_2d\": [714.3, 184.63, 851.33, 284.68], \"box_3d\": [1.51, 1.69, 4.04, 3.29, 1.75, 13.34], \"ry\": -1.14}, {\"category\": \"Car\", \"angle\": -1.89, \"box_2d\": [616.22, 184.59, 732.26, 254.93], \"box_3d\": [1.46, 1.58, 3.93, 1.37, 1.76, 17.4], \"ry\": -1.81}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [275.62, 184.17, 374.83, 250.27], \"box_3d\": [1.5, 1.68, 4.03, -7.34, 1.81, 18.85], \"ry\": 1.38}, {\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [568.08, 178.75, 641.39, 231.16], \"box_3d\": [1.6, 1.7, 3.56, -0.29, 1.81, 24.12], \"ry\": -1.8}, {\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [541.75, 180.94, 606.06, 222.38], \"box_3d\": [1.52, 1.73, 3.89, -1.56, 1.86, 28.95], \"ry\": -1.85}, {\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [523.6, 183.33, 567.71, 214.35], \"box_3d\": [1.37, 1.51, 3.49, -3.12, 1.87, 34.15], \"ry\": -1.83}]\n```" - } - ] - }, - { - "id": 923, - "data_source": "Mono3DRefer", - "scene_name": "004373", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [369.31, 181.38, 506.72, 269.29], \"box_3d\": [1.5, 1.7, 3.77, -3.3, 1.69, 14.39], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": -1.52, \"box_2d\": [525.9, 171.16, 553.8, 197.67], \"box_3d\": [1.57, 1.55, 3.32, -4.35, 1.49, 44.83], \"ry\": -1.62}]\n```" - } - ] - }, - { - "id": 924, - "data_source": "Mono3DRefer", - "scene_name": "004382", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [567.13, 183.02, 618.01, 232.61], \"box_3d\": [1.61, 1.66, 3.2, -0.63, 1.98, 25.44], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [376.72, 195.85, 445.51, 240.22], \"box_3d\": [1.4, 1.51, 3.7, -7.16, 2.28, 26.15], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [504.41, 188.57, 546.99, 220.58], \"box_3d\": [1.46, 1.66, 4.05, -4.22, 2.29, 36.38], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 925, - "data_source": "Mono3DRefer", - "scene_name": "004394", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [514.13, 182.3, 559.61, 214.37], \"box_3d\": [1.45, 1.77, 4.04, -3.55, 1.93, 35.34], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 926, - "data_source": "Mono3DRefer", - "scene_name": "004395", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [433.99, 201.16, 529.68, 274.15], \"box_3d\": [1.42, 1.66, 4.11, -3.01, 2.17, 17.46], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.6, \"box_2d\": [563.6, 178.05, 586.59, 201.68], \"box_3d\": [1.97, 1.87, 4.53, -3.03, 2.44, 63.07], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 927, - "data_source": "Mono3DRefer", - "scene_name": "004399", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [453.62, 189.21, 509.4, 242.1], \"box_3d\": [1.65, 1.75, 3.39, -4.43, 2.24, 25.05], \"ry\": 1.44}]\n```" - } - ] - }, - { - "id": 928, - "data_source": "Mono3DRefer", - "scene_name": "004432", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.56, \"box_2d\": [388.21, 187.93, 612.46, 284.43], \"box_3d\": [1.41, 1.62, 3.6, -1.9, 1.71, 12.77], \"ry\": -0.7}, {\"category\": \"Car\", \"angle\": -0.52, \"box_2d\": [294.55, 170.84, 533.92, 271.11], \"box_3d\": [1.77, 1.96, 4.39, -3.98, 1.75, 14.99], \"ry\": -0.78}, {\"category\": \"Car\", \"angle\": 1.41, \"box_2d\": [107.76, 189.87, 172.65, 226.0], \"box_3d\": [1.37, 1.6, 4.5, -19.87, 2.11, 30.37], \"ry\": 0.84}]\n```" - } - ] - }, - { - "id": 929, - "data_source": "Mono3DRefer", - "scene_name": "004443", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [270.92, 180.48, 437.5, 290.97], \"box_3d\": [1.57, 1.67, 4.14, -4.07, 1.56, 12.46], \"ry\": 1.51}]\n```" - } - ] - }, - { - "id": 930, - "data_source": "Mono3DRefer", - "scene_name": "004446", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [497.91, 183.9, 547.28, 222.52], \"box_3d\": [1.45, 1.53, 3.98, -3.58, 1.93, 29.82], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 2.06, \"box_2d\": [0.0, 186.33, 378.73, 374.0], \"box_3d\": [1.65, 1.56, 3.98, -3.73, 1.8, 6.96], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 931, - "data_source": "Mono3DRefer", - "scene_name": "004453", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [556.07, 179.69, 607.06, 212.24], \"box_3d\": [1.44, 1.68, 4.39, -1.31, 1.77, 34.34], \"ry\": 1.71}]\n```" - } - ] - }, - { - "id": 932, - "data_source": "Mono3DRefer", - "scene_name": "004457", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [536.91, 178.01, 566.42, 205.44], \"box_3d\": [1.7, 1.53, 3.82, -3.8, 2.06, 47.14], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [458.48, 170.55, 537.73, 239.7], \"box_3d\": [1.97, 1.87, 4.53, -3.49, 1.94, 23.05], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 933, - "data_source": "Mono3DRefer", - "scene_name": "004464", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [500.83, 174.96, 553.53, 224.44], \"box_3d\": [1.97, 1.87, 4.53, -3.54, 2.09, 31.31], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [548.51, 180.06, 573.01, 203.23], \"box_3d\": [1.7, 1.53, 3.82, -3.77, 2.27, 55.45], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 934, - "data_source": "Mono3DRefer", - "scene_name": "004466", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.3, \"box_2d\": [0.0, 203.55, 134.13, 356.05], \"box_3d\": [1.48, 1.61, 3.61, -7.04, 1.87, 7.56], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [611.29, 178.89, 634.99, 201.43], \"box_3d\": [1.54, 1.63, 3.67, 0.91, 1.99, 51.82], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [466.05, 182.8, 514.69, 211.97], \"box_3d\": [1.56, 1.59, 3.69, -6.78, 2.14, 41.07], \"ry\": 1.74}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [491.82, 174.86, 534.1, 206.91], \"box_3d\": [1.93, 1.59, 4.01, -6.13, 2.08, 45.81], \"ry\": 1.72}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [656.0, 175.32, 681.62, 199.52], \"box_3d\": [1.58, 1.67, 3.84, 3.99, 1.77, 49.58], \"ry\": -1.51}]\n```" - } - ] - }, - { - "id": 935, - "data_source": "Mono3DRefer", - "scene_name": "004475", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.86, 718.86), (cx, cy) = (607.19, 185.22). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [651.7, 168.28, 682.85, 194.29], \"box_3d\": [1.6, 1.76, 3.84, 3.8, 0.54, 46.6], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [586.44, 180.86, 656.14, 243.0], \"box_3d\": [1.39, 1.56, 3.45, 0.29, 1.29, 17.97], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [585.35, 176.49, 636.9, 220.53], \"box_3d\": [1.41, 1.66, 3.76, 0.04, 1.12, 25.18], \"ry\": -1.59}]\n```" - } - ] - }, - { - "id": 936, - "data_source": "Mono3DRefer", - "scene_name": "004497", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.46, \"box_2d\": [205.36, 190.23, 388.81, 242.34], \"box_3d\": [1.36, 1.68, 4.49, -8.83, 1.89, 20.45], \"ry\": 0.06}]\n```" - } - ] - }, - { - "id": 937, - "data_source": "Mono3DRefer", - "scene_name": "004523", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [493.39, 178.96, 536.26, 214.38], \"box_3d\": [1.7, 1.72, 4.11, -4.86, 2.03, 37.07], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 938, - "data_source": "Mono3DRefer", - "scene_name": "004535", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [548.6, 181.36, 574.6, 202.02], \"box_3d\": [1.52, 1.73, 4.28, -3.74, 2.2, 56.11], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 939, - "data_source": "Mono3DRefer", - "scene_name": "004540", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [365.28, 187.66, 520.17, 301.14], \"box_3d\": [1.55, 1.68, 4.42, -2.7, 1.84, 12.46], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [677.26, 182.17, 760.22, 247.22], \"box_3d\": [1.6, 1.53, 4.18, 2.87, 1.87, 20.09], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [660.23, 182.38, 722.89, 230.65], \"box_3d\": [1.56, 1.57, 3.98, 2.8, 1.92, 25.79], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": 1.42, \"box_2d\": [666.59, 174.51, 694.56, 196.51], \"box_3d\": [1.79, 1.72, 4.35, 5.96, 1.96, 61.43], \"ry\": 1.52}]\n```" - } - ] - }, - { - "id": 940, - "data_source": "Mono3DRefer", - "scene_name": "004550", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.14, \"box_2d\": [0.0, 201.81, 163.42, 271.82], \"box_3d\": [1.35, 1.5, 3.64, -12.39, 2.1, 17.03], \"ry\": 1.52}]\n```" - } - ] - }, - { - "id": 941, - "data_source": "Mono3DRefer", - "scene_name": "004554", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.83, \"box_2d\": [316.52, 174.35, 465.22, 244.07], \"box_3d\": [1.86, 1.69, 3.69, -6.18, 1.93, 20.49], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": -2.88, \"box_2d\": [360.82, 179.81, 490.98, 231.06], \"box_3d\": [1.68, 1.69, 4.07, -6.36, 1.94, 25.06], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [791.31, 160.92, 851.6, 201.86], \"box_3d\": [2.04, 1.99, 4.82, 11.21, 1.47, 38.67], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": -2.93, \"box_2d\": [401.52, 182.85, 508.54, 220.41], \"box_3d\": [1.48, 1.56, 4.12, -6.44, 1.92, 30.1], \"ry\": 3.14}, {\"category\": \"Car\", \"angle\": -2.95, \"box_2d\": [426.71, 176.34, 527.14, 213.87], \"box_3d\": [1.74, 1.75, 4.57, -6.48, 1.94, 35.27], \"ry\": -3.13}]\n```" - } - ] - }, - { - "id": 942, - "data_source": "Mono3DRefer", - "scene_name": "004559", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.53, \"box_2d\": [691.43, 182.13, 834.56, 311.59], \"box_3d\": [1.57, 1.73, 4.15, 2.15, 1.73, 11.02], \"ry\": 1.72}]\n```" - } - ] - }, - { - "id": 943, - "data_source": "Mono3DRefer", - "scene_name": "004566", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.57, \"box_2d\": [593.37, 197.6, 623.38, 227.37], \"box_3d\": [1.5, 1.62, 4.19, -0.12, 2.98, 41.33], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 944, - "data_source": "Mono3DRefer", - "scene_name": "004587", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [553.38, 185.59, 722.91, 345.88], \"box_3d\": [1.49, 1.6, 3.83, 0.22, 1.66, 8.8], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -0.91, \"box_2d\": [0.0, 181.49, 318.09, 374.0], \"box_3d\": [1.59, 1.65, 4.04, -3.31, 1.65, 3.98], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.38, \"box_2d\": [415.41, 179.57, 516.21, 244.37], \"box_3d\": [1.46, 1.73, 4.19, -3.61, 1.65, 18.57], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [694.09, 177.37, 755.45, 218.02], \"box_3d\": [1.42, 1.66, 4.18, 4.28, 1.62, 27.69], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 945, - "data_source": "Mono3DRefer", - "scene_name": "004592", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [581.75, 178.79, 607.51, 198.98], \"box_3d\": [1.42, 1.66, 4.11, -1.12, 1.87, 53.44], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": -0.11, \"box_2d\": [996.08, 123.5, 1241.0, 244.39], \"box_3d\": [1.66, 1.15, 3.39, 7.78, 1.0, 10.99], \"ry\": 0.5}, {\"category\": \"Car\", \"angle\": -1.3, \"box_2d\": [364.46, 177.98, 527.12, 291.17], \"box_3d\": [1.77, 1.75, 4.67, -2.9, 1.88, 13.66], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [506.41, 178.06, 572.52, 229.59], \"box_3d\": [1.6, 1.68, 3.94, -2.35, 1.79, 24.52], \"ry\": 1.62}]\n```" - } - ] - }, - { - "id": 946, - "data_source": "Mono3DRefer", - "scene_name": "004613", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [524.74, 180.94, 557.57, 207.87], \"box_3d\": [1.45, 1.53, 3.98, -3.94, 1.93, 41.46], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [400.06, 180.98, 501.21, 256.34], \"box_3d\": [1.65, 1.56, 3.98, -3.86, 1.86, 17.93], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 947, - "data_source": "Mono3DRefer", - "scene_name": "004618", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [382.2, 177.57, 492.51, 252.79], \"box_3d\": [1.5, 1.62, 4.19, -3.87, 1.63, 16.68], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 948, - "data_source": "Mono3DRefer", - "scene_name": "004636", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.04, \"box_2d\": [0.0, 187.01, 441.49, 374.0], \"box_3d\": [1.45, 1.67, 4.43, -2.92, 1.6, 6.12], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": -2.02, \"box_2d\": [796.11, 202.2, 1233.32, 374.0], \"box_3d\": [1.47, 1.59, 3.62, 2.84, 1.78, 6.11], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [321.46, 175.38, 487.47, 285.08], \"box_3d\": [1.56, 1.68, 3.69, -3.31, 1.62, 12.19], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.35, \"box_2d\": [711.76, 193.16, 887.55, 311.17], \"box_3d\": [1.44, 1.64, 4.5, 2.75, 1.82, 11.66], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [464.26, 174.97, 533.24, 230.42], \"box_3d\": [1.46, 1.51, 3.24, -3.17, 1.54, 20.81], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.46, \"box_2d\": [673.15, 180.64, 732.21, 223.35], \"box_3d\": [1.38, 1.59, 4.19, 3.17, 1.66, 25.65], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [504.23, 177.47, 556.51, 215.92], \"box_3d\": [1.38, 1.69, 3.74, -3.06, 1.57, 28.04], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [651.41, 175.54, 679.51, 200.62], \"box_3d\": [1.51, 1.56, 3.81, 3.46, 1.7, 45.76], \"ry\": -1.55}]\n```" - } - ] - }, - { - "id": 949, - "data_source": "Mono3DRefer", - "scene_name": "004651", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [646.59, 173.54, 787.36, 300.47], \"box_3d\": [1.46, 1.53, 3.17, 1.27, 1.49, 9.96], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -2.68, \"box_2d\": [248.19, 175.72, 393.66, 240.84], \"box_3d\": [1.59, 1.45, 3.15, -7.43, 1.68, 18.57], \"ry\": -3.06}]\n```" - } - ] - }, - { - "id": 950, - "data_source": "Mono3DRefer", - "scene_name": "004654", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.49, \"box_2d\": [693.15, 169.2, 722.86, 193.97], \"box_3d\": [1.54, 1.63, 3.67, 6.43, 1.34, 47.25], \"ry\": -1.36}, {\"category\": \"Car\", \"angle\": 2.31, \"box_2d\": [0.0, 150.57, 223.75, 355.59], \"box_3d\": [1.93, 1.59, 4.01, -6.47, 1.75, 8.24], \"ry\": 1.66}, {\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [744.85, 170.3, 906.16, 280.52], \"box_3d\": [1.58, 1.67, 3.84, 3.44, 1.57, 12.38], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [587.74, 167.55, 638.9, 202.17], \"box_3d\": [1.53, 1.67, 3.78, 0.18, 1.32, 33.97], \"ry\": 1.78}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [738.23, 149.19, 828.63, 247.6], \"box_3d\": [2.05, 1.5, 3.83, 3.94, 1.59, 17.06], \"ry\": -1.48}]\n```" - } - ] - }, - { - "id": 951, - "data_source": "Mono3DRefer", - "scene_name": "004655", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [646.34, 171.76, 709.47, 217.89], \"box_3d\": [1.41, 1.58, 4.36, 2.17, 1.41, 24.48], \"ry\": -1.6}]\n```" - } - ] - }, - { - "id": 952, - "data_source": "Mono3DRefer", - "scene_name": "004656", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [158.48, 196.01, 445.7, 366.46], \"box_3d\": [1.45, 1.77, 4.04, -3.33, 1.78, 8.62], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 953, - "data_source": "Mono3DRefer", - "scene_name": "004717", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [673.67, 181.72, 838.37, 325.21], \"box_3d\": [1.46, 1.53, 3.17, 1.59, 1.58, 9.01], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": -2.65, \"box_2d\": [247.28, 182.87, 399.57, 251.47], \"box_3d\": [1.59, 1.45, 3.15, -7.01, 1.85, 17.69], \"ry\": -3.02}]\n```" - } - ] - }, - { - "id": 954, - "data_source": "Mono3DRefer", - "scene_name": "004727", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [255.36, 178.9, 479.5, 356.53], \"box_3d\": [1.65, 1.75, 3.39, -2.64, 1.73, 8.53], \"ry\": 1.48}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [428.19, 181.65, 531.67, 263.45], \"box_3d\": [1.57, 1.54, 4.01, -2.79, 1.78, 16.04], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 955, - "data_source": "Mono3DRefer", - "scene_name": "004741", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.6, \"box_2d\": [448.33, 196.11, 520.26, 266.37], \"box_3d\": [1.65, 1.75, 3.39, -3.4, 2.33, 19.66], \"ry\": 1.43}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [479.14, 195.63, 534.21, 243.69], \"box_3d\": [1.57, 1.54, 4.01, -3.84, 2.48, 27.09], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 956, - "data_source": "Mono3DRefer", - "scene_name": "004751", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.99, \"box_2d\": [10.44, 184.64, 339.19, 364.34], \"box_3d\": [1.6, 1.66, 4.33, -4.85, 1.76, 8.8], \"ry\": 1.5}, {\"category\": \"Car\", \"angle\": 2.45, \"box_2d\": [970.41, 143.85, 1241.0, 320.97], \"box_3d\": [1.71, 1.67, 3.7, 6.2, 1.43, 7.93], \"ry\": 3.09}]\n```" - } - ] - }, - { - "id": 957, - "data_source": "Mono3DRefer", - "scene_name": "004803", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.15, \"box_2d\": [0.0, 158.49, 360.61, 374.0], \"box_3d\": [1.97, 1.87, 4.53, -3.33, 1.95, 4.6], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [314.86, 206.9, 477.51, 318.14], \"box_3d\": [1.47, 1.62, 3.79, -3.52, 2.14, 12.43], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [495.4, 194.63, 548.11, 243.38], \"box_3d\": [1.7, 1.53, 3.82, -3.45, 2.61, 28.47], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [534.94, 194.32, 567.46, 224.82], \"box_3d\": [1.5, 1.53, 3.38, -3.18, 2.71, 39.18], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [561.99, 192.96, 596.92, 215.6], \"box_3d\": [1.56, 1.56, 3.6, -2.26, 3.1, 53.92], \"ry\": 1.84}]\n```" - } - ] - }, - { - "id": 958, - "data_source": "Mono3DRefer", - "scene_name": "004805", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [512.54, 174.61, 553.89, 200.31], \"box_3d\": [1.42, 1.78, 3.62, -3.93, 1.04, 42.03], \"ry\": 1.65}]\n```" - } - ] - }, - { - "id": 959, - "data_source": "Mono3DRefer", - "scene_name": "004809", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [407.79, 182.47, 489.89, 218.71], \"box_3d\": [1.42, 1.67, 4.53, -6.88, 1.86, 31.15], \"ry\": 1.79}, {\"category\": \"Car\", \"angle\": -1.06, \"box_2d\": [693.65, 169.64, 746.99, 195.62], \"box_3d\": [1.54, 1.63, 3.74, 6.85, 1.36, 44.72], \"ry\": -0.91}, {\"category\": \"Car\", \"angle\": 2.35, \"box_2d\": [0.0, 189.5, 153.05, 260.81], \"box_3d\": [1.45, 1.5, 4.53, -13.46, 1.89, 17.38], \"ry\": 1.7}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [628.51, 174.33, 697.59, 239.23], \"box_3d\": [1.56, 1.65, 3.69, 1.4, 1.62, 19.3], \"ry\": -1.49}, {\"category\": \"Car\", \"angle\": -1.23, \"box_2d\": [869.97, 172.32, 1143.38, 329.83], \"box_3d\": [1.39, 1.61, 3.65, 4.58, 1.4, 8.19], \"ry\": -0.74}, {\"category\": \"Car\", \"angle\": 2.33, \"box_2d\": [0.0, 199.81, 221.51, 374.0], \"box_3d\": [1.45, 1.84, 4.59, -6.27, 1.79, 7.3], \"ry\": 1.64}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [508.06, 171.61, 564.95, 213.73], \"box_3d\": [1.76, 1.58, 3.68, -3.24, 1.73, 32.13], \"ry\": 1.74}, {\"category\": \"Car\", \"angle\": 2.19, \"box_2d\": [157.26, 184.0, 323.72, 254.77], \"box_3d\": [1.56, 1.62, 4.19, -9.25, 1.87, 18.41], \"ry\": 1.73}]\n```" - } - ] - }, - { - "id": 960, - "data_source": "Mono3DRefer", - "scene_name": "004820", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [305.31, 196.47, 462.08, 312.77], \"box_3d\": [1.5, 1.53, 3.38, -3.5, 1.93, 11.6], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [409.77, 189.0, 502.26, 257.82], \"box_3d\": [1.52, 1.54, 3.56, -3.83, 1.95, 18.27], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [481.28, 180.2, 554.86, 228.11], \"box_3d\": [1.56, 1.56, 3.6, -3.2, 1.83, 25.48], \"ry\": 1.75}]\n```" - } - ] - }, - { - "id": 961, - "data_source": "Mono3DRefer", - "scene_name": "004829", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [514.99, 181.35, 570.98, 219.34], \"box_3d\": [1.5, 1.7, 3.77, -2.83, 1.88, 30.86], \"ry\": 1.67}]\n```" - } - ] - }, - { - "id": 962, - "data_source": "Mono3DRefer", - "scene_name": "004831", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.29, \"box_2d\": [967.27, 180.47, 1110.12, 262.31], \"box_3d\": [1.47, 1.6, 3.66, 8.95, 1.65, 15.02], \"ry\": -0.76}, {\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [1149.95, 173.36, 1235.21, 219.39], \"box_3d\": [1.7, 1.63, 4.08, 23.19, 1.74, 28.71], \"ry\": 2.51}, {\"category\": \"Car\", \"angle\": -0.86, \"box_2d\": [486.08, 168.36, 817.42, 374.0], \"box_3d\": [1.6, 1.57, 3.23, 0.51, 1.58, 7.24], \"ry\": -0.79}, {\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [763.78, 173.41, 983.25, 307.38], \"box_3d\": [1.57, 1.5, 3.68, 3.83, 1.59, 10.29], \"ry\": 2.36}]\n```" - } - ] - }, - { - "id": 963, - "data_source": "Mono3DRefer", - "scene_name": "004834", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.42, \"box_2d\": [429.8, 188.09, 524.87, 270.47], \"box_3d\": [1.54, 1.5, 3.48, -2.82, 1.9, 15.7], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [465.92, 190.83, 539.97, 247.68], \"box_3d\": [1.44, 1.63, 3.32, -3.04, 2.0, 20.82], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [656.67, 186.96, 734.48, 247.66], \"box_3d\": [1.55, 1.55, 3.27, 2.32, 1.97, 20.63], \"ry\": -1.66}, {\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [637.17, 189.45, 695.55, 236.79], \"box_3d\": [1.48, 1.6, 3.79, 1.88, 2.1, 25.49], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [523.53, 187.67, 563.3, 221.68], \"box_3d\": [1.56, 1.55, 3.64, -3.3, 2.32, 35.96], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [541.15, 178.34, 576.23, 213.5], \"box_3d\": [1.91, 1.66, 4.22, -2.93, 2.24, 41.56], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [558.35, 182.46, 590.72, 211.96], \"box_3d\": [1.83, 1.69, 4.28, -2.31, 2.47, 47.52], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [317.91, 186.84, 505.98, 329.03], \"box_3d\": [1.63, 1.6, 3.73, -2.65, 1.85, 10.36], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -1.88, \"box_2d\": [736.03, 191.58, 968.07, 367.67], \"box_3d\": [1.59, 1.49, 3.34, 2.54, 1.85, 8.49], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [673.78, 193.44, 796.18, 290.16], \"box_3d\": [1.54, 1.6, 3.52, 2.21, 1.97, 13.85], \"ry\": -1.61}]\n```" - } - ] - }, - { - "id": 964, - "data_source": "Mono3DRefer", - "scene_name": "004848", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.96, \"box_2d\": [160.81, 192.5, 440.68, 358.23], \"box_3d\": [1.5, 1.7, 3.77, -3.45, 1.78, 8.76], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": -1.51, \"box_2d\": [502.66, 181.67, 535.49, 212.24], \"box_3d\": [1.57, 1.55, 3.32, -4.93, 2.06, 39.22], \"ry\": -1.63}]\n```" - } - ] - }, - { - "id": 965, - "data_source": "Mono3DRefer", - "scene_name": "004865", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [446.05, 180.23, 533.3, 255.1], \"box_3d\": [1.97, 1.87, 4.53, -3.48, 2.2, 21.46], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [533.22, 188.45, 564.01, 217.61], \"box_3d\": [1.7, 1.53, 3.82, -3.86, 2.72, 45.52], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 966, - "data_source": "Mono3DRefer", - "scene_name": "004875", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.97, \"box_2d\": [161.42, 167.07, 383.51, 239.0], \"box_3d\": [1.49, 1.67, 4.5, -7.87, 1.38, 16.93], \"ry\": 2.54}, {\"category\": \"Car\", \"angle\": -1.09, \"box_2d\": [967.05, 151.32, 1170.6, 253.18], \"box_3d\": [1.65, 1.67, 3.64, 8.46, 1.31, 13.32], \"ry\": -0.53}, {\"category\": \"Car\", \"angle\": 2.23, \"box_2d\": [1188.1, 153.77, 1241.0, 202.7], \"box_3d\": [1.59, 1.55, 3.93, 22.11, 0.97, 24.62], \"ry\": 2.95}]\n```" - } - ] - }, - { - "id": 967, - "data_source": "Mono3DRefer", - "scene_name": "004877", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.02, \"box_2d\": [1203.93, 189.08, 1241.0, 374.0], \"box_3d\": [1.49, 1.66, 4.16, 5.38, 1.55, 3.25], \"ry\": -0.03}, {\"category\": \"Car\", \"angle\": 2.57, \"box_2d\": [819.01, 144.17, 1241.0, 332.35], \"box_3d\": [1.85, 1.68, 4.34, 4.76, 1.58, 8.16], \"ry\": 3.08}, {\"category\": \"Car\", \"angle\": -2.57, \"box_2d\": [0.0, 197.24, 267.59, 316.2], \"box_3d\": [1.43, 1.56, 3.85, -7.05, 1.81, 9.9], \"ry\": 3.1}, {\"category\": \"Car\", \"angle\": -2.84, \"box_2d\": [198.58, 189.23, 376.55, 256.06], \"box_3d\": [1.42, 1.62, 3.5, -7.53, 1.84, 17.02], \"ry\": 3.03}, {\"category\": \"Car\", \"angle\": -2.96, \"box_2d\": [363.25, 184.05, 462.11, 221.33], \"box_3d\": [1.41, 1.54, 3.6, -7.91, 1.88, 29.01], \"ry\": 3.06}]\n```" - } - ] - }, - { - "id": 968, - "data_source": "Mono3DRefer", - "scene_name": "004878", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [304.39, 197.38, 488.18, 321.28], \"box_3d\": [1.42, 1.66, 4.11, -3.05, 1.85, 11.01], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.6, \"box_2d\": [553.51, 168.41, 579.51, 194.88], \"box_3d\": [1.97, 1.87, 4.53, -3.38, 1.67, 56.54], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 969, - "data_source": "Mono3DRefer", - "scene_name": "004899", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.22, \"box_2d\": [79.75, 182.19, 197.09, 229.98], \"box_3d\": [1.5, 1.47, 3.62, -16.14, 1.84, 24.84], \"ry\": 1.65}, {\"category\": \"Car\", \"angle\": 2.31, \"box_2d\": [0.0, 173.03, 89.55, 241.84], \"box_3d\": [1.78, 1.6, 4.29, -16.56, 1.8, 19.54], \"ry\": 1.62}]\n```" - } - ] - }, - { - "id": 970, - "data_source": "Mono3DRefer", - "scene_name": "004904", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [822.37, 152.47, 973.13, 266.61], \"box_3d\": [1.61, 1.56, 4.12, 4.65, 1.35, 12.34], \"ry\": -1.4}]\n```" - } - ] - }, - { - "id": 971, - "data_source": "Mono3DRefer", - "scene_name": "004908", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [294.81, 185.06, 435.2, 260.47], \"box_3d\": [1.51, 1.81, 4.07, -5.6, 1.82, 16.94], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 972, - "data_source": "Mono3DRefer", - "scene_name": "004920", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.65, \"box_2d\": [843.51, 165.11, 1157.85, 279.86], \"box_3d\": [1.76, 1.71, 4.4, 6.41, 1.67, 12.24], \"ry\": 3.12}, {\"category\": \"Car\", \"angle\": 0.65, \"box_2d\": [0.0, 120.76, 250.68, 330.14], \"box_3d\": [2.22, 1.68, 4.08, -6.77, 1.69, 8.55], \"ry\": -0.0}, {\"category\": \"Car\", \"angle\": -2.6, \"box_2d\": [16.62, 184.83, 290.86, 286.71], \"box_3d\": [1.4, 1.46, 3.23, -6.84, 1.6, 10.95], \"ry\": 3.13}, {\"category\": \"Car\", \"angle\": -2.8, \"box_2d\": [246.65, 178.72, 441.92, 241.52], \"box_3d\": [1.46, 1.61, 4.31, -6.61, 1.63, 18.12], \"ry\": 3.14}, {\"category\": \"Car\", \"angle\": -3.0, \"box_2d\": [341.58, 167.9, 490.62, 228.11], \"box_3d\": [1.8, 1.77, 4.41, -6.15, 1.68, 23.17], \"ry\": 3.02}, {\"category\": \"Car\", \"angle\": -2.93, \"box_2d\": [439.48, 174.32, 507.59, 205.24], \"box_3d\": [1.58, 1.54, 3.3, -7.25, 1.68, 38.37], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": 0.16, \"box_2d\": [456.71, 174.35, 533.23, 205.26], \"box_3d\": [1.68, 1.71, 4.07, -6.53, 1.8, 41.08], \"ry\": 0.0}]\n```" - } - ] - }, - { - "id": 973, - "data_source": "Mono3DRefer", - "scene_name": "004936", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [498.0, 183.15, 549.86, 220.28], \"box_3d\": [1.44, 1.72, 4.14, -3.63, 1.9, 30.78], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 974, - "data_source": "Mono3DRefer", - "scene_name": "004951", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [394.14, 179.76, 488.61, 249.48], \"box_3d\": [1.7, 1.72, 4.11, -4.55, 1.9, 19.86], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 975, - "data_source": "Mono3DRefer", - "scene_name": "004961", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.08, \"box_2d\": [0.0, 147.24, 370.95, 374.0], \"box_3d\": [1.83, 1.69, 4.28, -3.01, 1.77, 4.52], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [487.98, 180.61, 585.26, 241.76], \"box_3d\": [1.51, 1.63, 4.03, -1.96, 1.74, 20.05], \"ry\": 1.75}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [556.32, 176.88, 625.48, 223.6], \"box_3d\": [1.57, 1.44, 3.6, -0.66, 1.73, 26.21], \"ry\": 1.86}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [616.77, 176.46, 671.85, 211.93], \"box_3d\": [1.47, 1.51, 3.5, 1.53, 1.64, 31.71], \"ry\": 1.89}, {\"category\": \"Car\", \"angle\": -1.27, \"box_2d\": [801.09, 163.04, 841.61, 186.59], \"box_3d\": [1.77, 1.75, 4.67, 16.67, 1.05, 56.81], \"ry\": -0.99}]\n```" - } - ] - }, - { - "id": 976, - "data_source": "Mono3DRefer", - "scene_name": "004969", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.61, \"box_2d\": [780.02, 174.72, 834.26, 222.35], \"box_3d\": [1.56, 1.65, 3.69, 6.88, 1.65, 25.58], \"ry\": -1.35}]\n```" - } - ] - }, - { - "id": 977, - "data_source": "Mono3DRefer", - "scene_name": "004982", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [579.82, 176.46, 630.27, 224.47], \"box_3d\": [1.54, 1.63, 3.67, -0.21, 1.69, 25.31], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [598.1, 170.54, 637.56, 211.7], \"box_3d\": [1.77, 1.69, 4.32, 0.31, 1.7, 33.48], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -2.14, \"box_2d\": [867.91, 128.79, 1241.0, 374.0], \"box_3d\": [1.74, 1.62, 4.13, 2.88, 1.61, 3.95], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.83, \"box_2d\": [735.3, 175.17, 881.73, 282.78], \"box_3d\": [1.53, 1.46, 3.26, 3.09, 1.58, 11.97], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [667.72, 172.88, 717.37, 216.07], \"box_3d\": [1.6, 1.56, 3.45, 3.19, 1.63, 28.69], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [651.43, 175.43, 694.0, 209.47], \"box_3d\": [1.52, 1.53, 4.07, 2.92, 1.66, 34.57], \"ry\": -1.61}]\n```" - } - ] - }, - { - "id": 978, - "data_source": "Mono3DRefer", - "scene_name": "004994", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [357.64, 191.12, 475.74, 274.64], \"box_3d\": [1.39, 1.48, 3.76, -3.78, 1.79, 14.51], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [521.82, 176.9, 560.3, 206.31], \"box_3d\": [1.46, 1.75, 4.23, -3.65, 1.7, 38.45], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 979, - "data_source": "Mono3DRefer", - "scene_name": "005001", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [250.36, 196.31, 436.69, 318.75], \"box_3d\": [1.39, 1.48, 3.76, -3.74, 1.78, 10.66], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [512.36, 179.61, 557.96, 212.96], \"box_3d\": [1.46, 1.75, 4.23, -3.51, 1.79, 34.16], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 980, - "data_source": "Mono3DRefer", - "scene_name": "005014", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [185.62, 185.17, 390.16, 307.82], \"box_3d\": [1.62, 1.66, 3.81, -5.0, 1.84, 11.7], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [434.58, 181.4, 502.7, 223.58], \"box_3d\": [1.46, 1.67, 4.92, -5.39, 1.81, 27.89], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [639.43, 175.67, 668.21, 199.31], \"box_3d\": [1.51, 1.84, 4.1, 2.93, 1.72, 48.64], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": -1.87, \"box_2d\": [740.21, 171.71, 1005.25, 354.03], \"box_3d\": [1.61, 1.67, 3.62, 2.63, 1.63, 8.25], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 981, - "data_source": "Mono3DRefer", - "scene_name": "005020", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.53, \"box_2d\": [1139.73, 185.57, 1193.05, 225.06], \"box_3d\": [1.43, 1.55, 3.91, 22.35, 1.97, 28.97], \"ry\": -0.88}]\n```" - } - ] - }, - { - "id": 982, - "data_source": "Mono3DRefer", - "scene_name": "005033", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.27, \"box_2d\": [0.0, 142.03, 297.72, 374.0], \"box_3d\": [1.83, 1.69, 4.28, -3.01, 1.73, 2.94], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [472.08, 183.53, 579.3, 251.74], \"box_3d\": [1.51, 1.63, 4.03, -2.05, 1.8, 18.36], \"ry\": 1.74}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [549.52, 180.14, 623.73, 230.69], \"box_3d\": [1.57, 1.44, 3.6, -0.75, 1.83, 24.51], \"ry\": 1.86}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [614.95, 180.95, 672.46, 218.88], \"box_3d\": [1.47, 1.51, 3.5, 1.43, 1.82, 30.04], \"ry\": 1.88}, {\"category\": \"Car\", \"angle\": -1.28, \"box_2d\": [803.7, 164.31, 845.08, 188.55], \"box_3d\": [1.77, 1.75, 4.67, 16.43, 1.16, 55.19], \"ry\": -0.99}]\n```" - } - ] - }, - { - "id": 983, - "data_source": "Mono3DRefer", - "scene_name": "005034", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.84, \"box_2d\": [740.46, 187.83, 927.98, 316.93], \"box_3d\": [1.51, 1.56, 3.81, 3.02, 1.76, 10.64], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [695.34, 180.68, 810.36, 262.46], \"box_3d\": [1.57, 1.68, 3.83, 2.97, 1.76, 15.98], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [503.3, 181.2, 546.87, 223.62], \"box_3d\": [1.52, 1.5, 3.85, -3.28, 1.85, 28.06], \"ry\": 1.5}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [651.37, 180.42, 693.31, 212.63], \"box_3d\": [1.43, 1.59, 3.68, 2.87, 1.79, 34.17], \"ry\": -1.59}]\n```" - } - ] - }, - { - "id": 984, - "data_source": "Mono3DRefer", - "scene_name": "005045", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.08, \"box_2d\": [800.7, 128.42, 1197.51, 358.58], \"box_3d\": [1.79, 1.72, 4.35, 3.62, 1.47, 7.86], \"ry\": 1.5}, {\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [125.57, 201.18, 395.6, 366.68], \"box_3d\": [1.45, 1.49, 4.07, -4.0, 1.87, 8.95], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [390.42, 197.36, 485.38, 261.47], \"box_3d\": [1.27, 1.54, 3.58, -4.03, 1.9, 17.23], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [443.88, 181.5, 512.77, 237.01], \"box_3d\": [1.63, 1.65, 3.91, -4.23, 1.93, 23.49], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [600.44, 180.05, 632.27, 208.41], \"box_3d\": [1.46, 1.53, 3.17, 0.28, 1.86, 39.17], \"ry\": -1.62}]\n```" - } - ] - }, - { - "id": 985, - "data_source": "Mono3DRefer", - "scene_name": "005047", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [547.16, 178.88, 574.03, 198.98], \"box_3d\": [1.45, 1.77, 4.04, -3.73, 1.92, 54.81], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 986, - "data_source": "Mono3DRefer", - "scene_name": "005051", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [114.51, 190.47, 376.51, 372.31], \"box_3d\": [1.6, 1.73, 3.62, -4.02, 1.85, 8.54], \"ry\": 1.42}, {\"category\": \"Car\", \"angle\": 2.95, \"box_2d\": [593.5, 171.69, 684.0, 204.79], \"box_3d\": [1.73, 1.74, 4.71, 1.53, 1.69, 39.72], \"ry\": 2.99}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [441.19, 181.58, 475.69, 206.63], \"box_3d\": [1.58, 1.76, 4.18, -10.09, 2.17, 48.18], \"ry\": 1.48}]\n```" - } - ] - }, - { - "id": 987, - "data_source": "Mono3DRefer", - "scene_name": "005052", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [686.29, 180.94, 794.57, 263.1], \"box_3d\": [1.6, 1.53, 4.18, 2.75, 1.79, 16.29], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [662.25, 179.78, 738.1, 236.9], \"box_3d\": [1.56, 1.57, 3.98, 2.61, 1.79, 22.0], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": 1.42, \"box_2d\": [663.31, 170.67, 693.26, 194.1], \"box_3d\": [1.79, 1.72, 4.35, 5.41, 1.65, 57.78], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [222.5, 186.88, 486.05, 364.82], \"box_3d\": [1.55, 1.68, 4.42, -2.74, 1.75, 8.73], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 988, - "data_source": "Mono3DRefer", - "scene_name": "005083", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [542.66, 180.83, 622.21, 237.88], \"box_3d\": [1.6, 1.7, 3.56, -0.98, 1.86, 22.35], \"ry\": -1.83}, {\"category\": \"Car\", \"angle\": -1.8, \"box_2d\": [516.18, 182.99, 585.35, 227.37], \"box_3d\": [1.52, 1.73, 3.89, -2.35, 1.92, 27.22], \"ry\": -1.89}, {\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [304.79, 190.21, 355.92, 233.76], \"box_3d\": [1.44, 1.6, 3.79, -10.41, 2.12, 26.95], \"ry\": 1.25}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [500.27, 185.97, 545.76, 219.23], \"box_3d\": [1.37, 1.51, 3.49, -3.97, 1.97, 32.26], \"ry\": -1.84}, {\"category\": \"Car\", \"angle\": -1.41, \"box_2d\": [713.75, 182.49, 862.61, 299.38], \"box_3d\": [1.51, 1.69, 4.04, 2.99, 1.69, 11.66], \"ry\": -1.17}, {\"category\": \"Car\", \"angle\": -1.89, \"box_2d\": [603.77, 186.12, 734.17, 266.12], \"box_3d\": [1.46, 1.58, 3.93, 1.08, 1.77, 15.63], \"ry\": -1.83}, {\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [210.07, 190.66, 330.62, 267.6], \"box_3d\": [1.5, 1.68, 4.03, -7.81, 1.95, 16.86], \"ry\": 1.36}]\n```" - } - ] - }, - { - "id": 989, - "data_source": "Mono3DRefer", - "scene_name": "005100", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [506.6, 184.1, 559.71, 223.71], \"box_3d\": [1.42, 1.66, 4.11, -3.01, 1.89, 28.63], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 990, - "data_source": "Mono3DRefer", - "scene_name": "005116", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [601.84, 185.02, 676.87, 246.95], \"box_3d\": [1.36, 1.69, 3.38, 0.62, 1.69, 18.05], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.39, \"box_2d\": [1108.54, 177.98, 1241.0, 226.89], \"box_3d\": [1.45, 1.67, 4.32, 18.48, 1.63, 22.9], \"ry\": 3.06}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [351.31, 186.89, 410.65, 221.36], \"box_3d\": [1.59, 1.72, 3.86, -11.47, 2.32, 36.29], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -2.35, \"box_2d\": [480.51, 182.6, 552.76, 214.87], \"box_3d\": [1.59, 1.63, 3.64, -4.92, 2.12, 37.6], \"ry\": -2.48}, {\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [583.32, 185.34, 627.46, 226.28], \"box_3d\": [1.52, 1.67, 3.61, -0.17, 2.04, 29.26], \"ry\": -1.55}]\n```" - } - ] - }, - { - "id": 991, - "data_source": "Mono3DRefer", - "scene_name": "005127", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.47, \"box_2d\": [933.5, 180.67, 1110.38, 317.29], \"box_3d\": [1.33, 1.59, 3.48, 5.24, 1.45, 9.04], \"ry\": -0.96}, {\"category\": \"Car\", \"angle\": -1.31, \"box_2d\": [878.58, 170.85, 966.35, 219.11], \"box_3d\": [1.47, 1.6, 4.41, 10.51, 1.42, 24.17], \"ry\": -0.9}, {\"category\": \"Car\", \"angle\": -1.42, \"box_2d\": [1041.91, 160.69, 1145.89, 230.81], \"box_3d\": [1.63, 1.69, 3.88, 12.59, 1.36, 18.72], \"ry\": -0.84}, {\"category\": \"Car\", \"angle\": -1.26, \"box_2d\": [922.95, 169.44, 1005.88, 210.27], \"box_3d\": [1.46, 1.68, 4.1, 13.69, 1.35, 27.83], \"ry\": -0.8}]\n```" - } - ] - }, - { - "id": 992, - "data_source": "Mono3DRefer", - "scene_name": "005128", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.12, \"box_2d\": [0.0, 183.38, 367.74, 374.0], \"box_3d\": [1.48, 1.64, 3.95, -3.32, 1.57, 5.27], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -2.02, \"box_2d\": [805.09, 199.02, 1241.0, 374.0], \"box_3d\": [1.35, 1.64, 3.84, 2.72, 1.6, 5.29], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.39, \"box_2d\": [685.45, 189.38, 796.62, 267.86], \"box_3d\": [1.42, 1.59, 4.19, 2.66, 1.81, 15.77], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [409.23, 187.34, 507.93, 260.39], \"box_3d\": [1.42, 1.64, 3.63, -3.35, 1.77, 16.36], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [667.33, 184.89, 736.59, 236.16], \"box_3d\": [1.38, 1.58, 3.66, 2.65, 1.76, 21.76], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.45, \"box_2d\": [655.92, 182.04, 712.56, 224.43], \"box_3d\": [1.42, 1.6, 3.78, 2.61, 1.77, 26.46], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 993, - "data_source": "Mono3DRefer", - "scene_name": "005130", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [406.82, 190.15, 517.67, 264.24], \"box_3d\": [1.44, 1.7, 3.65, -3.29, 1.87, 16.51], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 994, - "data_source": "Mono3DRefer", - "scene_name": "005132", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [706.61, 172.69, 777.69, 238.48], \"box_3d\": [1.56, 1.65, 3.69, 3.36, 1.58, 19.03], \"ry\": -1.43}, {\"category\": \"Car\", \"angle\": 2.05, \"box_2d\": [292.27, 173.54, 475.95, 289.2], \"box_3d\": [1.76, 1.58, 3.68, -3.88, 1.79, 12.92], \"ry\": 1.76}, {\"category\": \"Car\", \"angle\": 2.38, \"box_2d\": [0.0, 191.21, 227.34, 317.4], \"box_3d\": [1.56, 1.62, 4.19, -8.36, 1.9, 11.51], \"ry\": 1.76}, {\"category\": \"Car\", \"angle\": 2.07, \"box_2d\": [395.39, 185.27, 507.21, 233.84], \"box_3d\": [1.42, 1.67, 4.53, -5.2, 1.87, 24.16], \"ry\": 1.86}, {\"category\": \"Car\", \"angle\": -0.64, \"box_2d\": [844.89, 163.73, 922.27, 192.84], \"box_3d\": [1.54, 1.63, 3.74, 15.04, 1.06, 39.78], \"ry\": -0.29}]\n```" - } - ] - }, - { - "id": 995, - "data_source": "Mono3DRefer", - "scene_name": "005133", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [599.03, 175.73, 626.41, 202.65], \"box_3d\": [1.65, 1.67, 3.64, 0.12, 1.86, 46.49], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.01, \"box_2d\": [0.0, 184.83, 394.18, 374.0], \"box_3d\": [1.61, 1.51, 4.5, -2.69, 1.7, 4.01], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 996, - "data_source": "Mono3DRefer", - "scene_name": "005169", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [444.32, 179.7, 502.89, 217.18], \"box_3d\": [1.51, 1.81, 4.07, -5.92, 1.83, 31.56], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 997, - "data_source": "Mono3DRefer", - "scene_name": "005170", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [245.37, 116.63, 516.15, 316.46], \"box_3d\": [2.11, 2.04, 4.62, -2.67, 1.43, 9.96], \"ry\": 1.64}]\n```" - } - ] - }, - { - "id": 998, - "data_source": "Mono3DRefer", - "scene_name": "005177", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [740.74, 168.37, 814.53, 234.01], \"box_3d\": [1.56, 1.65, 3.69, 4.3, 1.48, 19.08], \"ry\": -1.41}, {\"category\": \"Car\", \"angle\": 2.41, \"box_2d\": [0.0, 170.19, 285.71, 374.0], \"box_3d\": [1.76, 1.58, 3.68, -4.47, 1.76, 5.65], \"ry\": 1.76}, {\"category\": \"Car\", \"angle\": 2.09, \"box_2d\": [381.53, 181.99, 510.77, 237.12], \"box_3d\": [1.42, 1.67, 4.53, -4.71, 1.71, 21.32], \"ry\": 1.87}, {\"category\": \"Car\", \"angle\": -0.53, \"box_2d\": [931.58, 158.46, 1021.01, 190.01], \"box_3d\": [1.54, 1.63, 3.74, 18.64, 0.84, 36.84], \"ry\": -0.07}]\n```" - } - ] - }, - { - "id": 999, - "data_source": "Mono3DRefer", - "scene_name": "005184", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [471.2, 180.84, 554.37, 244.73], \"box_3d\": [1.55, 1.68, 4.42, -2.6, 1.78, 19.94], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [667.84, 181.43, 724.36, 227.38], \"box_3d\": [1.6, 1.53, 4.18, 3.17, 1.93, 27.51], \"ry\": -1.6}]\n```" - } - ] - }, - { - "id": 1000, - "data_source": "Mono3DRefer", - "scene_name": "005185", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [726.63, 182.52, 783.35, 231.04], \"box_3d\": [1.52, 1.54, 4.5, 4.95, 1.88, 25.27], \"ry\": -1.46}, {\"category\": \"Car\", \"angle\": -2.64, \"box_2d\": [298.06, 190.93, 456.39, 265.55], \"box_3d\": [1.59, 1.45, 3.15, -5.36, 2.03, 16.61], \"ry\": -2.94}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [750.29, 186.32, 995.55, 374.0], \"box_3d\": [1.46, 1.53, 3.17, 2.25, 1.61, 7.07], \"ry\": -1.46}]\n```" - } - ] - }, - { - "id": 1001, - "data_source": "Mono3DRefer", - "scene_name": "005190", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [636.82, 179.45, 671.82, 207.4], \"box_3d\": [1.45, 1.59, 3.59, 2.37, 1.83, 39.77], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [653.78, 180.36, 707.72, 223.76], \"box_3d\": [1.54, 1.67, 3.72, 2.62, 1.84, 27.74], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [675.59, 171.99, 789.57, 259.34], \"box_3d\": [1.76, 1.71, 4.35, 2.66, 1.77, 16.87], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": -2.57, \"box_2d\": [55.98, 182.73, 335.49, 286.27], \"box_3d\": [1.52, 1.58, 3.61, -6.56, 1.7, 11.56], \"ry\": -3.08}, {\"category\": \"Car\", \"angle\": 0.4, \"box_2d\": [129.34, 182.12, 405.35, 269.44], \"box_3d\": [1.52, 1.71, 4.49, -6.47, 1.72, 13.9], \"ry\": -0.03}, {\"category\": \"Car\", \"angle\": -2.73, \"box_2d\": [222.13, 185.49, 421.82, 256.77], \"box_3d\": [1.49, 1.61, 3.87, -6.47, 1.8, 16.33], \"ry\": -3.1}, {\"category\": \"Car\", \"angle\": 0.27, \"box_2d\": [331.19, 168.77, 486.61, 232.58], \"box_3d\": [1.95, 1.64, 4.57, -6.47, 1.85, 23.37], \"ry\": 0.0}, {\"category\": \"Car\", \"angle\": -2.84, \"box_2d\": [322.67, 177.32, 452.88, 237.53], \"box_3d\": [1.66, 1.58, 3.31, -6.47, 1.81, 21.11], \"ry\": -3.14}, {\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [663.71, 181.74, 737.17, 232.91], \"box_3d\": [1.44, 1.66, 3.25, 2.67, 1.72, 22.24], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": -2.89, \"box_2d\": [362.21, 169.59, 473.21, 224.73], \"box_3d\": [1.87, 1.61, 3.53, -6.85, 1.79, 25.78], \"ry\": 3.13}, {\"category\": \"Car\", \"angle\": -2.89, \"box_2d\": [389.46, 174.76, 501.54, 217.63], \"box_3d\": [1.72, 1.65, 4.33, -6.92, 1.83, 30.47], \"ry\": -3.11}]\n```" - } - ] - }, - { - "id": 1002, - "data_source": "Mono3DRefer", - "scene_name": "005199", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [523.85, 176.31, 560.8, 204.88], \"box_3d\": [1.46, 1.75, 4.23, -3.68, 1.67, 39.52], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [375.25, 189.91, 483.02, 267.0], \"box_3d\": [1.39, 1.48, 3.76, -3.79, 1.79, 15.5], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 1003, - "data_source": "Mono3DRefer", - "scene_name": "005213", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [591.05, 177.23, 618.21, 204.09], \"box_3d\": [1.65, 1.67, 3.64, -0.4, 1.95, 46.62], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.32, \"box_2d\": [316.57, 179.38, 502.96, 316.33], \"box_3d\": [1.61, 1.51, 4.5, -2.76, 1.72, 10.83], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 2.18, \"box_2d\": [0.0, 198.56, 161.34, 281.41], \"box_3d\": [1.61, 1.72, 3.65, -12.43, 2.26, 16.76], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 1004, - "data_source": "Mono3DRefer", - "scene_name": "005223", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [310.24, 179.53, 487.55, 326.43], \"box_3d\": [1.65, 1.75, 3.39, -2.74, 1.75, 9.95], \"ry\": 1.47}, {\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [438.81, 182.62, 532.43, 257.3], \"box_3d\": [1.57, 1.54, 4.01, -2.92, 1.82, 17.44], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 1005, - "data_source": "Mono3DRefer", - "scene_name": "005224", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.58, \"box_2d\": [61.33, 196.37, 314.95, 267.04], \"box_3d\": [1.36, 1.68, 4.49, -8.93, 1.9, 15.39], \"ry\": 0.06}]\n```" - } - ] - }, - { - "id": 1006, - "data_source": "Mono3DRefer", - "scene_name": "005225", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [537.64, 181.89, 568.81, 206.48], \"box_3d\": [1.52, 1.73, 4.28, -3.72, 2.13, 47.57], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 1007, - "data_source": "Mono3DRefer", - "scene_name": "005250", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.53, \"box_2d\": [507.07, 176.93, 644.01, 307.39], \"box_3d\": [1.54, 1.63, 3.67, -0.47, 1.62, 10.47], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [562.26, 163.25, 645.7, 250.31], \"box_3d\": [1.77, 1.69, 4.32, -0.16, 1.6, 16.96], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.85, \"box_2d\": [744.32, 172.67, 946.11, 319.15], \"box_3d\": [1.6, 1.56, 3.45, 2.87, 1.62, 9.68], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [680.11, 177.53, 791.54, 259.96], \"box_3d\": [1.52, 1.53, 4.07, 2.5, 1.64, 15.48], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [652.66, 174.5, 719.48, 232.12], \"box_3d\": [1.52, 1.53, 4.16, 2.11, 1.59, 21.28], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 0.4, \"box_2d\": [64.36, 184.27, 146.91, 215.74], \"box_3d\": [1.74, 1.87, 3.51, -29.47, 2.43, 42.22], \"ry\": -0.21}, {\"category\": \"Car\", \"angle\": 0.03, \"box_2d\": [341.88, 177.8, 401.36, 203.2], \"box_3d\": [1.53, 1.85, 3.5, -15.06, 1.86, 45.64], \"ry\": -0.29}, {\"category\": \"Car\", \"angle\": 0.17, \"box_2d\": [242.58, 182.02, 307.88, 207.96], \"box_3d\": [1.51, 1.85, 3.37, -20.55, 2.09, 44.36], \"ry\": -0.26}, {\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [641.49, 174.93, 681.92, 211.83], \"box_3d\": [1.4, 1.53, 3.24, 2.02, 1.5, 29.19], \"ry\": -1.52}]\n```" - } - ] - }, - { - "id": 1008, - "data_source": "Mono3DRefer", - "scene_name": "005252", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [522.09, 174.68, 553.06, 201.83], \"box_3d\": [1.65, 1.56, 3.98, -4.62, 1.79, 46.17], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 1009, - "data_source": "Mono3DRefer", - "scene_name": "005257", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.03, \"box_2d\": [22.08, 215.46, 402.67, 374.0], \"box_3d\": [1.47, 1.62, 3.79, -3.48, 1.99, 7.11], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [468.92, 187.45, 537.53, 247.28], \"box_3d\": [1.7, 1.53, 3.82, -3.37, 2.2, 23.09], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [524.75, 186.54, 563.2, 221.36], \"box_3d\": [1.5, 1.53, 3.38, -3.08, 2.17, 33.76], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [558.19, 184.56, 597.24, 209.18], \"box_3d\": [1.56, 1.56, 3.6, -2.15, 2.37, 48.54], \"ry\": 1.85}]\n```" - } - ] - }, - { - "id": 1010, - "data_source": "Mono3DRefer", - "scene_name": "005269", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [668.09, 185.87, 760.37, 253.39], \"box_3d\": [1.41, 1.64, 3.77, 2.37, 1.75, 17.49], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [496.46, 185.87, 536.08, 212.25], \"box_3d\": [1.43, 1.76, 3.97, -5.5, 2.22, 42.51], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [386.27, 191.05, 474.17, 248.09], \"box_3d\": [1.49, 1.61, 3.94, -5.33, 2.07, 21.72], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [429.1, 188.93, 499.05, 233.64], \"box_3d\": [1.48, 1.66, 3.9, -5.38, 2.11, 26.86], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.96, \"box_2d\": [213.87, 204.86, 393.96, 296.26], \"box_3d\": [1.31, 1.58, 4.23, -5.6, 2.0, 13.69], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 1011, - "data_source": "Mono3DRefer", - "scene_name": "005300", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.59, \"box_2d\": [735.59, 184.3, 1069.83, 308.98], \"box_3d\": [1.49, 1.66, 4.16, 4.02, 1.68, 10.2], \"ry\": -0.22}, {\"category\": \"Car\", \"angle\": 2.7, \"box_2d\": [623.45, 166.99, 851.49, 266.56], \"box_3d\": [1.85, 1.68, 4.34, 2.57, 1.76, 14.96], \"ry\": 2.87}, {\"category\": \"Car\", \"angle\": -2.58, \"box_2d\": [0.0, 207.3, 128.8, 373.2], \"box_3d\": [1.51, 1.53, 4.46, -7.68, 1.92, 7.18], \"ry\": 2.91}, {\"category\": \"Car\", \"angle\": -2.8, \"box_2d\": [8.77, 194.13, 255.89, 275.96], \"box_3d\": [1.43, 1.56, 3.85, -9.47, 1.9, 14.54], \"ry\": 2.92}, {\"category\": \"Car\", \"angle\": -2.93, \"box_2d\": [163.33, 191.24, 306.27, 244.53], \"box_3d\": [1.42, 1.62, 3.5, -11.08, 2.01, 21.45], \"ry\": 2.88}, {\"category\": \"Car\", \"angle\": -2.99, \"box_2d\": [274.96, 187.17, 363.91, 220.04], \"box_3d\": [1.41, 1.54, 3.6, -13.35, 2.09, 33.24], \"ry\": 2.91}]\n```" - } - ] - }, - { - "id": 1012, - "data_source": "Mono3DRefer", - "scene_name": "005305", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [500.7, 174.51, 546.08, 202.61], \"box_3d\": [1.42, 1.78, 3.62, -4.14, 1.07, 38.6], \"ry\": 1.64}]\n```" - } - ] - }, - { - "id": 1013, - "data_source": "Mono3DRefer", - "scene_name": "005307", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [699.4, 188.3, 841.79, 283.23], \"box_3d\": [1.41, 1.58, 4.36, 2.72, 1.73, 13.38], \"ry\": -1.59}]\n```" - } - ] - }, - { - "id": 1014, - "data_source": "Mono3DRefer", - "scene_name": "005315", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.06, \"box_2d\": [799.95, 187.55, 1241.0, 374.0], \"box_3d\": [1.56, 1.57, 3.98, 2.7, 1.7, 5.19], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -2.82, \"box_2d\": [342.46, 175.9, 443.48, 220.24], \"box_3d\": [1.61, 1.57, 3.35, -8.27, 1.75, 27.52], \"ry\": -3.11}, {\"category\": \"Car\", \"angle\": 1.38, \"box_2d\": [691.64, 165.75, 736.9, 199.3], \"box_3d\": [1.79, 1.72, 4.35, 5.83, 1.44, 41.0], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": 1.58, \"box_2d\": [563.42, 176.72, 590.52, 202.99], \"box_3d\": [1.45, 1.49, 4.07, -1.95, 1.69, 42.23], \"ry\": 1.53}]\n```" - } - ] - }, - { - "id": 1015, - "data_source": "Mono3DRefer", - "scene_name": "005319", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [288.08, 188.3, 488.4, 329.72], \"box_3d\": [1.46, 1.63, 4.47, -2.83, 1.71, 10.08], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [466.94, 172.0, 557.5, 236.34], \"box_3d\": [1.7, 1.82, 4.45, -2.81, 1.71, 21.41], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": -1.18, \"box_2d\": [644.57, 172.92, 691.66, 204.2], \"box_3d\": [1.57, 1.56, 2.62, 3.04, 1.59, 37.78], \"ry\": -1.1}]\n```" - } - ] - }, - { - "id": 1016, - "data_source": "Mono3DRefer", - "scene_name": "005331", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [502.46, 183.0, 554.4, 219.32], \"box_3d\": [1.45, 1.77, 4.04, -3.52, 1.91, 31.51], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 1017, - "data_source": "Mono3DRefer", - "scene_name": "005340", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [633.46, 178.01, 676.43, 215.8], \"box_3d\": [1.37, 1.56, 3.54, 1.74, 1.59, 28.22], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": -1.46, \"box_2d\": [580.14, 170.49, 624.38, 205.66], \"box_3d\": [1.61, 1.69, 4.26, -0.35, 1.53, 35.46], \"ry\": -1.47}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [649.93, 170.89, 682.87, 201.63], \"box_3d\": [1.53, 1.63, 4.5, 3.01, 1.46, 38.53], \"ry\": -1.48}, {\"category\": \"Car\", \"angle\": -0.72, \"box_2d\": [959.15, 152.38, 1072.5, 190.96], \"box_3d\": [1.48, 1.6, 3.84, 16.21, 0.69, 28.93], \"ry\": -0.21}, {\"category\": \"Car\", \"angle\": -0.65, \"box_2d\": [917.34, 154.2, 1014.56, 190.49], \"box_3d\": [1.55, 1.61, 3.64, 15.8, 0.75, 32.14], \"ry\": -0.2}]\n```" - } - ] - }, - { - "id": 1018, - "data_source": "Mono3DRefer", - "scene_name": "005351", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.61, \"box_2d\": [69.63, 195.63, 311.11, 289.84], \"box_3d\": [1.41, 1.62, 3.6, -7.66, 1.88, 13.45], \"ry\": -1.12}, {\"category\": \"Car\", \"angle\": -0.55, \"box_2d\": [0.0, 179.69, 223.93, 285.07], \"box_3d\": [1.77, 1.96, 4.39, -10.48, 1.93, 14.54], \"ry\": -1.16}]\n```" - } - ] - }, - { - "id": 1019, - "data_source": "Mono3DRefer", - "scene_name": "005357", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.48, \"box_2d\": [691.42, 178.61, 722.06, 203.49], \"box_3d\": [1.54, 1.63, 3.67, 6.3, 1.93, 46.91], \"ry\": -1.34}, {\"category\": \"Car\", \"angle\": 2.43, \"box_2d\": [0.0, 191.55, 144.96, 353.49], \"box_3d\": [1.56, 1.59, 3.69, -6.83, 1.78, 7.21], \"ry\": 1.69}, {\"category\": \"Car\", \"angle\": 2.15, \"box_2d\": [80.86, 160.14, 331.94, 302.83], \"box_3d\": [1.93, 1.59, 4.01, -6.34, 1.79, 11.84], \"ry\": 1.67}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [727.82, 179.82, 844.89, 262.87], \"box_3d\": [1.58, 1.67, 3.84, 3.67, 1.75, 15.81], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [617.63, 177.5, 658.73, 205.34], \"box_3d\": [1.53, 1.67, 3.78, 1.65, 1.81, 41.69], \"ry\": 1.8}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [725.85, 161.63, 795.7, 241.57], \"box_3d\": [2.05, 1.5, 3.83, 4.17, 1.79, 20.55], \"ry\": -1.48}]\n```" - } - ] - }, - { - "id": 1020, - "data_source": "Mono3DRefer", - "scene_name": "005358", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.63, \"box_2d\": [2.16, 193.43, 305.02, 298.49], \"box_3d\": [1.41, 1.54, 3.6, -6.81, 1.75, 10.96], \"ry\": 3.1}]\n```" - } - ] - }, - { - "id": 1021, - "data_source": "Mono3DRefer", - "scene_name": "005379", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.97, \"box_2d\": [234.79, 189.26, 469.38, 322.54], \"box_3d\": [1.44, 1.64, 3.78, -3.34, 1.7, 10.02], \"ry\": 1.66}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [382.44, 167.04, 530.35, 272.51], \"box_3d\": [1.77, 1.71, 4.28, -2.88, 1.7, 14.33], \"ry\": 1.65}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [719.11, 180.0, 799.11, 230.77], \"box_3d\": [1.45, 1.71, 3.93, 4.57, 1.69, 22.81], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [501.69, 174.12, 573.89, 232.52], \"box_3d\": [1.73, 1.65, 3.96, -2.29, 1.79, 23.48], \"ry\": 1.65}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [699.19, 169.75, 763.55, 218.23], \"box_3d\": [1.78, 1.89, 4.37, 4.73, 1.69, 28.91], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [534.65, 174.89, 583.25, 220.42], \"box_3d\": [1.68, 1.51, 3.29, -1.99, 1.78, 28.43], \"ry\": 1.62}]\n```" - } - ] - }, - { - "id": 1022, - "data_source": "Mono3DRefer", - "scene_name": "005393", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [386.94, 179.27, 437.26, 217.81], \"box_3d\": [1.49, 1.54, 2.54, -8.11, 1.77, 29.59], \"ry\": 1.49}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [426.42, 178.24, 454.58, 195.49], \"box_3d\": [1.48, 1.97, 4.33, -15.23, 1.98, 64.94], \"ry\": 1.46}]\n```" - } - ] - }, - { - "id": 1023, - "data_source": "Mono3DRefer", - "scene_name": "005402", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [379.23, 189.49, 502.65, 269.93], \"box_3d\": [1.44, 1.72, 4.14, -3.54, 1.84, 15.65], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 1024, - "data_source": "Mono3DRefer", - "scene_name": "005415", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.09, \"box_2d\": [0.0, 205.4, 415.67, 374.0], \"box_3d\": [1.4, 1.67, 3.72, -3.06, 1.75, 6.03], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": 1.46, \"box_2d\": [634.52, 182.45, 699.39, 234.6], \"box_3d\": [1.49, 1.61, 3.74, 1.68, 1.8, 22.82], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": -2.97, \"box_2d\": [423.49, 180.42, 512.12, 210.47], \"box_3d\": [1.47, 1.68, 4.26, -7.32, 1.88, 37.26], \"ry\": 3.12}]\n```" - } - ] - }, - { - "id": 1025, - "data_source": "Mono3DRefer", - "scene_name": "005421", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [285.21, 180.77, 455.4, 288.78], \"box_3d\": [1.59, 1.63, 3.83, -4.03, 1.74, 12.66], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [659.37, 171.43, 692.23, 202.77], \"box_3d\": [1.62, 1.7, 3.29, 3.53, 1.58, 39.43], \"ry\": -1.5}]\n```" - } - ] - }, - { - "id": 1026, - "data_source": "Mono3DRefer", - "scene_name": "005424", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [500.5, 178.45, 540.82, 212.43], \"box_3d\": [1.7, 1.72, 4.11, -4.75, 2.01, 38.56], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 1027, - "data_source": "Mono3DRefer", - "scene_name": "005435", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.33, \"box_2d\": [357.5, 183.96, 471.07, 216.94], \"box_3d\": [1.36, 1.68, 4.49, -8.56, 1.87, 31.62], \"ry\": 0.07}]\n```" - } - ] - }, - { - "id": 1028, - "data_source": "Mono3DRefer", - "scene_name": "005446", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [222.47, 196.84, 453.68, 351.22], \"box_3d\": [1.45, 1.56, 3.47, -3.15, 1.79, 8.95], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 1029, - "data_source": "Mono3DRefer", - "scene_name": "005448", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [595.61, 171.69, 622.85, 198.58], \"box_3d\": [1.65, 1.67, 3.64, -0.11, 1.61, 46.55], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.19, \"box_2d\": [53.67, 181.94, 455.0, 374.0], \"box_3d\": [1.61, 1.51, 4.5, -2.75, 1.71, 6.74], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 1030, - "data_source": "Mono3DRefer", - "scene_name": "005449", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -3.01, \"box_2d\": [414.12, 176.27, 513.31, 210.78], \"box_3d\": [1.47, 1.53, 4.21, -6.59, 1.66, 32.62], \"ry\": 3.08}, {\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [651.59, 172.12, 704.25, 215.68], \"box_3d\": [1.64, 1.62, 4.2, 2.68, 1.65, 29.62], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [512.58, 180.69, 565.21, 221.62], \"box_3d\": [1.35, 1.47, 3.43, -2.53, 1.64, 25.88], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -2.87, \"box_2d\": [228.66, 178.3, 434.31, 247.06], \"box_3d\": [1.48, 1.61, 4.24, -6.48, 1.64, 17.02], \"ry\": 3.05}, {\"category\": \"Car\", \"angle\": 0.4, \"box_2d\": [171.23, 180.32, 402.55, 261.41], \"box_3d\": [1.5, 1.8, 3.88, -6.45, 1.67, 14.62], \"ry\": -0.01}, {\"category\": \"Car\", \"angle\": 0.54, \"box_2d\": [63.84, 185.32, 366.35, 295.97], \"box_3d\": [1.51, 1.55, 3.72, -5.84, 1.72, 10.83], \"ry\": 0.05}, {\"category\": \"Car\", \"angle\": 0.58, \"box_2d\": [0.0, 179.53, 307.09, 338.79], \"box_3d\": [1.58, 1.66, 4.21, -5.91, 1.68, 8.1], \"ry\": -0.04}, {\"category\": \"Car\", \"angle\": -1.82, \"box_2d\": [725.42, 192.9, 902.84, 300.41], \"box_3d\": [1.27, 1.58, 4.03, 2.86, 1.63, 11.14], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 1031, - "data_source": "Mono3DRefer", - "scene_name": "005451", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [566.63, 178.08, 626.99, 224.61], \"box_3d\": [1.38, 1.51, 4.1, -0.55, 1.57, 23.79], \"ry\": -1.71}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [618.6, 170.52, 770.45, 289.11], \"box_3d\": [1.6, 1.66, 3.55, 1.15, 1.59, 11.67], \"ry\": -1.68}]\n```" - } - ] - }, - { - "id": 1032, - "data_source": "Mono3DRefer", - "scene_name": "005455", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [669.71, 171.35, 740.14, 237.04], \"box_3d\": [1.56, 1.65, 3.69, 2.44, 1.55, 19.07], \"ry\": -1.45}, {\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [434.74, 171.56, 534.2, 240.07], \"box_3d\": [1.76, 1.58, 3.68, -3.49, 1.75, 20.47], \"ry\": 1.76}, {\"category\": \"Car\", \"angle\": 2.3, \"box_2d\": [31.71, 186.71, 275.02, 283.64], \"box_3d\": [1.56, 1.62, 4.19, -8.7, 1.87, 14.17], \"ry\": 1.75}, {\"category\": \"Car\", \"angle\": 2.11, \"box_2d\": [395.29, 183.24, 499.6, 226.22], \"box_3d\": [1.42, 1.67, 4.53, -5.94, 1.83, 26.81], \"ry\": 1.89}, {\"category\": \"Car\", \"angle\": -0.77, \"box_2d\": [772.82, 166.81, 840.9, 194.19], \"box_3d\": [1.54, 1.63, 3.74, 11.5, 1.2, 42.19], \"ry\": -0.5}]\n```" - } - ] - }, - { - "id": 1033, - "data_source": "Mono3DRefer", - "scene_name": "005456", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [324.9, 192.77, 483.74, 304.13], \"box_3d\": [1.45, 1.56, 3.47, -3.17, 1.81, 11.61], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 1034, - "data_source": "Mono3DRefer", - "scene_name": "005466", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.87, \"box_2d\": [729.89, 166.12, 1034.32, 360.25], \"box_3d\": [1.49, 1.64, 4.17, 2.41, 1.47, 7.68], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -2.47, \"box_2d\": [0.0, 191.94, 191.11, 295.06], \"box_3d\": [1.47, 1.58, 3.55, -8.88, 1.8, 11.35], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": -1.51, \"box_2d\": [479.54, 168.98, 548.27, 229.34], \"box_3d\": [1.6, 1.75, 3.89, -2.78, 1.52, 21.29], \"ry\": -1.64}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [661.2, 168.09, 749.79, 232.4], \"box_3d\": [1.59, 1.64, 4.03, 2.5, 1.5, 20.07], \"ry\": -1.64}, {\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [650.56, 171.14, 709.97, 216.45], \"box_3d\": [1.5, 1.8, 4.5, 2.45, 1.48, 26.43], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -1.46, \"box_2d\": [528.6, 174.1, 568.41, 206.09], \"box_3d\": [1.46, 1.56, 3.21, -2.97, 1.55, 34.94], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [638.0, 163.18, 675.9, 201.47], \"box_3d\": [1.83, 1.75, 4.52, 2.32, 1.39, 36.98], \"ry\": -1.54}]\n```" - } - ] - }, - { - "id": 1035, - "data_source": "Mono3DRefer", - "scene_name": "005478", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.36, \"box_2d\": [466.4, 155.43, 569.53, 231.44], \"box_3d\": [1.77, 1.75, 4.67, -2.34, 1.39, 19.25], \"ry\": -1.48}, {\"category\": \"Car\", \"angle\": -0.69, \"box_2d\": [994.19, 127.25, 1241.0, 270.94], \"box_3d\": [1.47, 1.52, 4.43, 6.96, 1.0, 8.21], \"ry\": -0.0}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [551.07, 156.42, 600.78, 197.8], \"box_3d\": [1.6, 1.68, 3.94, -1.39, 0.98, 30.09], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [608.02, 155.07, 631.32, 173.77], \"box_3d\": [1.42, 1.66, 4.11, 0.81, 0.05, 58.95], \"ry\": 1.64}]\n```" - } - ] - }, - { - "id": 1036, - "data_source": "Mono3DRefer", - "scene_name": "005485", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [500.88, 175.87, 563.24, 207.67], \"box_3d\": [1.37, 1.78, 4.65, -3.59, 1.54, 33.77], \"ry\": 1.72}]\n```" - } - ] - }, - { - "id": 1037, - "data_source": "Mono3DRefer", - "scene_name": "005486", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.02, \"box_2d\": [8.49, 174.1, 410.14, 374.0], \"box_3d\": [1.56, 1.68, 3.69, -3.31, 1.59, 6.78], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.17, \"box_2d\": [774.02, 200.57, 1241.0, 374.0], \"box_3d\": [1.44, 1.64, 4.5, 2.73, 1.76, 6.27], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [405.06, 174.56, 505.82, 251.51], \"box_3d\": [1.46, 1.51, 3.24, -3.24, 1.52, 15.46], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.43, \"box_2d\": [682.83, 179.89, 762.83, 235.31], \"box_3d\": [1.38, 1.59, 4.19, 3.01, 1.58, 20.22], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [474.14, 174.19, 542.37, 222.85], \"box_3d\": [1.38, 1.69, 3.74, -3.13, 1.44, 22.56], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [652.24, 173.49, 684.56, 202.01], \"box_3d\": [1.51, 1.56, 3.81, 3.21, 1.57, 40.46], \"ry\": -1.55}]\n```" - } - ] - }, - { - "id": 1038, - "data_source": "Mono3DRefer", - "scene_name": "005493", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.04, \"box_2d\": [0.0, 173.54, 369.3, 374.0], \"box_3d\": [1.68, 1.63, 3.9, -3.12, 1.72, 5.11], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 1.43, \"box_2d\": [355.02, 186.44, 389.32, 212.23], \"box_3d\": [1.65, 1.75, 3.39, -16.18, 2.59, 48.92], \"ry\": 1.12}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [395.3, 187.82, 481.31, 255.67], \"box_3d\": [1.47, 1.7, 4.39, -4.31, 1.88, 18.5], \"ry\": 1.44}]\n```" - } - ] - }, - { - "id": 1039, - "data_source": "Mono3DRefer", - "scene_name": "005520", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.27, \"box_2d\": [7.6, 181.39, 147.73, 235.06], \"box_3d\": [1.5, 1.47, 3.62, -16.3, 1.78, 22.26], \"ry\": 1.65}]\n```" - } - ] - }, - { - "id": 1040, - "data_source": "Mono3DRefer", - "scene_name": "005522", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [427.24, 187.28, 519.63, 249.88], \"box_3d\": [1.44, 1.72, 4.14, -3.58, 1.86, 19.38], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 1041, - "data_source": "Mono3DRefer", - "scene_name": "005545", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [582.37, 162.49, 615.58, 189.9], \"box_3d\": [1.49, 1.68, 4.08, -0.09, 0.45, 41.51], \"ry\": 1.62}]\n```" - } - ] - }, - { - "id": 1042, - "data_source": "Mono3DRefer", - "scene_name": "005550", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [541.54, 176.64, 568.0, 201.51], \"box_3d\": [1.7, 1.72, 4.11, -3.96, 1.99, 51.93], \"ry\": 1.53}]\n```" - } - ] - }, - { - "id": 1043, - "data_source": "Mono3DRefer", - "scene_name": "005558", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.86, 718.86), (cx, cy) = (607.19, 185.22). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [641.99, 169.33, 681.31, 200.83], \"box_3d\": [1.6, 1.76, 3.84, 2.83, 0.78, 38.81], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [572.31, 174.05, 613.93, 211.06], \"box_3d\": [1.39, 1.56, 3.45, -0.6, 0.96, 28.94], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [569.91, 172.51, 601.27, 199.16], \"box_3d\": [1.41, 1.66, 3.76, -1.31, 0.73, 40.37], \"ry\": -1.63}]\n```" - } - ] - }, - { - "id": 1044, - "data_source": "Mono3DRefer", - "scene_name": "005563", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.55, \"box_2d\": [516.7, 180.7, 552.08, 217.25], \"box_3d\": [1.68, 1.63, 3.9, -3.8, 2.08, 35.59], \"ry\": 1.45}, {\"category\": \"Car\", \"angle\": 1.45, \"box_2d\": [489.89, 181.89, 521.85, 203.8], \"box_3d\": [1.47, 1.7, 4.39, -7.46, 2.12, 51.25], \"ry\": 1.3}]\n```" - } - ] - }, - { - "id": 1045, - "data_source": "Mono3DRefer", - "scene_name": "005571", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [592.47, 173.46, 619.64, 200.32], \"box_3d\": [1.65, 1.67, 3.64, -0.31, 1.71, 46.6], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.29, \"box_2d\": [262.0, 180.24, 490.69, 343.09], \"box_3d\": [1.61, 1.51, 4.5, -2.76, 1.72, 9.47], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.26, \"box_2d\": [0.0, 195.43, 92.94, 284.22], \"box_3d\": [1.61, 1.72, 3.65, -12.34, 2.1, 14.15], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 1046, - "data_source": "Mono3DRefer", - "scene_name": "005585", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.51, \"box_2d\": [114.21, 193.45, 443.79, 294.37], \"box_3d\": [1.38, 1.66, 4.38, -5.02, 1.74, 11.14], \"ry\": 0.09}, {\"category\": \"Car\", \"angle\": -0.54, \"box_2d\": [861.34, 171.86, 1241.0, 360.18], \"box_3d\": [1.48, 1.86, 4.18, 4.67, 1.48, 6.61], \"ry\": 0.06}, {\"category\": \"Car\", \"angle\": -0.46, \"box_2d\": [784.17, 156.61, 1163.94, 299.11], \"box_3d\": [1.74, 1.66, 4.37, 4.75, 1.57, 9.91], \"ry\": -0.03}, {\"category\": \"Car\", \"angle\": -2.82, \"box_2d\": [358.42, 184.4, 516.0, 236.6], \"box_3d\": [1.47, 1.67, 4.28, -5.16, 1.84, 21.64], \"ry\": -3.05}, {\"category\": \"Car\", \"angle\": -3.04, \"box_2d\": [412.21, 181.08, 535.6, 220.68], \"box_3d\": [1.38, 1.67, 4.37, -5.03, 1.71, 26.93], \"ry\": 3.06}]\n```" - } - ] - }, - { - "id": 1047, - "data_source": "Mono3DRefer", - "scene_name": "005587", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [593.84, 173.7, 624.99, 204.48], \"box_3d\": [1.65, 1.67, 3.64, -0.03, 1.72, 40.89], \"ry\": -1.55}]\n```" - } - ] - }, - { - "id": 1048, - "data_source": "Mono3DRefer", - "scene_name": "005597", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.6, \"box_2d\": [536.05, 173.58, 563.77, 200.01], \"box_3d\": [1.7, 1.72, 4.11, -4.06, 1.78, 48.97], \"ry\": 1.52}]\n```" - } - ] - }, - { - "id": 1049, - "data_source": "Mono3DRefer", - "scene_name": "005600", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [130.72, 182.66, 430.59, 359.6], \"box_3d\": [1.7, 1.82, 4.45, -3.82, 1.84, 9.27], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.48, \"box_2d\": [662.02, 166.67, 700.92, 201.74], \"box_3d\": [1.57, 1.56, 2.62, 3.36, 1.31, 33.93], \"ry\": -1.38}, {\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [612.62, 167.63, 649.49, 189.56], \"box_3d\": [1.43, 1.68, 4.19, 1.47, 1.11, 49.53], \"ry\": 1.81}]\n```" - } - ] - }, - { - "id": 1050, - "data_source": "Mono3DRefer", - "scene_name": "005614", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [945.93, 180.13, 984.8, 212.53], \"box_3d\": [1.43, 1.55, 3.91, 16.87, 1.79, 34.29], \"ry\": -1.1}]\n```" - } - ] - }, - { - "id": 1051, - "data_source": "Mono3DRefer", - "scene_name": "005618", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.29, \"box_2d\": [0.0, 199.58, 248.12, 334.54], \"box_3d\": [1.47, 1.58, 3.55, -6.79, 1.84, 9.25], \"ry\": -2.91}, {\"category\": \"Car\", \"angle\": -1.48, \"box_2d\": [602.27, 175.1, 687.7, 248.26], \"box_3d\": [1.6, 1.75, 3.89, 0.93, 1.67, 17.8], \"ry\": -1.43}, {\"category\": \"Car\", \"angle\": -1.85, \"box_2d\": [826.11, 170.63, 964.02, 255.23], \"box_3d\": [1.59, 1.64, 4.03, 5.95, 1.58, 15.66], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [805.45, 173.53, 896.11, 228.75], \"box_3d\": [1.5, 1.8, 4.5, 7.15, 1.55, 21.98], \"ry\": -1.43}, {\"category\": \"Car\", \"angle\": -1.47, \"box_2d\": [664.36, 179.28, 707.87, 215.01], \"box_3d\": [1.46, 1.56, 3.21, 3.31, 1.75, 31.32], \"ry\": -1.37}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [786.87, 166.52, 833.84, 210.57], \"box_3d\": [1.83, 1.75, 4.52, 8.87, 1.59, 32.32], \"ry\": -1.36}, {\"category\": \"Car\", \"angle\": -2.15, \"box_2d\": [1013.19, 162.64, 1241.0, 374.0], \"box_3d\": [1.49, 1.64, 4.17, 3.59, 1.44, 3.51], \"ry\": -1.39}]\n```" - } - ] - }, - { - "id": 1052, - "data_source": "Mono3DRefer", - "scene_name": "005635", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.86, 718.86), (cx, cy) = (607.19, 185.22). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.5, \"box_2d\": [737.29, 179.57, 791.89, 225.64], \"box_3d\": [1.6, 1.76, 3.84, 5.96, 1.4, 27.19], \"ry\": -1.29}, {\"category\": \"Car\", \"angle\": -1.31, \"box_2d\": [57.47, 211.17, 140.91, 253.16], \"box_3d\": [1.36, 1.66, 4.18, -20.29, 2.5, 28.82], \"ry\": -1.92}, {\"category\": \"Car\", \"angle\": -1.39, \"box_2d\": [141.73, 200.37, 195.88, 237.95], \"box_3d\": [1.7, 1.76, 3.21, -22.19, 2.52, 36.41], \"ry\": -1.93}, {\"category\": \"Car\", \"angle\": -1.42, \"box_2d\": [798.0, 180.12, 827.44, 200.15], \"box_3d\": [1.39, 1.56, 3.45, 14.88, 1.03, 52.15], \"ry\": -1.15}]\n```" - } - ] - }, - { - "id": 1053, - "data_source": "Mono3DRefer", - "scene_name": "005655", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [531.41, 172.47, 555.54, 194.19], \"box_3d\": [1.65, 1.56, 3.98, -5.26, 1.65, 57.22], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 1054, - "data_source": "Mono3DRefer", - "scene_name": "005657", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.57, \"box_2d\": [502.53, 184.82, 540.39, 224.32], \"box_3d\": [1.68, 1.63, 3.9, -4.14, 2.26, 33.45], \"ry\": 1.45}, {\"category\": \"Car\", \"angle\": 1.45, \"box_2d\": [478.04, 186.53, 512.48, 210.62], \"box_3d\": [1.47, 1.7, 4.39, -7.64, 2.39, 47.59], \"ry\": 1.29}]\n```" - } - ] - }, - { - "id": 1055, - "data_source": "Mono3DRefer", - "scene_name": "005678", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [249.23, 193.18, 470.25, 328.94], \"box_3d\": [1.52, 1.73, 4.28, -3.44, 1.87, 10.71], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [514.03, 184.36, 560.4, 217.89], \"box_3d\": [1.44, 1.7, 3.65, -3.37, 2.0, 33.68], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 1056, - "data_source": "Mono3DRefer", - "scene_name": "005690", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.69, \"box_2d\": [91.58, 189.09, 342.74, 277.39], \"box_3d\": [1.41, 1.54, 3.6, -6.87, 1.72, 12.8], \"ry\": 3.1}]\n```" - } - ] - }, - { - "id": 1057, - "data_source": "Mono3DRefer", - "scene_name": "005701", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [541.55, 179.49, 564.85, 199.26], \"box_3d\": [1.45, 1.53, 3.98, -4.37, 1.98, 55.65], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [492.74, 178.54, 540.61, 218.41], \"box_3d\": [1.65, 1.56, 3.98, -4.12, 1.91, 32.07], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 1058, - "data_source": "Mono3DRefer", - "scene_name": "005710", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.62, \"box_2d\": [530.02, 170.72, 565.05, 205.31], \"box_3d\": [1.97, 1.87, 4.53, -3.76, 1.88, 43.8], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 1059, - "data_source": "Mono3DRefer", - "scene_name": "005713", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [594.98, 177.0, 625.69, 207.34], \"box_3d\": [1.65, 1.67, 3.64, 0.03, 1.9, 41.44], \"ry\": -1.55}]\n```" - } - ] - }, - { - "id": 1060, - "data_source": "Mono3DRefer", - "scene_name": "005716", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [565.33, 175.42, 594.26, 193.52], \"box_3d\": [1.56, 1.56, 3.6, -2.68, 1.81, 64.46], \"ry\": 1.84}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [358.32, 166.3, 504.92, 281.1], \"box_3d\": [1.97, 1.87, 4.53, -3.47, 1.89, 14.78], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [454.44, 187.17, 528.31, 239.57], \"box_3d\": [1.47, 1.62, 3.79, -3.7, 1.95, 22.85], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [520.68, 178.2, 556.95, 211.66], \"box_3d\": [1.7, 1.53, 3.82, -3.83, 2.01, 38.99], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 1061, - "data_source": "Mono3DRefer", - "scene_name": "005720", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [651.01, 178.38, 675.58, 201.47], \"box_3d\": [1.54, 1.63, 3.67, 3.74, 1.94, 50.56], \"ry\": -1.48}, {\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [746.73, 165.7, 870.86, 273.32], \"box_3d\": [1.88, 1.57, 3.71, 3.82, 1.79, 14.58], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [804.83, 188.4, 1023.41, 348.42], \"box_3d\": [1.5, 1.56, 3.69, 3.41, 1.72, 8.9], \"ry\": -1.43}, {\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [711.99, 176.85, 775.63, 231.93], \"box_3d\": [1.7, 1.62, 3.76, 4.4, 1.85, 24.35], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [700.81, 182.37, 736.53, 212.59], \"box_3d\": [1.4, 1.53, 3.56, 5.32, 1.89, 35.84], \"ry\": -1.48}]\n```" - } - ] - }, - { - "id": 1062, - "data_source": "Mono3DRefer", - "scene_name": "005734", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.93, \"box_2d\": [28.99, 180.6, 134.87, 225.12], \"box_3d\": [1.57, 1.57, 3.32, -20.01, 1.88, 27.45], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 2.14, \"box_2d\": [123.57, 179.74, 222.66, 217.38], \"box_3d\": [1.57, 1.2, 5.24, -19.82, 1.89, 32.96], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [471.89, 182.18, 534.06, 224.86], \"box_3d\": [1.41, 1.69, 4.36, -3.87, 1.77, 26.49], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 2.35, \"box_2d\": [0.0, 167.6, 239.16, 374.0], \"box_3d\": [1.72, 1.66, 4.32, -3.75, 1.71, 3.46], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [589.75, 174.88, 628.33, 213.31], \"box_3d\": [1.62, 1.63, 4.5, -0.07, 1.74, 33.02], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 1063, - "data_source": "Mono3DRefer", - "scene_name": "005752", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [545.95, 180.02, 570.61, 201.91], \"box_3d\": [1.66, 1.67, 4.49, -4.11, 2.24, 57.6], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.25, \"box_2d\": [357.3, 183.33, 389.16, 202.24], \"box_3d\": [1.56, 1.72, 3.09, -20.39, 2.48, 62.2], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [584.33, 179.73, 621.78, 217.03], \"box_3d\": [1.62, 1.63, 4.5, -0.34, 1.96, 33.95], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 1064, - "data_source": "Mono3DRefer", - "scene_name": "005753", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [618.68, 175.95, 648.15, 195.36], \"box_3d\": [1.44, 1.68, 4.39, 1.84, 1.7, 56.13], \"ry\": 1.74}]\n```" - } - ] - }, - { - "id": 1065, - "data_source": "Mono3DRefer", - "scene_name": "005768", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.47, \"box_2d\": [638.8, 188.15, 1061.82, 336.76], \"box_3d\": [1.45, 1.68, 4.53, 2.7, 1.67, 8.62], \"ry\": -0.18}]\n```" - } - ] - }, - { - "id": 1066, - "data_source": "Mono3DRefer", - "scene_name": "005774", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [243.36, 171.66, 456.35, 309.38], \"box_3d\": [1.66, 1.67, 4.49, -3.7, 1.68, 11.05], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.06, \"box_2d\": [166.46, 179.87, 236.52, 213.07], \"box_3d\": [1.56, 1.72, 3.09, -20.24, 1.92, 35.82], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [487.19, 177.53, 538.69, 214.86], \"box_3d\": [1.46, 1.59, 4.12, -4.08, 1.67, 30.58], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.18, \"box_2d\": [318.92, 175.8, 358.01, 197.76], \"box_3d\": [1.57, 1.57, 3.32, -20.26, 1.81, 53.9], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [341.65, 170.37, 380.85, 190.83], \"box_3d\": [1.57, 1.2, 5.24, -20.11, 1.41, 58.48], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [534.03, 173.39, 563.14, 199.92], \"box_3d\": [1.72, 1.66, 4.32, -4.19, 1.79, 49.47], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [584.99, 173.97, 622.61, 211.13], \"box_3d\": [1.62, 1.63, 4.5, -0.27, 1.7, 34.06], \"ry\": -1.55}]\n```" - } - ] - }, - { - "id": 1067, - "data_source": "Mono3DRefer", - "scene_name": "005778", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.62, \"box_2d\": [776.35, 150.41, 1162.94, 308.93], \"box_3d\": [1.85, 1.68, 4.34, 4.56, 1.6, 9.58], \"ry\": 3.05}, {\"category\": \"Car\", \"angle\": -2.63, \"box_2d\": [0.0, 196.03, 287.15, 302.54], \"box_3d\": [1.43, 1.56, 3.85, -7.28, 1.83, 11.06], \"ry\": 3.08}, {\"category\": \"Car\", \"angle\": -2.86, \"box_2d\": [214.4, 189.49, 379.35, 251.68], \"box_3d\": [1.42, 1.62, 3.5, -7.84, 1.88, 18.23], \"ry\": 3.02}, {\"category\": \"Car\", \"angle\": -2.97, \"box_2d\": [361.61, 184.12, 456.46, 219.96], \"box_3d\": [1.41, 1.54, 3.6, -8.38, 1.9, 30.18], \"ry\": 3.04}, {\"category\": \"Car\", \"angle\": -0.88, \"box_2d\": [1041.09, 182.72, 1241.0, 374.0], \"box_3d\": [1.49, 1.66, 4.16, 5.31, 1.55, 4.68], \"ry\": -0.06}]\n```" - } - ] - }, - { - "id": 1068, - "data_source": "Mono3DRefer", - "scene_name": "005788", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.48, \"box_2d\": [688.59, 183.2, 719.38, 208.7], \"box_3d\": [1.54, 1.63, 3.67, 6.06, 2.23, 46.39], \"ry\": -1.35}, {\"category\": \"Car\", \"angle\": 2.18, \"box_2d\": [107.57, 194.69, 332.42, 305.08], \"box_3d\": [1.56, 1.59, 3.69, -6.61, 1.99, 12.62], \"ry\": 1.7}, {\"category\": \"Car\", \"angle\": 2.04, \"box_2d\": [279.47, 176.19, 424.55, 268.22], \"box_3d\": [1.93, 1.59, 4.01, -6.02, 2.03, 17.25], \"ry\": 1.71}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [715.67, 182.71, 795.41, 243.42], \"box_3d\": [1.58, 1.67, 3.84, 4.1, 1.89, 21.07], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [650.8, 183.18, 683.67, 204.96], \"box_3d\": [1.53, 1.67, 3.78, 4.26, 2.31, 53.47], \"ry\": 1.86}, {\"category\": \"Car\", \"angle\": -1.61, \"box_2d\": [721.95, 169.71, 770.2, 232.13], \"box_3d\": [2.05, 1.5, 3.83, 4.76, 1.98, 25.75], \"ry\": -1.43}]\n```" - } - ] - }, - { - "id": 1069, - "data_source": "Mono3DRefer", - "scene_name": "005805", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [652.14, 183.67, 683.48, 208.32], \"box_3d\": [1.41, 1.58, 4.36, 3.49, 2.09, 44.47], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 1070, - "data_source": "Mono3DRefer", - "scene_name": "005809", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [137.45, 192.13, 456.93, 373.43], \"box_3d\": [1.44, 1.68, 4.39, -3.15, 1.7, 8.25], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 1071, - "data_source": "Mono3DRefer", - "scene_name": "005821", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [0.0, 206.06, 363.26, 374.0], \"box_3d\": [1.39, 1.48, 3.76, -3.71, 1.78, 6.82], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [499.08, 182.83, 554.34, 222.05], \"box_3d\": [1.46, 1.75, 4.23, -3.37, 1.89, 29.61], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 1072, - "data_source": "Mono3DRefer", - "scene_name": "005823", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [519.6, 181.7, 555.36, 210.87], \"box_3d\": [1.45, 1.53, 3.98, -3.85, 1.94, 38.54], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [357.29, 182.31, 485.27, 274.32], \"box_3d\": [1.65, 1.56, 3.98, -3.82, 1.86, 15.1], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 1073, - "data_source": "Mono3DRefer", - "scene_name": "005829", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.17, \"box_2d\": [752.68, 140.24, 1004.25, 297.51], \"box_3d\": [1.79, 1.72, 4.35, 3.49, 1.44, 10.48], \"ry\": 1.49}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [244.28, 195.26, 426.5, 314.58], \"box_3d\": [1.45, 1.49, 4.07, -4.14, 1.85, 11.42], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [412.92, 189.48, 489.95, 243.04], \"box_3d\": [1.27, 1.54, 3.58, -4.28, 1.75, 19.73], \"ry\": 1.5}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [453.25, 175.8, 513.27, 225.24], \"box_3d\": [1.63, 1.65, 3.91, -4.53, 1.76, 26.07], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [591.93, 173.76, 621.62, 200.29], \"box_3d\": [1.46, 1.53, 3.17, -0.25, 1.54, 41.74], \"ry\": -1.63}]\n```" - } - ] - }, - { - "id": 1074, - "data_source": "Mono3DRefer", - "scene_name": "005833", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.23, \"box_2d\": [363.49, 180.01, 394.51, 198.62], \"box_3d\": [1.57, 1.57, 3.32, -20.25, 2.21, 63.32], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [586.19, 175.8, 623.41, 212.88], \"box_3d\": [1.62, 1.63, 4.5, -0.25, 1.78, 34.14], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [479.58, 174.91, 539.02, 222.74], \"box_3d\": [1.66, 1.67, 4.49, -3.79, 1.77, 27.58], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.15, \"box_2d\": [262.4, 181.94, 312.13, 208.14], \"box_3d\": [1.56, 1.72, 3.09, -20.18, 2.14, 45.16], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [531.34, 179.2, 561.7, 202.7], \"box_3d\": [1.46, 1.59, 4.12, -4.15, 1.89, 47.37], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 1075, - "data_source": "Mono3DRefer", - "scene_name": "005842", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.52, \"box_2d\": [906.51, 146.5, 1241.0, 246.56], \"box_3d\": [1.47, 1.52, 4.43, 7.28, 1.1, 11.63], \"ry\": 0.03}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [577.99, 168.76, 621.69, 205.41], \"box_3d\": [1.6, 1.68, 3.94, -0.45, 1.44, 33.69], \"ry\": 1.65}, {\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [628.41, 166.53, 650.53, 183.72], \"box_3d\": [1.42, 1.66, 4.11, 2.57, 0.92, 62.45], \"ry\": 1.67}, {\"category\": \"Car\", \"angle\": -1.38, \"box_2d\": [512.71, 170.56, 595.87, 233.0], \"box_3d\": [1.77, 1.75, 4.67, -1.68, 1.73, 22.89], \"ry\": -1.46}, {\"category\": \"Car\", \"angle\": -1.97, \"box_2d\": [0.0, 233.22, 124.91, 321.47], \"box_3d\": [1.72, 1.39, 3.07, -13.0, 3.23, 16.94], \"ry\": -2.62}, {\"category\": \"Car\", \"angle\": -0.74, \"box_2d\": [1122.31, 139.29, 1241.0, 278.63], \"box_3d\": [1.48, 1.63, 3.98, 8.54, 1.11, 8.42], \"ry\": 0.03}]\n```" - } - ] - }, - { - "id": 1076, - "data_source": "Mono3DRefer", - "scene_name": "005845", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.12, \"box_2d\": [149.83, 199.6, 235.11, 235.87], \"box_3d\": [1.42, 1.7, 4.14, -18.96, 2.7, 32.91], \"ry\": -1.64}, {\"category\": \"Car\", \"angle\": -1.06, \"box_2d\": [83.01, 198.32, 176.54, 237.07], \"box_3d\": [1.52, 1.71, 4.11, -21.67, 2.74, 32.71], \"ry\": -1.64}]\n```" - } - ] - }, - { - "id": 1077, - "data_source": "Mono3DRefer", - "scene_name": "005847", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.24, \"box_2d\": [1037.44, 98.98, 1241.0, 374.0], \"box_3d\": [1.85, 1.68, 4.34, 4.98, 1.5, 3.98], \"ry\": 3.1}, {\"category\": \"Car\", \"angle\": -2.34, \"box_2d\": [0.0, 213.55, 96.44, 374.0], \"box_3d\": [1.43, 1.56, 3.85, -6.84, 1.83, 6.03], \"ry\": 3.12}, {\"category\": \"Car\", \"angle\": -2.72, \"box_2d\": [92.2, 197.76, 334.3, 286.74], \"box_3d\": [1.42, 1.62, 3.5, -7.11, 1.92, 13.13], \"ry\": 3.07}, {\"category\": \"Car\", \"angle\": -2.9, \"box_2d\": [343.97, 185.45, 459.71, 228.66], \"box_3d\": [1.41, 1.54, 3.6, -7.23, 1.87, 25.14], \"ry\": 3.1}]\n```" - } - ] - }, - { - "id": 1078, - "data_source": "Mono3DRefer", - "scene_name": "005851", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.54, \"box_2d\": [390.06, 192.98, 429.51, 230.07], \"box_3d\": [1.65, 1.75, 3.39, -9.88, 2.67, 35.35], \"ry\": 1.27}]\n```" - } - ] - }, - { - "id": 1079, - "data_source": "Mono3DRefer", - "scene_name": "005854", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [342.91, 191.45, 506.77, 303.07], \"box_3d\": [1.46, 1.75, 4.23, -2.92, 1.82, 12.12], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 1080, - "data_source": "Mono3DRefer", - "scene_name": "005860", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.53, \"box_2d\": [659.5, 178.82, 685.37, 202.1], \"box_3d\": [1.54, 1.63, 3.67, 4.35, 1.97, 50.12], \"ry\": -1.44}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [739.74, 168.08, 839.58, 257.25], \"box_3d\": [1.88, 1.57, 3.71, 4.12, 1.81, 17.21], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [776.98, 185.34, 923.77, 299.82], \"box_3d\": [1.5, 1.56, 3.69, 3.62, 1.72, 11.57], \"ry\": -1.43}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [715.56, 176.77, 771.77, 226.12], \"box_3d\": [1.7, 1.62, 3.76, 4.89, 1.87, 26.95], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": -1.61, \"box_2d\": [707.92, 182.32, 740.49, 210.4], \"box_3d\": [1.4, 1.53, 3.56, 6.01, 1.92, 38.43], \"ry\": -1.46}]\n```" - } - ] - }, - { - "id": 1081, - "data_source": "Mono3DRefer", - "scene_name": "005871", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.48, \"box_2d\": [689.55, 185.3, 720.4, 210.89], \"box_3d\": [1.54, 1.63, 3.67, 6.13, 2.37, 46.49], \"ry\": -1.35}, {\"category\": \"Car\", \"angle\": 2.24, \"box_2d\": [3.11, 196.35, 287.74, 329.4], \"box_3d\": [1.56, 1.59, 3.69, -6.7, 1.96, 10.85], \"ry\": 1.7}, {\"category\": \"Car\", \"angle\": 2.08, \"box_2d\": [227.96, 176.3, 398.6, 280.92], \"box_3d\": [1.93, 1.59, 4.01, -6.17, 2.02, 15.42], \"ry\": 1.7}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [718.5, 185.49, 806.6, 252.84], \"box_3d\": [1.58, 1.67, 3.84, 3.93, 1.94, 19.33], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [640.53, 186.68, 675.43, 210.52], \"box_3d\": [1.53, 1.67, 3.78, 3.31, 2.5, 49.51], \"ry\": 1.83}, {\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [721.09, 172.29, 775.95, 239.52], \"box_3d\": [2.05, 1.5, 3.83, 4.51, 2.06, 24.05], \"ry\": -1.45}]\n```" - } - ] - }, - { - "id": 1082, - "data_source": "Mono3DRefer", - "scene_name": "005884", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [535.27, 165.72, 590.08, 209.97], \"box_3d\": [1.6, 1.68, 3.94, -1.81, 1.36, 28.25], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [597.37, 165.53, 621.4, 184.37], \"box_3d\": [1.42, 1.66, 4.11, -0.03, 0.89, 57.14], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": -1.34, \"box_2d\": [437.74, 165.04, 555.64, 250.28], \"box_3d\": [1.77, 1.75, 4.67, -2.59, 1.64, 17.4], \"ry\": -1.49}, {\"category\": \"Car\", \"angle\": -0.81, \"box_2d\": [1077.66, 122.42, 1241.0, 308.94], \"box_3d\": [1.47, 1.52, 4.43, 6.86, 1.06, 6.47], \"ry\": -0.02}]\n```" - } - ] - }, - { - "id": 1083, - "data_source": "Mono3DRefer", - "scene_name": "005885", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.89, \"box_2d\": [764.11, 191.63, 1013.31, 356.61], \"box_3d\": [1.51, 1.56, 3.81, 3.04, 1.78, 8.83], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.8, \"box_2d\": [706.62, 181.61, 842.33, 276.23], \"box_3d\": [1.57, 1.68, 3.83, 3.0, 1.76, 14.11], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.62, \"box_2d\": [496.08, 181.11, 543.39, 226.64], \"box_3d\": [1.52, 1.5, 3.85, -3.26, 1.83, 26.27], \"ry\": 1.5}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [654.88, 183.19, 699.53, 217.67], \"box_3d\": [1.43, 1.59, 3.68, 2.92, 1.9, 32.34], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [526.7, 182.15, 560.69, 207.77], \"box_3d\": [1.29, 1.59, 4.12, -3.59, 1.81, 39.25], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 1084, - "data_source": "Mono3DRefer", - "scene_name": "005901", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.89, \"box_2d\": [705.93, 189.12, 855.05, 284.64], \"box_3d\": [1.41, 1.51, 3.45, 2.82, 1.72, 12.82], \"ry\": -1.68}, {\"category\": \"Car\", \"angle\": -2.5, \"box_2d\": [0.0, 194.48, 281.46, 330.2], \"box_3d\": [1.44, 1.64, 4.01, -6.42, 1.74, 8.76], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": 0.57, \"box_2d\": [50.51, 187.48, 337.69, 294.42], \"box_3d\": [1.52, 1.54, 3.63, -6.4, 1.77, 11.24], \"ry\": 0.07}, {\"category\": \"Car\", \"angle\": -2.68, \"box_2d\": [116.68, 190.51, 365.54, 276.35], \"box_3d\": [1.43, 1.6, 3.78, -6.74, 1.79, 13.35], \"ry\": 3.14}, {\"category\": \"Car\", \"angle\": -2.72, \"box_2d\": [208.49, 186.66, 407.25, 260.89], \"box_3d\": [1.48, 1.59, 3.64, -6.5, 1.81, 15.64], \"ry\": -3.1}, {\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [651.65, 183.89, 727.38, 235.22], \"box_3d\": [1.42, 1.63, 4.15, 2.35, 1.79, 22.6], \"ry\": -1.64}, {\"category\": \"Car\", \"angle\": 0.26, \"box_2d\": [378.18, 184.11, 503.94, 227.12], \"box_3d\": [1.47, 1.65, 4.15, -6.11, 1.9, 26.21], \"ry\": 0.03}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [644.43, 182.19, 690.79, 213.99], \"box_3d\": [1.35, 1.58, 4.05, 2.57, 1.8, 33.36], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": -2.93, \"box_2d\": [391.71, 182.17, 490.71, 222.71], \"box_3d\": [1.53, 1.5, 3.61, -6.73, 1.93, 28.83], \"ry\": 3.12}, {\"category\": \"Car\", \"angle\": -2.94, \"box_2d\": [408.9, 174.11, 491.57, 215.32], \"box_3d\": [1.7, 1.61, 3.21, -6.88, 1.78, 31.12], \"ry\": 3.13}, {\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [630.14, 174.8, 658.28, 200.12], \"box_3d\": [1.76, 1.71, 4.35, 2.44, 1.93, 52.81], \"ry\": -1.6}]\n```" - } - ] - }, - { - "id": 1085, - "data_source": "Mono3DRefer", - "scene_name": "005920", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [595.08, 183.24, 649.48, 227.88], \"box_3d\": [1.36, 1.69, 3.38, 0.35, 1.73, 24.25], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [422.95, 180.68, 460.51, 203.35], \"box_3d\": [1.59, 1.72, 3.86, -12.36, 2.18, 53.13], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [580.98, 179.74, 619.18, 214.59], \"box_3d\": [1.52, 1.67, 3.61, -0.45, 1.85, 33.54], \"ry\": -1.56}]\n```" - } - ] - }, - { - "id": 1086, - "data_source": "Mono3DRefer", - "scene_name": "005930", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [452.53, 189.5, 505.85, 239.63], \"box_3d\": [1.65, 1.75, 3.39, -4.75, 2.29, 26.36], \"ry\": 1.44}]\n```" - } - ] - }, - { - "id": 1087, - "data_source": "Mono3DRefer", - "scene_name": "005941", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [522.94, 182.65, 561.84, 212.67], \"box_3d\": [1.52, 1.73, 4.28, -3.67, 2.07, 39.47], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 1088, - "data_source": "Mono3DRefer", - "scene_name": "005955", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.17, \"box_2d\": [117.19, 200.33, 444.28, 374.0], \"box_3d\": [1.4, 1.67, 3.72, -3.1, 1.75, 7.59], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": 1.47, \"box_2d\": [631.88, 183.42, 691.01, 231.76], \"box_3d\": [1.49, 1.61, 3.74, 1.64, 1.86, 24.54], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 1089, - "data_source": "Mono3DRefer", - "scene_name": "005956", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.48, \"box_2d\": [731.85, 184.44, 898.26, 330.53], \"box_3d\": [1.57, 1.73, 4.15, 2.57, 1.76, 10.08], \"ry\": 1.73}]\n```" - } - ] - }, - { - "id": 1090, - "data_source": "Mono3DRefer", - "scene_name": "005967", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.8, \"box_2d\": [290.05, 171.95, 452.72, 247.73], \"box_3d\": [1.86, 1.69, 3.69, -6.21, 1.87, 18.92], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": -2.87, \"box_2d\": [342.75, 178.04, 482.44, 232.81], \"box_3d\": [1.68, 1.69, 4.07, -6.39, 1.87, 23.5], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [797.25, 158.53, 861.02, 201.3], \"box_3d\": [2.04, 1.99, 4.82, 11.14, 1.38, 37.11], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": -2.92, \"box_2d\": [388.8, 179.6, 502.32, 219.2], \"box_3d\": [1.48, 1.56, 4.12, -6.47, 1.77, 28.5], \"ry\": 3.14}, {\"category\": \"Car\", \"angle\": -2.94, \"box_2d\": [416.24, 174.03, 521.81, 213.38], \"box_3d\": [1.74, 1.75, 4.57, -6.55, 1.83, 33.68], \"ry\": -3.13}]\n```" - } - ] - }, - { - "id": 1091, - "data_source": "Mono3DRefer", - "scene_name": "005990", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.39, \"box_2d\": [0.0, 202.0, 203.28, 374.0], \"box_3d\": [1.7, 1.43, 3.2, -3.18, 1.86, 2.68], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.53, \"box_2d\": [525.17, 179.93, 552.0, 206.03], \"box_3d\": [1.68, 1.63, 3.9, -4.91, 2.18, 49.03], \"ry\": 1.43}]\n```" - } - ] - }, - { - "id": 1092, - "data_source": "Mono3DRefer", - "scene_name": "005993", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.54, \"box_2d\": [542.15, 187.44, 571.87, 217.78], \"box_3d\": [1.68, 1.63, 3.9, -3.24, 2.58, 43.29], \"ry\": 1.46}]\n```" - } - ] - }, - { - "id": 1093, - "data_source": "Mono3DRefer", - "scene_name": "006003", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.99, \"box_2d\": [796.74, 177.46, 1233.37, 374.0], \"box_3d\": [1.51, 1.75, 3.78, 2.99, 1.57, 6.41], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.19, \"box_2d\": [0.0, 209.3, 252.64, 374.0], \"box_3d\": [1.36, 1.51, 4.02, -6.11, 1.89, 8.65], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [277.0, 179.06, 412.99, 264.72], \"box_3d\": [1.74, 1.65, 4.11, -6.03, 1.89, 16.83], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [532.42, 179.56, 568.89, 204.22], \"box_3d\": [1.41, 1.76, 4.4, -3.58, 1.83, 43.88], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [371.87, 180.4, 457.22, 235.45], \"box_3d\": [1.61, 1.64, 4.0, -6.24, 1.86, 23.31], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [448.51, 183.28, 499.25, 215.99], \"box_3d\": [1.43, 1.7, 3.74, -6.41, 1.94, 34.17], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [635.41, 174.46, 659.61, 195.4], \"box_3d\": [1.55, 1.66, 4.15, 2.86, 1.7, 55.95], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.97, \"box_2d\": [697.34, 167.83, 770.89, 190.79], \"box_3d\": [1.47, 1.7, 4.65, 8.32, 1.17, 48.71], \"ry\": 3.13}]\n```" - } - ] - }, - { - "id": 1094, - "data_source": "Mono3DRefer", - "scene_name": "006021", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.59, \"box_2d\": [688.5, 171.45, 844.08, 240.82], \"box_3d\": [1.62, 1.47, 3.66, 3.94, 1.6, 18.38], \"ry\": -0.38}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [386.5, 195.34, 547.03, 325.13], \"box_3d\": [1.38, 1.43, 3.64, -1.83, 1.74, 10.0], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 1095, - "data_source": "Mono3DRefer", - "scene_name": "006022", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 3.06, \"box_2d\": [488.6, 176.73, 748.49, 265.57], \"box_3d\": [1.47, 1.77, 4.49, 0.07, 1.57, 13.34], \"ry\": 3.07}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [543.44, 184.37, 570.85, 208.65], \"box_3d\": [1.52, 1.62, 4.5, -3.61, 2.31, 48.47], \"ry\": -1.69}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [448.24, 188.15, 482.79, 212.62], \"box_3d\": [1.42, 1.66, 4.5, -9.19, 2.43, 46.05], \"ry\": 1.48}]\n```" - } - ] - }, - { - "id": 1096, - "data_source": "Mono3DRefer", - "scene_name": "006030", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -3.14, \"box_2d\": [930.38, 183.43, 1096.8, 240.04], \"box_3d\": [1.54, 1.68, 4.16, 11.87, 1.88, 21.51], \"ry\": -2.64}, {\"category\": \"Car\", \"angle\": -0.78, \"box_2d\": [704.31, 201.57, 783.27, 231.54], \"box_3d\": [1.37, 1.56, 4.19, 7.05, 2.95, 38.0], \"ry\": -0.6}]\n```" - } - ] - }, - { - "id": 1097, - "data_source": "Mono3DRefer", - "scene_name": "006040", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.61, \"box_2d\": [724.03, 171.48, 796.27, 237.24], \"box_3d\": [1.56, 1.65, 3.69, 3.83, 1.55, 19.04], \"ry\": -1.42}, {\"category\": \"Car\", \"angle\": 2.17, \"box_2d\": [113.53, 173.48, 413.73, 347.6], \"box_3d\": [1.76, 1.58, 3.68, -4.14, 1.79, 9.24], \"ry\": 1.76}, {\"category\": \"Car\", \"angle\": 2.42, \"box_2d\": [0.0, 192.03, 193.7, 322.96], \"box_3d\": [1.56, 1.62, 4.19, -8.2, 1.88, 10.16], \"ry\": 1.75}, {\"category\": \"Car\", \"angle\": 2.07, \"box_2d\": [391.11, 184.63, 509.79, 236.37], \"box_3d\": [1.42, 1.67, 4.53, -4.92, 1.82, 22.77], \"ry\": 1.86}, {\"category\": \"Car\", \"angle\": -0.59, \"box_2d\": [886.99, 161.48, 969.95, 191.7], \"box_3d\": [1.54, 1.63, 3.74, 16.89, 0.97, 38.38], \"ry\": -0.18}]\n```" - } - ] - }, - { - "id": 1098, - "data_source": "Mono3DRefer", - "scene_name": "006050", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.1, \"box_2d\": [0.0, 198.26, 393.11, 374.0], \"box_3d\": [1.44, 1.64, 3.78, -3.54, 1.73, 6.68], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [274.48, 171.48, 486.58, 316.38], \"box_3d\": [1.77, 1.71, 4.28, -3.24, 1.79, 11.02], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [712.86, 185.68, 808.69, 246.67], \"box_3d\": [1.45, 1.71, 3.93, 3.94, 1.82, 19.67], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [460.78, 179.44, 548.05, 248.66], \"box_3d\": [1.73, 1.65, 3.96, -2.87, 1.92, 20.12], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [688.4, 174.43, 763.85, 229.63], \"box_3d\": [1.78, 1.89, 4.37, 3.99, 1.86, 25.68], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [504.42, 180.75, 560.15, 233.03], \"box_3d\": [1.68, 1.51, 3.29, -2.67, 1.96, 25.01], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 1099, - "data_source": "Mono3DRefer", - "scene_name": "006104", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [530.93, 181.36, 566.94, 207.34], \"box_3d\": [1.45, 1.77, 4.04, -3.62, 1.97, 43.0], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 1100, - "data_source": "Mono3DRefer", - "scene_name": "006105", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.56, \"box_2d\": [512.15, 178.6, 548.32, 215.94], \"box_3d\": [1.68, 1.63, 3.9, -3.91, 1.98, 34.88], \"ry\": 1.45}, {\"category\": \"Car\", \"angle\": 1.45, \"box_2d\": [485.62, 182.01, 519.12, 204.98], \"box_3d\": [1.47, 1.7, 4.39, -7.37, 2.1, 48.99], \"ry\": 1.3}]\n```" - } - ] - }, - { - "id": 1101, - "data_source": "Mono3DRefer", - "scene_name": "006120", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [258.92, 181.86, 437.49, 300.48], \"box_3d\": [1.7, 1.72, 4.11, -4.35, 1.87, 12.57], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 1102, - "data_source": "Mono3DRefer", - "scene_name": "006126", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.87, \"box_2d\": [384.13, 179.51, 466.67, 216.25], \"box_3d\": [1.61, 1.57, 3.35, -8.45, 1.94, 33.07], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": 1.4, \"box_2d\": [676.16, 167.65, 714.78, 197.0], \"box_3d\": [1.79, 1.72, 4.35, 5.44, 1.5, 46.57], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 1.58, \"box_2d\": [563.71, 178.0, 587.46, 201.02], \"box_3d\": [1.45, 1.49, 4.07, -2.29, 1.8, 47.9], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": -2.1, \"box_2d\": [818.29, 185.3, 1241.0, 374.0], \"box_3d\": [1.6, 1.53, 4.18, 2.78, 1.7, 4.83], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.84, \"box_2d\": [711.02, 183.38, 898.76, 315.03], \"box_3d\": [1.56, 1.57, 3.98, 2.61, 1.74, 10.76], \"ry\": -1.61}]\n```" - } - ] - }, - { - "id": 1103, - "data_source": "Mono3DRefer", - "scene_name": "006128", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [412.14, 175.48, 521.7, 265.77], \"box_3d\": [1.97, 1.87, 4.53, -3.47, 2.06, 18.18], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [526.22, 184.95, 559.67, 216.12], \"box_3d\": [1.7, 1.53, 3.82, -3.91, 2.43, 42.28], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 1104, - "data_source": "Mono3DRefer", - "scene_name": "006131", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.92, \"box_2d\": [389.48, 177.76, 476.48, 211.08], \"box_3d\": [1.49, 1.64, 3.7, -8.33, 1.75, 34.0], \"ry\": 3.13}]\n```" - } - ] - }, - { - "id": 1105, - "data_source": "Mono3DRefer", - "scene_name": "006140", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [494.35, 180.82, 551.09, 233.46], \"box_3d\": [1.97, 1.87, 4.53, -3.52, 2.31, 29.6], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [547.28, 186.22, 572.64, 210.6], \"box_3d\": [1.7, 1.53, 3.82, -3.71, 2.72, 53.74], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 1106, - "data_source": "Mono3DRefer", - "scene_name": "006146", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.31, \"box_2d\": [902.1, 189.62, 1241.0, 374.0], \"box_3d\": [1.6, 1.53, 4.18, 2.79, 1.7, 2.89], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.89, \"box_2d\": [732.13, 184.04, 981.66, 352.15], \"box_3d\": [1.56, 1.57, 3.98, 2.66, 1.72, 8.88], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -2.85, \"box_2d\": [371.92, 178.46, 459.84, 217.42], \"box_3d\": [1.61, 1.57, 3.35, -8.39, 1.87, 31.22], \"ry\": -3.11}, {\"category\": \"Car\", \"angle\": 1.39, \"box_2d\": [681.86, 167.06, 722.46, 197.69], \"box_3d\": [1.79, 1.72, 4.35, 5.63, 1.48, 44.72], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": 1.58, \"box_2d\": [564.83, 179.05, 589.67, 203.14], \"box_3d\": [1.45, 1.49, 4.07, -2.08, 1.85, 45.88], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 1107, - "data_source": "Mono3DRefer", - "scene_name": "006147", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.83, \"box_2d\": [728.78, 187.46, 949.01, 330.02], \"box_3d\": [1.41, 1.64, 3.77, 2.62, 1.63, 9.31], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [486.55, 182.8, 535.06, 215.27], \"box_3d\": [1.43, 1.76, 3.97, -4.71, 1.93, 34.54], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [257.83, 194.77, 422.98, 291.6], \"box_3d\": [1.49, 1.61, 3.94, -4.96, 1.95, 13.74], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [361.96, 189.0, 472.29, 255.25], \"box_3d\": [1.48, 1.66, 3.9, -4.91, 1.93, 18.75], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 2.33, \"box_2d\": [0.0, 228.54, 201.75, 374.0], \"box_3d\": [1.31, 1.58, 4.23, -5.39, 1.91, 5.83], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 1108, - "data_source": "Mono3DRefer", - "scene_name": "006155", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [680.22, 189.59, 800.0, 275.75], \"box_3d\": [1.41, 1.64, 3.77, 2.38, 1.78, 14.28], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [487.42, 189.01, 531.91, 217.99], \"box_3d\": [1.43, 1.76, 3.97, -5.44, 2.35, 39.31], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [345.08, 195.9, 454.41, 265.14], \"box_3d\": [1.49, 1.61, 3.94, -5.29, 2.13, 18.51], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [404.36, 191.66, 486.31, 243.51], \"box_3d\": [1.48, 1.66, 3.9, -5.32, 2.14, 23.63], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 2.07, \"box_2d\": [62.0, 211.17, 340.29, 341.36], \"box_3d\": [1.31, 1.58, 4.23, -5.56, 1.96, 10.44], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 1109, - "data_source": "Mono3DRefer", - "scene_name": "006159", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.29, \"box_2d\": [0.0, 199.4, 124.08, 374.0], \"box_3d\": [1.47, 1.67, 4.28, -6.49, 1.72, 5.54], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": -2.76, \"box_2d\": [0.0, 192.18, 327.83, 300.19], \"box_3d\": [1.38, 1.67, 4.37, -6.75, 1.71, 10.77], \"ry\": 2.98}, {\"category\": \"Car\", \"angle\": -2.18, \"box_2d\": [1030.14, 150.29, 1241.0, 265.62], \"box_3d\": [1.54, 1.44, 2.54, 7.83, 1.26, 10.99], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -2.35, \"box_2d\": [1152.04, 148.78, 1241.0, 253.82], \"box_3d\": [1.59, 1.8, 3.41, 10.39, 1.22, 10.91], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 0.5, \"box_2d\": [189.97, 183.03, 388.05, 244.51], \"box_3d\": [1.42, 1.41, 4.3, -7.86, 1.69, 17.73], \"ry\": 0.08}]\n```" - } - ] - }, - { - "id": 1110, - "data_source": "Mono3DRefer", - "scene_name": "006181", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.18, \"box_2d\": [0.0, 216.23, 341.61, 374.0], \"box_3d\": [1.42, 1.66, 4.11, -3.06, 1.77, 3.88], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [539.74, 167.27, 570.24, 197.89], \"box_3d\": [1.97, 1.87, 4.53, -3.72, 1.64, 49.18], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 1111, - "data_source": "Mono3DRefer", - "scene_name": "006187", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.34, \"box_2d\": [0.0, 204.64, 177.98, 374.0], \"box_3d\": [1.49, 1.59, 4.44, -6.57, 1.82, 6.39], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": -2.51, \"box_2d\": [0.0, 192.33, 321.69, 328.55], \"box_3d\": [1.48, 1.63, 4.1, -5.99, 1.76, 8.93], \"ry\": -3.09}, {\"category\": \"Car\", \"angle\": 0.58, \"box_2d\": [25.16, 194.7, 349.68, 291.03], \"box_3d\": [1.39, 1.6, 4.35, -6.69, 1.77, 11.58], \"ry\": 0.07}, {\"category\": \"Car\", \"angle\": 0.49, \"box_2d\": [186.59, 184.43, 385.8, 268.22], \"box_3d\": [1.52, 1.48, 3.23, -6.29, 1.77, 14.1], \"ry\": 0.07}, {\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [711.71, 181.47, 875.67, 288.35], \"box_3d\": [1.41, 1.56, 4.08, 2.71, 1.57, 11.74], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 0.36, \"box_2d\": [369.01, 178.5, 496.15, 227.66], \"box_3d\": [1.52, 1.64, 3.65, -5.75, 1.72, 23.42], \"ry\": 0.12}, {\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [670.82, 174.23, 722.02, 215.83], \"box_3d\": [1.43, 1.58, 3.37, 3.11, 1.5, 26.75], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -2.96, \"box_2d\": [404.57, 179.55, 509.9, 215.0], \"box_3d\": [1.41, 1.53, 4.11, -6.4, 1.71, 30.31], \"ry\": 3.12}, {\"category\": \"Car\", \"angle\": -2.98, \"box_2d\": [445.02, 176.45, 523.04, 207.29], \"box_3d\": [1.46, 1.57, 3.6, -6.25, 1.66, 35.86], \"ry\": 3.13}]\n```" - } - ] - }, - { - "id": 1112, - "data_source": "Mono3DRefer", - "scene_name": "006207", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [458.25, 185.73, 531.58, 236.93], \"box_3d\": [1.44, 1.72, 4.14, -3.62, 1.88, 23.11], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 1113, - "data_source": "Mono3DRefer", - "scene_name": "006214", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [505.53, 175.38, 564.53, 205.52], \"box_3d\": [1.37, 1.78, 4.65, -3.63, 1.52, 35.51], \"ry\": 1.72}]\n```" - } - ] - }, - { - "id": 1114, - "data_source": "Mono3DRefer", - "scene_name": "006216", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [506.66, 169.14, 555.74, 215.8], \"box_3d\": [1.97, 1.87, 4.53, -3.56, 1.85, 33.06], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [549.66, 175.74, 573.34, 198.19], \"box_3d\": [1.7, 1.53, 3.82, -3.83, 1.95, 57.18], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 1115, - "data_source": "Mono3DRefer", - "scene_name": "006223", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [326.49, 193.32, 489.67, 298.86], \"box_3d\": [1.44, 1.7, 3.65, -3.27, 1.83, 12.25], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 1116, - "data_source": "Mono3DRefer", - "scene_name": "006239", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [341.31, 188.06, 504.26, 287.38], \"box_3d\": [1.48, 1.87, 4.42, -3.28, 1.8, 13.43], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 1117, - "data_source": "Mono3DRefer", - "scene_name": "006262", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.34, \"box_2d\": [460.69, 184.74, 543.21, 222.04], \"box_3d\": [1.59, 1.63, 3.64, -4.98, 2.15, 32.83], \"ry\": -2.49}, {\"category\": \"Car\", \"angle\": -2.42, \"box_2d\": [484.67, 185.81, 561.44, 213.02], \"box_3d\": [1.39, 1.61, 4.09, -4.79, 2.12, 39.29], \"ry\": -2.54}, {\"category\": \"Car\", \"angle\": -1.5, \"box_2d\": [581.95, 184.03, 629.99, 225.17], \"box_3d\": [1.52, 1.67, 3.61, -0.14, 1.98, 29.0], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [596.57, 188.79, 676.63, 255.72], \"box_3d\": [1.36, 1.69, 3.38, 0.52, 1.77, 17.02], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [310.34, 186.7, 383.11, 226.61], \"box_3d\": [1.59, 1.72, 3.86, -11.46, 2.22, 31.58], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 1118, - "data_source": "Mono3DRefer", - "scene_name": "006263", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.03, \"box_2d\": [118.13, 200.07, 327.37, 310.69], \"box_3d\": [1.42, 1.54, 3.5, -6.09, 1.92, 11.73], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [699.66, 174.85, 820.16, 257.4], \"box_3d\": [1.51, 1.75, 3.78, 2.97, 1.58, 15.27], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [295.07, 192.88, 418.59, 260.65], \"box_3d\": [1.36, 1.51, 4.02, -5.98, 1.89, 17.39], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [404.39, 178.97, 478.32, 232.65], \"box_3d\": [1.74, 1.65, 4.11, -5.92, 1.97, 25.63], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [445.38, 182.16, 500.58, 221.15], \"box_3d\": [1.61, 1.64, 4.0, -6.09, 2.04, 32.28], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -1.61, \"box_2d\": [637.12, 176.03, 657.9, 194.08], \"box_3d\": [1.55, 1.66, 4.15, 3.32, 1.85, 64.57], \"ry\": -1.56}]\n```" - } - ] - }, - { - "id": 1119, - "data_source": "Mono3DRefer", - "scene_name": "006278", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [354.98, 183.69, 509.58, 296.77], \"box_3d\": [1.46, 1.63, 4.47, -2.72, 1.66, 11.84], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [487.29, 170.52, 568.76, 229.55], \"box_3d\": [1.7, 1.82, 4.45, -2.55, 1.66, 23.15], \"ry\": 1.64}, {\"category\": \"Car\", \"angle\": -1.12, \"box_2d\": [641.73, 171.74, 689.99, 202.4], \"box_3d\": [1.57, 1.56, 2.62, 2.98, 1.53, 38.53], \"ry\": -1.04}]\n```" - } - ] - }, - { - "id": 1120, - "data_source": "Mono3DRefer", - "scene_name": "006283", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.14, \"box_2d\": [199.24, 148.55, 266.0, 178.11], \"box_3d\": [1.52, 1.71, 4.11, -20.85, 0.28, 39.98], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": 2.09, \"box_2d\": [106.03, 153.83, 201.53, 194.17], \"box_3d\": [1.42, 1.55, 3.45, -17.28, 0.77, 27.45], \"ry\": 1.53}]\n```" - } - ] - }, - { - "id": 1121, - "data_source": "Mono3DRefer", - "scene_name": "006296", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.21, \"box_2d\": [0.0, 200.13, 207.34, 354.3], \"box_3d\": [1.48, 1.61, 3.61, -6.99, 1.88, 9.19], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [614.05, 173.28, 637.51, 195.59], \"box_3d\": [1.54, 1.63, 3.67, 1.11, 1.6, 52.34], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [473.83, 179.79, 519.93, 207.52], \"box_3d\": [1.56, 1.59, 3.69, -6.68, 1.98, 42.75], \"ry\": 1.74}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [498.12, 172.84, 538.64, 203.73], \"box_3d\": [1.93, 1.59, 4.01, -6.0, 1.96, 47.47], \"ry\": 1.72}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [656.45, 171.75, 681.25, 195.19], \"box_3d\": [1.58, 1.67, 3.84, 4.12, 1.54, 51.15], \"ry\": -1.5}]\n```" - } - ] - }, - { - "id": 1122, - "data_source": "Mono3DRefer", - "scene_name": "006309", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [456.35, 182.93, 551.77, 244.49], \"box_3d\": [1.44, 1.68, 4.39, -2.75, 1.73, 19.39], \"ry\": 1.63}]\n```" - } - ] - }, - { - "id": 1123, - "data_source": "Mono3DRefer", - "scene_name": "006310", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.4, \"box_2d\": [588.92, 167.06, 650.63, 213.57], \"box_3d\": [1.77, 1.75, 4.67, 0.46, 1.58, 29.92], \"ry\": -1.38}, {\"category\": \"Car\", \"angle\": -2.11, \"box_2d\": [240.56, 205.69, 328.61, 261.06], \"box_3d\": [1.72, 1.39, 3.07, -11.33, 2.9, 24.99], \"ry\": -2.53}, {\"category\": \"Car\", \"angle\": -2.01, \"box_2d\": [120.31, 208.59, 242.21, 277.81], \"box_3d\": [1.46, 1.52, 2.75, -10.5, 2.37, 17.57], \"ry\": -2.55}, {\"category\": \"Car\", \"angle\": -0.46, \"box_2d\": [962.8, 156.25, 1215.42, 234.7], \"box_3d\": [1.48, 1.63, 3.98, 9.57, 1.2, 14.73], \"ry\": 0.1}, {\"category\": \"Car\", \"angle\": -0.34, \"box_2d\": [855.69, 157.89, 1064.93, 220.9], \"box_3d\": [1.47, 1.52, 4.43, 8.57, 1.15, 18.01], \"ry\": 0.1}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [636.73, 163.86, 674.96, 193.98], \"box_3d\": [1.6, 1.68, 3.94, 2.6, 1.14, 40.57], \"ry\": 1.75}]\n```" - } - ] - }, - { - "id": 1124, - "data_source": "Mono3DRefer", - "scene_name": "006311", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [475.81, 179.19, 526.94, 220.45], \"box_3d\": [1.7, 1.72, 4.11, -4.8, 2.0, 32.12], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 1125, - "data_source": "Mono3DRefer", - "scene_name": "006317", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.75, \"box_2d\": [260.07, 175.08, 390.28, 230.79], \"box_3d\": [1.61, 1.57, 3.35, -8.69, 1.7, 22.08], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": 1.36, \"box_2d\": [688.6, 164.77, 742.24, 203.58], \"box_3d\": [1.79, 1.72, 4.35, 5.12, 1.44, 35.75], \"ry\": 1.5}, {\"category\": \"Car\", \"angle\": 1.6, \"box_2d\": [542.45, 175.67, 573.88, 205.91], \"box_3d\": [1.45, 1.49, 4.07, -2.64, 1.61, 36.96], \"ry\": 1.52}]\n```" - } - ] - }, - { - "id": 1126, - "data_source": "Mono3DRefer", - "scene_name": "006329", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.51, \"box_2d\": [922.68, 147.43, 1241.0, 297.73], \"box_3d\": [1.71, 1.67, 3.7, 6.19, 1.43, 9.23], \"ry\": 3.09}, {\"category\": \"Car\", \"angle\": 2.2, \"box_2d\": [0.0, 189.66, 206.25, 374.0], \"box_3d\": [1.64, 1.63, 4.51, -4.43, 1.78, 4.51], \"ry\": 1.45}, {\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [111.3, 184.93, 369.81, 334.65], \"box_3d\": [1.6, 1.66, 4.33, -4.84, 1.79, 10.12], \"ry\": 1.5}]\n```" - } - ] - }, - { - "id": 1127, - "data_source": "Mono3DRefer", - "scene_name": "006347", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.43, \"box_2d\": [804.94, 177.15, 1056.88, 372.98], \"box_3d\": [1.57, 1.73, 4.15, 3.13, 1.63, 7.99], \"ry\": 1.79}]\n```" - } - ] - }, - { - "id": 1128, - "data_source": "Mono3DRefer", - "scene_name": "006348", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [360.26, 176.94, 484.27, 260.22], \"box_3d\": [1.5, 1.62, 4.19, -3.84, 1.61, 15.27], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 1129, - "data_source": "Mono3DRefer", - "scene_name": "006350", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [579.52, 174.68, 608.81, 201.05], \"box_3d\": [1.43, 1.55, 3.91, -0.9, 1.56, 41.47], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 1.12, \"box_2d\": [850.97, 183.5, 1241.0, 374.0], \"box_3d\": [1.52, 1.61, 4.23, 3.94, 1.65, 7.46], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.26, \"box_2d\": [803.42, 177.16, 906.11, 236.81], \"box_3d\": [1.45, 1.51, 3.7, 6.46, 1.58, 19.53], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 2.94, \"box_2d\": [663.68, 170.93, 764.58, 206.58], \"box_3d\": [1.51, 1.65, 4.21, 4.58, 1.45, 32.32], \"ry\": 3.08}]\n```" - } - ] - }, - { - "id": 1130, - "data_source": "Mono3DRefer", - "scene_name": "006374", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [608.74, 183.7, 684.94, 239.66], \"box_3d\": [1.38, 1.51, 4.1, 0.89, 1.7, 20.33], \"ry\": -1.68}, {\"category\": \"Car\", \"angle\": -1.93, \"box_2d\": [706.41, 165.0, 966.17, 337.93], \"box_3d\": [1.6, 1.66, 3.55, 2.34, 1.55, 8.57], \"ry\": -1.67}]\n```" - } - ] - }, - { - "id": 1131, - "data_source": "Mono3DRefer", - "scene_name": "006375", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [588.9, 171.86, 625.37, 207.02], \"box_3d\": [1.6, 1.66, 3.55, -0.17, 1.58, 34.93], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 1132, - "data_source": "Mono3DRefer", - "scene_name": "006388", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [618.05, 176.04, 675.62, 230.04], \"box_3d\": [1.82, 1.65, 3.67, 1.24, 1.96, 26.39], \"ry\": -1.64}, {\"category\": \"Car\", \"angle\": -2.21, \"box_2d\": [819.46, 167.94, 1241.0, 374.0], \"box_3d\": [1.72, 1.67, 4.42, 2.97, 1.72, 4.62], \"ry\": -1.67}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [249.02, 198.77, 455.84, 351.9], \"box_3d\": [1.44, 1.56, 3.84, -3.08, 1.83, 9.28], \"ry\": 1.49}, {\"category\": \"Car\", \"angle\": -1.53, \"box_2d\": [474.74, 184.78, 522.6, 231.05], \"box_3d\": [1.5, 1.55, 4.21, -4.01, 1.95, 26.2], \"ry\": -1.69}]\n```" - } - ] - }, - { - "id": 1133, - "data_source": "Mono3DRefer", - "scene_name": "006392", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.58, \"box_2d\": [573.8, 182.27, 604.71, 210.07], \"box_3d\": [1.41, 1.59, 4.47, -1.14, 1.95, 39.67], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 1134, - "data_source": "Mono3DRefer", - "scene_name": "006398", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [430.11, 194.95, 514.28, 265.09], \"box_3d\": [1.5, 1.53, 3.38, -3.39, 2.1, 18.04], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [515.37, 185.79, 574.18, 223.66], \"box_3d\": [1.56, 1.56, 3.6, -2.89, 2.16, 32.31], \"ry\": 1.8}, {\"category\": \"Car\", \"angle\": 1.98, \"box_2d\": [67.93, 190.52, 403.33, 374.0], \"box_3d\": [1.7, 1.53, 3.82, -3.49, 1.92, 7.49], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 1135, - "data_source": "Mono3DRefer", - "scene_name": "006404", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.25, \"box_2d\": [0.0, 186.85, 366.97, 374.0], \"box_3d\": [1.63, 1.6, 3.73, -2.46, 1.71, 2.85], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.94, \"box_2d\": [754.68, 194.61, 1143.81, 374.0], \"box_3d\": [1.54, 1.6, 3.52, 2.37, 1.77, 6.13], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.31, \"box_2d\": [239.24, 193.42, 474.01, 374.0], \"box_3d\": [1.54, 1.5, 3.48, -2.58, 1.81, 8.02], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [388.04, 191.53, 517.23, 287.36], \"box_3d\": [1.44, 1.63, 3.32, -2.75, 1.82, 13.05], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.85, \"box_2d\": [696.89, 184.23, 839.72, 287.1], \"box_3d\": [1.55, 1.55, 3.27, 2.61, 1.77, 12.78], \"ry\": -1.66}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [663.3, 184.86, 751.53, 253.74], \"box_3d\": [1.48, 1.6, 3.79, 2.26, 1.8, 17.85], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [511.31, 184.7, 563.41, 227.93], \"box_3d\": [1.56, 1.55, 3.64, -2.83, 2.04, 28.42], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [541.33, 174.98, 585.34, 218.53], \"box_3d\": [1.91, 1.66, 4.22, -2.16, 2.03, 33.94], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [562.72, 176.97, 601.63, 212.1], \"box_3d\": [1.83, 1.69, 4.28, -1.51, 2.07, 39.95], \"ry\": 1.64}]\n```" - } - ] - }, - { - "id": 1136, - "data_source": "Mono3DRefer", - "scene_name": "006406", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [359.06, 184.74, 488.35, 263.44], \"box_3d\": [1.42, 1.75, 3.45, -3.78, 1.69, 15.07], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 1137, - "data_source": "Mono3DRefer", - "scene_name": "006455", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [389.59, 179.69, 482.63, 232.2], \"box_3d\": [1.42, 1.78, 3.62, -4.86, 1.38, 21.52], \"ry\": 1.62}]\n```" - } - ] - }, - { - "id": 1138, - "data_source": "Mono3DRefer", - "scene_name": "006456", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [437.62, 186.53, 527.89, 246.88], \"box_3d\": [1.45, 1.77, 4.04, -3.44, 1.86, 20.0], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 1139, - "data_source": "Mono3DRefer", - "scene_name": "006484", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [584.58, 175.21, 611.02, 201.34], \"box_3d\": [1.65, 1.67, 3.64, -0.84, 1.83, 47.85], \"ry\": -1.59}]\n```" - } - ] - }, - { - "id": 1140, - "data_source": "Mono3DRefer", - "scene_name": "006485", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [428.02, 185.27, 522.09, 254.24], \"box_3d\": [1.45, 1.56, 3.47, -3.17, 1.77, 17.29], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 1141, - "data_source": "Mono3DRefer", - "scene_name": "006487", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.57, \"box_2d\": [312.94, 184.4, 533.14, 276.59], \"box_3d\": [1.41, 1.62, 3.6, -3.37, 1.65, 13.25], \"ry\": -0.82}, {\"category\": \"Car\", \"angle\": -0.53, \"box_2d\": [212.5, 169.5, 458.14, 269.75], \"box_3d\": [1.77, 1.96, 4.39, -5.61, 1.73, 15.08], \"ry\": -0.88}, {\"category\": \"Car\", \"angle\": 1.4, \"box_2d\": [0.0, 188.39, 72.59, 225.66], \"box_3d\": [1.37, 1.6, 4.5, -23.27, 2.01, 29.13], \"ry\": 0.73}]\n```" - } - ] - }, - { - "id": 1142, - "data_source": "Mono3DRefer", - "scene_name": "006488", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [434.75, 170.39, 528.51, 239.21], \"box_3d\": [1.57, 1.67, 4.14, -3.0, 1.32, 18.62], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 1143, - "data_source": "Mono3DRefer", - "scene_name": "006510", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [538.42, 174.83, 571.38, 195.29], \"box_3d\": [1.42, 1.78, 3.62, -3.33, 0.96, 52.33], \"ry\": 1.68}]\n```" - } - ] - }, - { - "id": 1144, - "data_source": "Mono3DRefer", - "scene_name": "006514", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 3.08, \"box_2d\": [970.84, 178.44, 1168.09, 243.46], \"box_3d\": [1.54, 1.68, 4.16, 11.74, 1.71, 18.82], \"ry\": -2.65}, {\"category\": \"Car\", \"angle\": -0.79, \"box_2d\": [706.08, 197.77, 791.01, 229.83], \"box_3d\": [1.37, 1.56, 4.19, 6.76, 2.64, 35.21], \"ry\": -0.6}]\n```" - } - ] - }, - { - "id": 1145, - "data_source": "Mono3DRefer", - "scene_name": "006518", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.62, \"box_2d\": [526.58, 167.23, 563.42, 203.42], \"box_3d\": [1.97, 1.87, 4.53, -3.75, 1.7, 41.96], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 1146, - "data_source": "Mono3DRefer", - "scene_name": "006520", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.29, \"box_2d\": [694.4, 182.49, 845.04, 246.77], \"box_3d\": [1.55, 1.57, 3.51, 4.07, 1.83, 18.85], \"ry\": -0.09}, {\"category\": \"Car\", \"angle\": -0.58, \"box_2d\": [819.81, 192.59, 1241.0, 350.39], \"box_3d\": [1.47, 1.74, 3.88, 4.53, 1.73, 8.08], \"ry\": -0.08}, {\"category\": \"Car\", \"angle\": 0.55, \"box_2d\": [0.0, 160.19, 200.46, 303.88], \"box_3d\": [1.68, 1.67, 3.79, -7.69, 1.55, 9.27], \"ry\": -0.12}, {\"category\": \"Car\", \"angle\": -2.63, \"box_2d\": [0.0, 183.1, 281.19, 280.67], \"box_3d\": [1.43, 1.6, 4.04, -7.8, 1.62, 11.77], \"ry\": 3.08}, {\"category\": \"Car\", \"angle\": 0.48, \"box_2d\": [87.46, 174.96, 338.92, 257.44], \"box_3d\": [1.53, 1.72, 4.15, -7.95, 1.6, 14.65], \"ry\": -0.01}, {\"category\": \"Car\", \"angle\": -2.83, \"box_2d\": [227.09, 177.81, 397.78, 238.06], \"box_3d\": [1.52, 1.57, 4.01, -8.05, 1.68, 19.67], \"ry\": 3.06}, {\"category\": \"Car\", \"angle\": -0.3, \"box_2d\": [671.72, 186.93, 802.0, 241.47], \"box_3d\": [1.48, 1.63, 3.43, 3.65, 1.92, 21.25], \"ry\": -0.13}, {\"category\": \"Car\", \"angle\": -1.86, \"box_2d\": [458.87, 179.36, 494.0, 201.23], \"box_3d\": [1.66, 1.66, 3.98, -10.66, 2.19, 57.29], \"ry\": -2.04}]\n```" - } - ] - }, - { - "id": 1147, - "data_source": "Mono3DRefer", - "scene_name": "006521", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [678.43, 192.2, 787.25, 268.94], \"box_3d\": [1.41, 1.58, 4.36, 2.57, 1.89, 16.27], \"ry\": -1.6}]\n```" - } - ] - }, - { - "id": 1148, - "data_source": "Mono3DRefer", - "scene_name": "006525", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.35, \"box_2d\": [356.45, 201.06, 518.16, 306.9], \"box_3d\": [1.33, 1.71, 4.09, -2.71, 1.87, 12.06], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [660.16, 183.14, 719.55, 229.16], \"box_3d\": [1.49, 1.56, 4.34, 2.76, 1.88, 26.05], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.42, \"box_2d\": [474.9, 189.99, 549.16, 246.89], \"box_3d\": [1.42, 1.53, 4.12, -2.79, 1.96, 21.01], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [703.24, 178.39, 835.25, 277.99], \"box_3d\": [1.59, 1.59, 3.89, 2.79, 1.71, 13.62], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [639.41, 183.88, 668.61, 208.62], \"box_3d\": [1.38, 1.64, 3.51, 2.62, 2.05, 42.87], \"ry\": -1.49}]\n```" - } - ] - }, - { - "id": 1149, - "data_source": "Mono3DRefer", - "scene_name": "006532", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.66, \"box_2d\": [70.57, 180.83, 373.65, 275.35], \"box_3d\": [1.46, 1.61, 4.31, -6.5, 1.62, 12.31], \"ry\": -3.14}, {\"category\": \"Car\", \"angle\": -2.93, \"box_2d\": [252.11, 163.61, 458.92, 245.2], \"box_3d\": [1.8, 1.77, 4.41, -6.01, 1.62, 17.36], \"ry\": 3.03}, {\"category\": \"Car\", \"angle\": -2.91, \"box_2d\": [413.37, 174.07, 494.56, 210.63], \"box_3d\": [1.58, 1.54, 3.3, -7.03, 1.66, 32.56], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": 0.19, \"box_2d\": [432.1, 173.33, 522.21, 209.5], \"box_3d\": [1.68, 1.71, 4.07, -6.47, 1.73, 35.24], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": 2.35, \"box_2d\": [1044.67, 160.49, 1241.0, 374.0], \"box_3d\": [1.76, 1.71, 4.4, 6.48, 1.66, 6.31], \"ry\": 3.12}]\n```" - } - ] - }, - { - "id": 1150, - "data_source": "Mono3DRefer", - "scene_name": "006534", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.47, \"box_2d\": [0.0, 209.84, 131.38, 374.0], \"box_3d\": [1.46, 1.67, 4.92, -4.9, 1.76, 3.58], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [424.03, 183.68, 523.08, 237.12], \"box_3d\": [1.42, 1.73, 4.41, -4.04, 1.77, 21.88], \"ry\": 1.68}, {\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [701.39, 176.87, 764.05, 227.51], \"box_3d\": [1.51, 1.84, 4.1, 3.91, 1.66, 23.7], \"ry\": -1.44}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [649.18, 170.29, 683.52, 191.53], \"box_3d\": [1.76, 1.77, 4.41, 4.89, 1.57, 62.25], \"ry\": 1.93}]\n```" - } - ] - }, - { - "id": 1151, - "data_source": "Mono3DRefer", - "scene_name": "006535", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.58, \"box_2d\": [35.69, 181.16, 302.0, 289.66], \"box_3d\": [1.58, 1.54, 3.3, -6.93, 1.73, 11.47], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": 0.43, \"box_2d\": [150.1, 174.44, 398.25, 267.92], \"box_3d\": [1.68, 1.71, 4.07, -6.49, 1.74, 14.15], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": -2.82, \"box_2d\": [212.48, 179.99, 431.66, 253.19], \"box_3d\": [1.53, 1.7, 4.32, -6.46, 1.71, 16.45], \"ry\": 3.09}, {\"category\": \"Car\", \"angle\": 0.28, \"box_2d\": [331.46, 183.07, 485.14, 237.19], \"box_3d\": [1.54, 1.78, 4.2, -6.13, 1.88, 22.12], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": 0.13, \"box_2d\": [416.34, 182.83, 523.89, 219.62], \"box_3d\": [1.5, 1.66, 4.4, -6.05, 1.96, 31.38], \"ry\": -0.05}, {\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [650.35, 182.97, 699.29, 220.22], \"box_3d\": [1.41, 1.62, 4.19, 2.6, 1.85, 30.05], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 1152, - "data_source": "Mono3DRefer", - "scene_name": "006550", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.98, \"box_2d\": [61.39, 197.26, 440.31, 374.0], \"box_3d\": [1.6, 1.68, 3.94, -3.03, 1.89, 7.01], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [530.55, 189.37, 570.75, 221.08], \"box_3d\": [1.42, 1.66, 4.11, -2.93, 2.28, 35.95], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 1153, - "data_source": "Mono3DRefer", - "scene_name": "006553", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.43, \"box_2d\": [1023.28, 176.49, 1120.47, 251.89], \"box_3d\": [1.66, 1.56, 3.42, 11.38, 1.76, 17.75], \"ry\": -0.87}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [943.33, 177.6, 999.2, 208.07], \"box_3d\": [1.45, 1.64, 4.21, 18.47, 1.71, 36.86], \"ry\": 2.26}]\n```" - } - ] - }, - { - "id": 1154, - "data_source": "Mono3DRefer", - "scene_name": "006556", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.71, \"box_2d\": [252.39, 173.92, 391.24, 236.11], \"box_3d\": [1.59, 1.45, 3.15, -7.76, 1.64, 19.46], \"ry\": -3.08}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [634.49, 174.28, 761.58, 288.51], \"box_3d\": [1.46, 1.53, 3.17, 1.15, 1.5, 10.9], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 1155, - "data_source": "Mono3DRefer", - "scene_name": "006566", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.58, \"box_2d\": [581.33, 183.06, 614.25, 212.78], \"box_3d\": [1.41, 1.59, 4.47, -0.64, 1.96, 37.37], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 1156, - "data_source": "Mono3DRefer", - "scene_name": "006568", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.57, \"box_2d\": [885.51, 163.43, 1241.0, 301.93], \"box_3d\": [1.76, 1.71, 4.4, 6.42, 1.66, 10.27], \"ry\": 3.12}, {\"category\": \"Car\", \"angle\": 0.78, \"box_2d\": [0.0, 98.45, 157.81, 374.0], \"box_3d\": [2.22, 1.68, 4.08, -6.77, 1.65, 6.62], \"ry\": 0.0}, {\"category\": \"Car\", \"angle\": -2.51, \"box_2d\": [0.0, 182.7, 228.06, 307.8], \"box_3d\": [1.4, 1.46, 3.23, -6.79, 1.54, 8.93], \"ry\": 3.14}, {\"category\": \"Car\", \"angle\": -2.76, \"box_2d\": [203.13, 179.32, 424.77, 250.04], \"box_3d\": [1.46, 1.61, 4.31, -6.56, 1.63, 16.18], \"ry\": 3.14}, {\"category\": \"Car\", \"angle\": -2.98, \"box_2d\": [317.45, 165.87, 481.83, 231.81], \"box_3d\": [1.8, 1.77, 4.41, -6.1, 1.63, 21.24], \"ry\": 3.02}, {\"category\": \"Car\", \"angle\": -2.93, \"box_2d\": [431.61, 173.12, 503.65, 205.74], \"box_3d\": [1.58, 1.54, 3.3, -7.18, 1.62, 36.41], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": 0.17, \"box_2d\": [449.24, 172.31, 529.82, 204.8], \"box_3d\": [1.68, 1.71, 4.07, -6.52, 1.68, 39.14], \"ry\": 0.0}]\n```" - } - ] - }, - { - "id": 1157, - "data_source": "Mono3DRefer", - "scene_name": "006594", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [397.99, 178.04, 529.12, 280.85], \"box_3d\": [1.63, 1.6, 3.73, -2.6, 1.74, 13.39], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -1.9, \"box_2d\": [819.99, 207.78, 1241.0, 374.0], \"box_3d\": [1.45, 1.5, 3.26, 2.58, 1.77, 5.28], \"ry\": -1.47}, {\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [712.13, 181.82, 862.56, 300.35], \"box_3d\": [1.59, 1.49, 3.34, 2.61, 1.75, 11.5], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [666.1, 179.34, 760.58, 253.55], \"box_3d\": [1.54, 1.6, 3.52, 2.26, 1.7, 16.88], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.43, \"box_2d\": [466.5, 189.49, 543.73, 257.26], \"box_3d\": [1.54, 1.5, 3.48, -2.68, 2.0, 18.77], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [491.64, 183.52, 554.08, 231.45], \"box_3d\": [1.44, 1.63, 3.32, -2.85, 1.81, 23.87], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [653.29, 177.19, 719.74, 228.67], \"box_3d\": [1.55, 1.55, 3.27, 2.4, 1.71, 23.62], \"ry\": -1.66}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [638.31, 176.77, 689.58, 217.44], \"box_3d\": [1.48, 1.6, 3.79, 2.03, 1.66, 28.51], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [534.4, 176.8, 570.54, 207.32], \"box_3d\": [1.56, 1.55, 3.64, -3.09, 1.79, 38.99], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [551.76, 170.2, 584.02, 202.83], \"box_3d\": [1.91, 1.66, 4.22, -2.58, 1.78, 44.61], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [565.17, 172.28, 595.41, 199.74], \"box_3d\": [1.83, 1.69, 4.28, -2.06, 1.82, 50.55], \"ry\": 1.63}]\n```" - } - ] - }, - { - "id": 1158, - "data_source": "Mono3DRefer", - "scene_name": "006597", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.02, \"box_2d\": [82.97, 186.3, 354.26, 339.32], \"box_3d\": [1.62, 1.66, 3.81, -5.0, 1.83, 9.79], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [420.67, 182.31, 495.67, 228.04], \"box_3d\": [1.46, 1.67, 4.92, -5.39, 1.82, 26.02], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [506.81, 181.69, 548.71, 206.11], \"box_3d\": [1.42, 1.73, 4.41, -5.1, 1.99, 45.04], \"ry\": 1.66}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [640.77, 176.66, 670.77, 201.3], \"box_3d\": [1.51, 1.84, 4.1, 2.95, 1.78, 46.76], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": -1.95, \"box_2d\": [771.49, 173.3, 1172.31, 374.0], \"box_3d\": [1.61, 1.67, 3.62, 2.63, 1.64, 6.33], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 1159, - "data_source": "Mono3DRefer", - "scene_name": "006606", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [531.05, 174.77, 566.34, 196.72], \"box_3d\": [1.42, 1.78, 3.62, -3.53, 0.98, 48.9], \"ry\": 1.67}]\n```" - } - ] - }, - { - "id": 1160, - "data_source": "Mono3DRefer", - "scene_name": "006610", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.25, \"box_2d\": [0.0, 171.72, 149.3, 236.4], \"box_3d\": [1.78, 1.6, 4.29, -16.4, 1.78, 22.09], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 2.18, \"box_2d\": [137.37, 182.86, 237.72, 225.93], \"box_3d\": [1.5, 1.47, 3.62, -15.98, 1.9, 27.43], \"ry\": 1.66}]\n```" - } - ] - }, - { - "id": 1161, - "data_source": "Mono3DRefer", - "scene_name": "006638", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.71, \"box_2d\": [134.83, 186.13, 383.17, 274.31], \"box_3d\": [1.5, 1.58, 3.88, -6.49, 1.78, 13.51], \"ry\": 3.13}, {\"category\": \"Car\", \"angle\": -0.42, \"box_2d\": [794.22, 181.25, 1034.69, 267.86], \"box_3d\": [1.5, 1.62, 3.88, 5.63, 1.68, 13.73], \"ry\": -0.04}, {\"category\": \"Car\", \"angle\": -2.63, \"box_2d\": [3.81, 190.44, 327.15, 297.4], \"box_3d\": [1.44, 1.55, 3.94, -6.63, 1.74, 10.97], \"ry\": 3.12}, {\"category\": \"Car\", \"angle\": -2.57, \"box_2d\": [0.0, 167.93, 265.56, 335.73], \"box_3d\": [1.77, 1.68, 4.12, -6.59, 1.74, 8.52], \"ry\": 3.07}, {\"category\": \"Car\", \"angle\": 0.8, \"box_2d\": [0.0, 197.02, 170.56, 374.0], \"box_3d\": [1.51, 1.59, 4.28, -6.27, 1.75, 5.9], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": 0.14, \"box_2d\": [430.43, 178.64, 524.26, 213.49], \"box_3d\": [1.58, 1.66, 4.21, -6.31, 1.88, 34.49], \"ry\": -0.04}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [641.02, 183.04, 693.73, 220.14], \"box_3d\": [1.32, 1.63, 4.1, 2.14, 1.73, 28.33], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 0.28, \"box_2d\": [316.37, 178.87, 472.74, 234.63], \"box_3d\": [1.58, 1.56, 4.25, -6.46, 1.78, 21.77], \"ry\": -0.0}]\n```" - } - ] - }, - { - "id": 1162, - "data_source": "Mono3DRefer", - "scene_name": "006641", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.59, \"box_2d\": [199.9, 197.12, 234.55, 226.28], \"box_3d\": [1.45, 1.62, 3.83, -21.9, 2.85, 40.27], \"ry\": 1.09}]\n```" - } - ] - }, - { - "id": 1163, - "data_source": "Mono3DRefer", - "scene_name": "006646", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [317.49, 185.75, 505.47, 316.44], \"box_3d\": [1.56, 1.56, 3.6, -2.74, 1.77, 10.62], \"ry\": 1.64}, {\"category\": \"Car\", \"angle\": 1.5, \"box_2d\": [478.34, 174.48, 501.07, 194.88], \"box_3d\": [1.68, 1.63, 3.9, -10.39, 1.85, 62.07], \"ry\": 1.33}, {\"category\": \"Car\", \"angle\": 2.24, \"box_2d\": [0.0, 202.75, 288.99, 374.0], \"box_3d\": [1.52, 1.54, 3.56, -3.55, 1.76, 4.3], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -2.45, \"box_2d\": [0.0, 202.54, 243.1, 311.98], \"box_3d\": [1.54, 1.64, 4.01, -8.28, 2.05, 11.47], \"ry\": -3.06}]\n```" - } - ] - }, - { - "id": 1164, - "data_source": "Mono3DRefer", - "scene_name": "006654", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.73, \"box_2d\": [990.34, 183.2, 1241.0, 293.19], \"box_3d\": [1.35, 1.49, 4.0, 7.65, 1.52, 10.12], \"ry\": -0.1}, {\"category\": \"Car\", \"angle\": -0.67, \"box_2d\": [947.54, 181.96, 1241.0, 269.58], \"box_3d\": [1.32, 1.62, 4.57, 8.35, 1.5, 12.28], \"ry\": -0.08}]\n```" - } - ] - }, - { - "id": 1165, - "data_source": "Mono3DRefer", - "scene_name": "006660", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.06, \"box_2d\": [837.77, 205.39, 1241.0, 374.0], \"box_3d\": [1.39, 1.51, 3.4, 2.96, 1.7, 5.35], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 0.82, \"box_2d\": [0.0, 165.59, 99.53, 374.0], \"box_3d\": [1.7, 1.57, 3.19, -5.88, 1.66, 5.19], \"ry\": -0.0}, {\"category\": \"Car\", \"angle\": -2.68, \"box_2d\": [0.0, 179.62, 302.25, 310.82], \"box_3d\": [1.63, 1.55, 3.35, -6.32, 1.74, 10.02], \"ry\": 3.05}, {\"category\": \"Car\", \"angle\": 2.75, \"box_2d\": [843.83, 191.45, 1118.16, 275.35], \"box_3d\": [1.34, 1.52, 3.99, 6.31, 1.69, 12.67], \"ry\": -3.08}, {\"category\": \"Car\", \"angle\": -0.28, \"box_2d\": [820.86, 178.17, 1015.77, 260.55], \"box_3d\": [1.63, 1.53, 3.45, 6.32, 1.76, 15.18], \"ry\": 0.11}, {\"category\": \"Car\", \"angle\": 2.85, \"box_2d\": [733.58, 170.35, 887.24, 229.13], \"box_3d\": [1.76, 1.71, 4.4, 6.28, 1.71, 23.05], \"ry\": 3.11}, {\"category\": \"Car\", \"angle\": 0.33, \"box_2d\": [262.23, 151.25, 438.49, 240.14], \"box_3d\": [2.22, 1.68, 4.08, -6.85, 1.7, 19.21], \"ry\": -0.01}, {\"category\": \"Car\", \"angle\": -2.92, \"box_2d\": [379.21, 178.17, 496.88, 217.06], \"box_3d\": [1.46, 1.61, 4.31, -6.83, 1.7, 28.78], \"ry\": 3.13}, {\"category\": \"Car\", \"angle\": -3.08, \"box_2d\": [423.78, 171.1, 521.95, 211.7], \"box_3d\": [1.8, 1.77, 4.41, -6.39, 1.75, 33.86], \"ry\": 3.02}]\n```" - } - ] - }, - { - "id": 1166, - "data_source": "Mono3DRefer", - "scene_name": "006663", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.05, \"box_2d\": [41.29, 188.01, 388.46, 374.0], \"box_3d\": [1.49, 1.52, 3.35, -3.57, 1.67, 7.14], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": -1.99, \"box_2d\": [814.85, 182.12, 1177.1, 374.0], \"box_3d\": [1.53, 1.65, 4.37, 3.68, 1.65, 7.99], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [355.54, 167.64, 492.44, 268.23], \"box_3d\": [1.74, 1.65, 4.18, -3.64, 1.68, 14.7], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [492.58, 174.07, 578.68, 234.16], \"box_3d\": [1.64, 1.45, 3.48, -2.16, 1.69, 21.49], \"ry\": 1.81}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [751.84, 175.21, 865.29, 262.3], \"box_3d\": [1.61, 1.68, 4.04, 4.03, 1.68, 15.4], \"ry\": -1.46}, {\"category\": \"Car\", \"angle\": -1.61, \"box_2d\": [745.83, 179.42, 810.05, 239.1], \"box_3d\": [1.52, 1.58, 4.28, 4.68, 1.72, 20.7], \"ry\": -1.39}, {\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [640.45, 174.9, 709.22, 212.25], \"box_3d\": [1.54, 1.64, 4.08, 2.91, 1.65, 31.98], \"ry\": 2.04}]\n```" - } - ] - }, - { - "id": 1167, - "data_source": "Mono3DRefer", - "scene_name": "006683", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.5, \"box_2d\": [352.54, 194.29, 389.92, 226.26], \"box_3d\": [1.65, 1.75, 3.39, -13.57, 2.9, 40.81], \"ry\": 1.18}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [165.86, 196.52, 440.09, 374.0], \"box_3d\": [1.47, 1.7, 4.39, -3.27, 1.81, 8.57], \"ry\": 1.49}]\n```" - } - ] - }, - { - "id": 1168, - "data_source": "Mono3DRefer", - "scene_name": "006697", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [522.52, 174.7, 560.57, 198.37], \"box_3d\": [1.42, 1.78, 3.62, -3.73, 1.01, 45.47], \"ry\": 1.66}]\n```" - } - ] - }, - { - "id": 1169, - "data_source": "Mono3DRefer", - "scene_name": "006705", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [444.65, 188.12, 533.76, 253.55], \"box_3d\": [1.42, 1.66, 4.11, -3.0, 1.84, 18.41], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.6, \"box_2d\": [564.21, 164.83, 586.84, 188.09], \"box_3d\": [1.97, 1.87, 4.53, -3.03, 1.32, 64.01], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 1170, - "data_source": "Mono3DRefer", - "scene_name": "006727", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [289.16, 196.22, 489.41, 337.0], \"box_3d\": [1.6, 1.68, 3.94, -3.04, 1.99, 10.66], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [540.79, 192.78, 576.66, 221.77], \"box_3d\": [1.42, 1.66, 4.11, -2.79, 2.56, 39.61], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 1171, - "data_source": "Mono3DRefer", - "scene_name": "006734", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [261.11, 191.23, 421.83, 275.91], \"box_3d\": [1.46, 1.67, 4.92, -5.62, 1.91, 15.7], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [469.19, 183.06, 527.83, 215.85], \"box_3d\": [1.42, 1.73, 4.41, -5.24, 1.93, 34.28], \"ry\": 1.66}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [643.48, 175.49, 682.53, 207.52], \"box_3d\": [1.51, 1.84, 4.1, 2.65, 1.66, 36.42], \"ry\": -1.49}]\n```" - } - ] - }, - { - "id": 1172, - "data_source": "Mono3DRefer", - "scene_name": "006739", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [426.81, 179.43, 488.89, 240.08], \"box_3d\": [1.68, 1.63, 3.9, -4.65, 1.9, 22.26], \"ry\": 1.42}, {\"category\": \"Car\", \"angle\": 1.47, \"box_2d\": [442.76, 185.0, 485.56, 216.5], \"box_3d\": [1.47, 1.7, 4.39, -7.5, 2.1, 36.7], \"ry\": 1.27}]\n```" - } - ] - }, - { - "id": 1173, - "data_source": "Mono3DRefer", - "scene_name": "006740", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [590.55, 179.6, 668.13, 241.25], \"box_3d\": [1.61, 1.66, 3.2, 0.44, 1.82, 20.74], \"ry\": -1.72}, {\"category\": \"Car\", \"angle\": 1.96, \"box_2d\": [295.26, 184.0, 355.78, 219.31], \"box_3d\": [1.5, 1.5, 3.1, -12.94, 2.03, 32.9], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [375.93, 184.02, 435.95, 220.97], \"box_3d\": [1.5, 1.67, 4.07, -9.03, 2.02, 32.12], \"ry\": 1.53}]\n```" - } - ] - }, - { - "id": 1174, - "data_source": "Mono3DRefer", - "scene_name": "006748", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.48, \"box_2d\": [349.01, 190.65, 385.28, 220.29], \"box_3d\": [1.65, 1.75, 3.39, -14.66, 2.75, 43.38], \"ry\": 1.15}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [293.72, 192.32, 460.88, 310.28], \"box_3d\": [1.47, 1.7, 4.39, -3.58, 1.83, 11.77], \"ry\": 1.47}]\n```" - } - ] - }, - { - "id": 1175, - "data_source": "Mono3DRefer", - "scene_name": "006750", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (707.05, 707.05), (cx, cy) = (604.08, 180.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.88, \"box_2d\": [511.6, 157.39, 746.62, 251.5], \"box_3d\": [1.46, 1.53, 3.8, 0.34, 1.09, 12.3], \"ry\": 2.91}, {\"category\": \"Car\", \"angle\": -3.02, \"box_2d\": [196.77, 146.21, 472.07, 251.61], \"box_3d\": [1.53, 1.4, 3.94, -4.24, 1.01, 11.47], \"ry\": 2.92}, {\"category\": \"Car\", \"angle\": -2.6, \"box_2d\": [0.0, 149.16, 86.36, 257.35], \"box_3d\": [1.48, 1.55, 3.29, -9.74, 1.07, 10.14], \"ry\": 2.93}]\n```" - } - ] - }, - { - "id": 1176, - "data_source": "Mono3DRefer", - "scene_name": "006757", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [567.01, 175.54, 620.22, 227.03], \"box_3d\": [1.61, 1.66, 3.2, -0.57, 1.72, 24.4], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [476.19, 179.29, 514.94, 210.39], \"box_3d\": [1.66, 1.61, 3.46, -6.42, 2.03, 40.59], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -2.08, \"box_2d\": [957.33, 165.23, 1036.32, 195.23], \"box_3d\": [1.36, 1.5, 4.21, 18.73, 1.04, 35.19], \"ry\": -1.59}]\n```" - } - ] - }, - { - "id": 1177, - "data_source": "Mono3DRefer", - "scene_name": "006763", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [344.1, 173.81, 468.62, 276.68], \"box_3d\": [1.68, 1.63, 3.9, -3.8, 1.73, 13.93], \"ry\": 1.48}, {\"category\": \"Car\", \"angle\": 1.53, \"box_2d\": [432.69, 178.77, 482.46, 219.98], \"box_3d\": [1.47, 1.7, 4.39, -6.08, 1.72, 28.29], \"ry\": 1.32}]\n```" - } - ] - }, - { - "id": 1178, - "data_source": "Mono3DRefer", - "scene_name": "006776", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [452.18, 185.94, 530.27, 242.72], \"box_3d\": [1.45, 1.53, 3.98, -3.4, 1.86, 21.05], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 1179, - "data_source": "Mono3DRefer", - "scene_name": "006785", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [593.89, 169.67, 621.08, 196.54], \"box_3d\": [1.65, 1.67, 3.64, -0.22, 1.48, 46.58], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.25, \"box_2d\": [182.05, 181.0, 475.2, 374.0], \"box_3d\": [1.61, 1.51, 4.5, -2.76, 1.72, 8.1], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 1180, - "data_source": "Mono3DRefer", - "scene_name": "006790", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [487.3, 183.87, 547.84, 225.76], \"box_3d\": [1.45, 1.77, 4.04, -3.5, 1.89, 27.67], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 1181, - "data_source": "Mono3DRefer", - "scene_name": "006802", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [495.73, 178.18, 561.95, 211.75], \"box_3d\": [1.37, 1.78, 4.65, -3.54, 1.62, 32.08], \"ry\": 1.72}]\n```" - } - ] - }, - { - "id": 1182, - "data_source": "Mono3DRefer", - "scene_name": "006808", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.92, \"box_2d\": [782.98, 185.09, 1021.2, 336.24], \"box_3d\": [1.72, 1.67, 4.42, 3.91, 1.93, 10.66], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [481.68, 201.55, 576.92, 283.19], \"box_3d\": [1.44, 1.56, 3.84, -1.68, 2.13, 15.74], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [571.43, 194.65, 608.21, 232.41], \"box_3d\": [1.5, 1.55, 4.21, -0.92, 2.54, 32.72], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [685.38, 186.07, 732.47, 229.98], \"box_3d\": [1.82, 1.65, 3.67, 4.36, 2.44, 32.46], \"ry\": -1.56}]\n```" - } - ] - }, - { - "id": 1183, - "data_source": "Mono3DRefer", - "scene_name": "006817", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.18, \"box_2d\": [0.0, 202.44, 344.89, 374.0], \"box_3d\": [1.57, 1.44, 3.6, -3.67, 1.87, 5.8], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [337.91, 197.53, 511.88, 312.73], \"box_3d\": [1.47, 1.51, 3.5, -2.82, 1.91, 11.56], \"ry\": 1.7}, {\"category\": \"Car\", \"angle\": -1.38, \"box_2d\": [699.31, 170.5, 748.17, 205.06], \"box_3d\": [1.77, 1.75, 4.67, 6.25, 1.67, 39.4], \"ry\": -1.22}, {\"category\": \"Car\", \"angle\": -2.22, \"box_2d\": [460.72, 193.79, 520.69, 230.46], \"box_3d\": [1.72, 1.39, 3.07, -6.06, 2.8, 36.34], \"ry\": -2.38}, {\"category\": \"Car\", \"angle\": -0.25, \"box_2d\": [943.42, 165.47, 1098.34, 214.84], \"box_3d\": [1.48, 1.63, 3.98, 12.89, 1.29, 22.94], \"ry\": 0.26}, {\"category\": \"Car\", \"angle\": -0.11, \"box_2d\": [884.07, 163.93, 1024.15, 206.56], \"box_3d\": [1.47, 1.52, 4.43, 12.34, 1.19, 26.23], \"ry\": 0.33}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [741.38, 168.65, 774.68, 193.07], \"box_3d\": [1.6, 1.68, 3.94, 10.16, 1.34, 49.46], \"ry\": 1.91}]\n```" - } - ] - }, - { - "id": 1184, - "data_source": "Mono3DRefer", - "scene_name": "006819", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [846.31, 117.55, 885.41, 145.91], \"box_3d\": [1.5, 1.62, 4.19, 16.19, -1.81, 45.61], \"ry\": 2.08}]\n```" - } - ] - }, - { - "id": 1185, - "data_source": "Mono3DRefer", - "scene_name": "006823", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [293.64, 173.56, 332.68, 192.96], \"box_3d\": [1.5, 1.62, 4.05, -24.01, 1.59, 58.44], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [568.98, 177.72, 618.58, 225.57], \"box_3d\": [1.6, 1.66, 3.55, -0.63, 1.79, 26.15], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 1.99, \"box_2d\": [255.69, 173.99, 303.92, 194.82], \"box_3d\": [1.43, 1.73, 3.87, -23.77, 1.54, 52.03], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 1186, - "data_source": "Mono3DRefer", - "scene_name": "006833", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.8, \"box_2d\": [0.0, 181.56, 164.89, 374.0], \"box_3d\": [1.53, 1.6, 3.62, -5.67, 1.62, 5.37], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": 0.69, \"box_2d\": [0.0, 175.15, 241.65, 349.09], \"box_3d\": [1.66, 1.57, 3.11, -5.97, 1.71, 7.76], \"ry\": 0.05}, {\"category\": \"Car\", \"angle\": 0.58, \"box_2d\": [0.0, 171.29, 348.48, 302.83], \"box_3d\": [1.66, 1.71, 4.34, -6.19, 1.67, 10.1], \"ry\": 0.04}, {\"category\": \"Car\", \"angle\": -2.64, \"box_2d\": [100.13, 173.28, 336.53, 271.35], \"box_3d\": [1.58, 1.62, 3.26, -6.79, 1.61, 12.65], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": -2.81, \"box_2d\": [197.33, 184.55, 410.45, 264.45], \"box_3d\": [1.51, 1.66, 3.73, -6.27, 1.78, 15.0], \"ry\": 3.08}, {\"category\": \"Car\", \"angle\": -2.75, \"box_2d\": [226.34, 174.59, 431.55, 243.76], \"box_3d\": [1.56, 1.57, 4.37, -6.76, 1.63, 17.51], \"ry\": -3.11}, {\"category\": \"Car\", \"angle\": -2.92, \"box_2d\": [318.96, 169.63, 461.62, 227.41], \"box_3d\": [1.65, 1.68, 3.88, -6.65, 1.58, 22.02], \"ry\": 3.07}, {\"category\": \"Car\", \"angle\": 0.25, \"box_2d\": [338.01, 159.64, 480.82, 224.41], \"box_3d\": [2.08, 1.8, 4.37, -6.78, 1.68, 24.58], \"ry\": -0.02}, {\"category\": \"Car\", \"angle\": -1.89, \"box_2d\": [768.71, 193.46, 1032.81, 365.48], \"box_3d\": [1.38, 1.49, 3.38, 2.79, 1.64, 7.81], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -0.39, \"box_2d\": [840.01, 173.06, 1110.5, 275.4], \"box_3d\": [1.67, 1.6, 3.94, 6.26, 1.71, 12.79], \"ry\": 0.06}, {\"category\": \"Car\", \"angle\": -0.32, \"box_2d\": [809.57, 179.05, 1003.89, 258.28], \"box_3d\": [1.54, 1.75, 3.36, 6.05, 1.69, 15.13], \"ry\": 0.05}, {\"category\": \"Car\", \"angle\": -0.29, \"box_2d\": [776.83, 180.03, 930.85, 243.05], \"box_3d\": [1.48, 1.52, 3.33, 5.97, 1.67, 18.03], \"ry\": 0.02}, {\"category\": \"Car\", \"angle\": 2.84, \"box_2d\": [730.98, 177.26, 869.89, 229.59], \"box_3d\": [1.57, 1.68, 3.97, 6.01, 1.74, 23.16], \"ry\": 3.09}, {\"category\": \"Car\", \"angle\": -0.27, \"box_2d\": [710.28, 179.02, 839.92, 221.58], \"box_3d\": [1.44, 1.65, 4.26, 5.86, 1.68, 26.06], \"ry\": -0.05}, {\"category\": \"Car\", \"angle\": -0.23, \"box_2d\": [712.67, 170.34, 820.3, 217.8], \"box_3d\": [1.77, 1.69, 3.82, 6.05, 1.7, 28.33], \"ry\": -0.02}, {\"category\": \"Car\", \"angle\": -2.97, \"box_2d\": [425.89, 173.06, 521.43, 204.81], \"box_3d\": [1.49, 1.59, 4.44, -6.74, 1.53, 35.78], \"ry\": 3.13}, {\"category\": \"Car\", \"angle\": -2.96, \"box_2d\": [450.54, 176.0, 532.91, 205.22], \"box_3d\": [1.48, 1.63, 4.1, -6.3, 1.68, 38.51], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [635.38, 177.61, 670.18, 203.7], \"box_3d\": [1.41, 1.56, 4.08, 2.39, 1.71, 41.61], \"ry\": -1.62}]\n```" - } - ] - }, - { - "id": 1187, - "data_source": "Mono3DRefer", - "scene_name": "006838", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [538.78, 181.77, 556.77, 198.72], \"box_3d\": [1.47, 1.43, 3.11, -5.6, 2.28, 64.96], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 2.61, \"box_2d\": [941.31, 192.36, 1022.31, 221.67], \"box_3d\": [1.68, 1.39, 4.26, 22.64, 2.91, 44.03], \"ry\": 3.08}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [561.19, 179.47, 615.52, 233.62], \"box_3d\": [1.62, 1.63, 4.5, -0.73, 1.86, 24.1], \"ry\": -1.59}]\n```" - } - ] - }, - { - "id": 1188, - "data_source": "Mono3DRefer", - "scene_name": "006860", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [298.44, 175.17, 463.0, 281.18], \"box_3d\": [1.5, 1.62, 4.19, -3.76, 1.57, 12.45], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 1189, - "data_source": "Mono3DRefer", - "scene_name": "006871", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [279.24, 191.71, 412.85, 269.64], \"box_3d\": [1.33, 1.56, 3.49, -5.29, 1.76, 14.78], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 2.18, \"box_2d\": [0.0, 198.1, 125.86, 313.84], \"box_3d\": [1.47, 1.72, 4.33, -9.08, 1.89, 10.26], \"ry\": 1.47}, {\"category\": \"Car\", \"angle\": 1.99, \"box_2d\": [94.6, 195.42, 262.86, 269.32], \"box_3d\": [1.31, 1.6, 4.09, -9.32, 1.86, 15.91], \"ry\": 1.47}, {\"category\": \"Car\", \"angle\": 2.18, \"box_2d\": [231.7, 188.47, 349.9, 239.57], \"box_3d\": [1.43, 1.58, 3.66, -10.01, 1.96, 22.83], \"ry\": 1.77}, {\"category\": \"Car\", \"angle\": -1.53, \"box_2d\": [658.28, 176.17, 690.11, 203.08], \"box_3d\": [1.37, 1.56, 3.54, 3.47, 1.57, 38.91], \"ry\": -1.44}]\n```" - } - ] - }, - { - "id": 1190, - "data_source": "Mono3DRefer", - "scene_name": "006880", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [538.97, 179.65, 563.71, 200.52], \"box_3d\": [1.45, 1.53, 3.98, -4.28, 1.96, 52.83], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [482.47, 178.44, 536.08, 222.42], \"box_3d\": [1.65, 1.56, 3.98, -4.05, 1.89, 29.25], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 1191, - "data_source": "Mono3DRefer", - "scene_name": "006885", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [311.28, 187.96, 343.31, 215.63], \"box_3d\": [1.46, 1.6, 3.71, -16.19, 2.34, 41.16], \"ry\": -1.97}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [382.31, 185.74, 471.47, 250.27], \"box_3d\": [1.57, 1.72, 3.52, -5.14, 1.93, 19.8], \"ry\": -2.01}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [0.0, 144.2, 170.2, 358.82], \"box_3d\": [2.01, 1.89, 4.91, -6.77, 1.77, 8.99], \"ry\": 1.07}, {\"category\": \"Car\", \"angle\": -1.53, \"box_2d\": [147.81, 189.89, 207.87, 241.61], \"box_3d\": [1.65, 1.67, 3.81, -15.38, 2.28, 25.75], \"ry\": -2.07}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [356.74, 187.97, 416.1, 235.56], \"box_3d\": [1.52, 1.51, 3.1, -7.94, 2.07, 25.37], \"ry\": -2.03}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [334.06, 188.05, 387.0, 229.12], \"box_3d\": [1.53, 1.58, 3.53, -10.26, 2.17, 29.44], \"ry\": -2.04}]\n```" - } - ] - }, - { - "id": 1192, - "data_source": "Mono3DRefer", - "scene_name": "006886", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.08, \"box_2d\": [118.97, 198.64, 249.84, 264.08], \"box_3d\": [1.51, 1.46, 3.63, -11.45, 2.26, 19.61], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [497.34, 172.58, 534.31, 205.18], \"box_3d\": [1.78, 1.72, 4.0, -5.43, 1.8, 41.83], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [364.6, 183.75, 431.57, 226.03], \"box_3d\": [1.47, 1.69, 3.74, -8.05, 1.9, 27.57], \"ry\": 1.5}, {\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [285.6, 187.68, 361.56, 228.47], \"box_3d\": [1.48, 1.54, 4.07, -11.53, 2.11, 29.23], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 0.15, \"box_2d\": [275.38, 184.24, 340.84, 204.32], \"box_3d\": [1.43, 1.55, 4.37, -22.75, 2.31, 54.45], \"ry\": -0.25}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [959.74, 158.2, 996.37, 177.42], \"box_3d\": [1.48, 1.61, 3.61, 29.56, 0.35, 57.94], \"ry\": 2.34}]\n```" - } - ] - }, - { - "id": 1193, - "data_source": "Mono3DRefer", - "scene_name": "006899", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [860.75, 169.73, 954.35, 262.12], \"box_3d\": [1.66, 1.56, 3.42, 6.01, 1.62, 14.81], \"ry\": -1.2}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [728.78, 175.42, 774.92, 207.1], \"box_3d\": [1.45, 1.64, 4.21, 6.96, 1.59, 35.29], \"ry\": 1.9}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [777.91, 175.23, 810.74, 196.91], \"box_3d\": [1.45, 1.67, 4.43, 12.98, 1.63, 50.71], \"ry\": 1.95}]\n```" - } - ] - }, - { - "id": 1194, - "data_source": "Mono3DRefer", - "scene_name": "006917", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.35, \"box_2d\": [718.74, 176.86, 876.02, 295.48], \"box_3d\": [1.62, 1.58, 3.89, 2.83, 1.7, 11.86], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [435.65, 185.21, 533.7, 251.68], \"box_3d\": [1.45, 1.67, 4.43, -3.1, 1.79, 18.46], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [678.45, 181.42, 767.52, 245.17], \"box_3d\": [1.47, 1.59, 3.62, 2.77, 1.7, 18.67], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": 1.47, \"box_2d\": [664.91, 181.45, 729.2, 229.75], \"box_3d\": [1.44, 1.64, 4.5, 2.78, 1.74, 24.15], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [524.58, 180.91, 562.53, 214.4], \"box_3d\": [1.46, 1.51, 3.24, -3.07, 1.84, 33.45], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 1195, - "data_source": "Mono3DRefer", - "scene_name": "006920", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.19, \"box_2d\": [0.0, 210.62, 234.86, 374.0], \"box_3d\": [1.4, 1.59, 3.89, -5.27, 1.84, 6.65], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 0.97, \"box_2d\": [884.06, 193.28, 1241.0, 374.0], \"box_3d\": [1.44, 1.59, 3.51, 2.51, 1.56, 3.03], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [664.84, 177.98, 740.54, 233.27], \"box_3d\": [1.41, 1.64, 3.77, 2.5, 1.58, 20.56], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [510.91, 180.25, 545.84, 204.2], \"box_3d\": [1.43, 1.76, 3.97, -5.15, 1.91, 45.72], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [303.42, 200.76, 434.85, 271.45], \"box_3d\": [1.31, 1.58, 4.23, -5.48, 2.03, 16.84], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 1196, - "data_source": "Mono3DRefer", - "scene_name": "006931", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [534.64, 176.52, 556.21, 196.16], \"box_3d\": [1.65, 1.56, 3.98, -5.63, 1.99, 63.09], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 1197, - "data_source": "Mono3DRefer", - "scene_name": "006948", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.16, \"box_2d\": [0.0, 184.67, 261.77, 370.54], \"box_3d\": [1.74, 1.65, 4.11, -6.17, 1.9, 8.91], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [490.79, 181.28, 545.61, 216.2], \"box_3d\": [1.41, 1.76, 4.4, -3.99, 1.79, 31.76], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.97, \"box_2d\": [215.83, 184.95, 375.94, 274.09], \"box_3d\": [1.61, 1.64, 4.0, -6.51, 1.89, 15.36], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [384.32, 186.37, 456.81, 230.16], \"box_3d\": [1.43, 1.7, 3.74, -6.84, 1.95, 26.27], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.61, \"box_2d\": [629.28, 173.77, 657.31, 198.31], \"box_3d\": [1.55, 1.66, 4.15, 2.15, 1.64, 48.04], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.95, \"box_2d\": [703.57, 164.65, 791.58, 192.02], \"box_3d\": [1.47, 1.7, 4.65, 7.74, 1.05, 40.97], \"ry\": 3.13}, {\"category\": \"Car\", \"angle\": 2.98, \"box_2d\": [699.12, 159.11, 775.54, 191.8], \"box_3d\": [1.91, 1.81, 4.33, 7.7, 1.12, 44.04], \"ry\": -3.13}]\n```" - } - ] - }, - { - "id": 1198, - "data_source": "Mono3DRefer", - "scene_name": "006949", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.88, \"box_2d\": [778.22, 176.23, 868.17, 224.82], \"box_3d\": [1.4, 1.55, 4.06, 6.68, 1.53, 23.14], \"ry\": -1.6}]\n```" - } - ] - }, - { - "id": 1199, - "data_source": "Mono3DRefer", - "scene_name": "006957", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [569.37, 181.72, 666.35, 273.86], \"box_3d\": [1.56, 1.65, 3.69, 0.09, 1.75, 14.24], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 0.58, \"box_2d\": [0.0, 196.71, 74.52, 233.31], \"box_3d\": [1.43, 1.45, 3.5, -24.7, 2.46, 30.08], \"ry\": -0.11}]\n```" - } - ] - }, - { - "id": 1200, - "data_source": "Mono3DRefer", - "scene_name": "006960", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [638.9, 175.74, 660.6, 194.44], \"box_3d\": [1.55, 1.66, 4.15, 3.39, 1.82, 62.42], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 3.0, \"box_2d\": [694.28, 169.37, 758.6, 189.56], \"box_3d\": [1.47, 1.7, 4.65, 8.86, 1.24, 55.26], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": 2.11, \"box_2d\": [0.0, 204.69, 276.28, 347.52], \"box_3d\": [1.42, 1.54, 3.5, -6.06, 1.91, 9.59], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [715.16, 174.28, 864.75, 273.16], \"box_3d\": [1.51, 1.75, 3.78, 3.02, 1.56, 13.06], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [246.26, 194.95, 396.18, 274.15], \"box_3d\": [1.36, 1.51, 4.02, -5.95, 1.88, 15.28], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [385.01, 179.45, 468.06, 238.34], \"box_3d\": [1.74, 1.65, 4.11, -5.91, 1.96, 23.55], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [434.11, 182.97, 494.56, 225.07], \"box_3d\": [1.61, 1.64, 4.0, -6.03, 2.05, 30.11], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [480.54, 183.05, 521.26, 210.2], \"box_3d\": [1.43, 1.7, 3.74, -6.14, 2.02, 40.76], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 1201, - "data_source": "Mono3DRefer", - "scene_name": "006967", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.47, \"box_2d\": [0.0, 194.88, 79.07, 283.64], \"box_3d\": [1.43, 1.58, 3.66, -10.94, 1.83, 11.67], \"ry\": 1.72}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [635.58, 177.68, 676.64, 213.79], \"box_3d\": [1.37, 1.56, 3.54, 1.87, 1.58, 29.45], \"ry\": -1.49}, {\"category\": \"Car\", \"angle\": -1.46, \"box_2d\": [586.71, 170.85, 629.13, 204.53], \"box_3d\": [1.61, 1.69, 4.26, -0.08, 1.54, 36.94], \"ry\": -1.46}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [654.16, 171.25, 685.9, 200.85], \"box_3d\": [1.53, 1.63, 4.5, 3.32, 1.48, 39.91], \"ry\": -1.48}, {\"category\": \"Car\", \"angle\": -0.7, \"box_2d\": [949.94, 157.49, 1058.3, 194.47], \"box_3d\": [1.48, 1.6, 3.84, 16.47, 0.87, 30.25], \"ry\": -0.2}, {\"category\": \"Car\", \"angle\": -0.64, \"box_2d\": [909.87, 155.07, 1003.09, 189.88], \"box_3d\": [1.55, 1.61, 3.64, 16.02, 0.76, 33.49], \"ry\": -0.19}]\n```" - } - ] - }, - { - "id": 1202, - "data_source": "Mono3DRefer", - "scene_name": "006968", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.41, \"box_2d\": [454.52, 188.34, 558.22, 256.76], \"box_3d\": [1.33, 1.71, 4.09, -2.31, 1.73, 16.75], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [676.58, 173.24, 728.91, 211.55], \"box_3d\": [1.49, 1.56, 4.34, 3.82, 1.54, 30.55], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.45, \"box_2d\": [524.71, 181.3, 581.82, 225.46], \"box_3d\": [1.42, 1.53, 4.12, -1.98, 1.74, 25.67], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [707.94, 173.73, 796.32, 244.02], \"box_3d\": [1.59, 1.59, 3.89, 3.48, 1.64, 18.46], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [661.64, 175.1, 687.85, 197.07], \"box_3d\": [1.38, 1.64, 3.51, 4.24, 1.54, 47.44], \"ry\": -1.48}]\n```" - } - ] - }, - { - "id": 1203, - "data_source": "Mono3DRefer", - "scene_name": "006975", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [552.71, 182.05, 603.27, 224.41], \"box_3d\": [1.38, 1.51, 4.1, -1.26, 1.72, 26.01], \"ry\": -1.69}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [590.5, 172.76, 706.49, 269.05], \"box_3d\": [1.6, 1.66, 3.55, 0.57, 1.62, 13.94], \"ry\": -1.68}, {\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [224.09, 181.85, 270.55, 202.78], \"box_3d\": [1.47, 1.62, 4.32, -26.87, 2.15, 53.56], \"ry\": 1.48}]\n```" - } - ] - }, - { - "id": 1204, - "data_source": "Mono3DRefer", - "scene_name": "006976", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.96, \"box_2d\": [446.53, 178.22, 513.66, 204.38], \"box_3d\": [1.49, 1.64, 3.7, -7.74, 1.83, 43.05], \"ry\": 3.14}]\n```" - } - ] - }, - { - "id": 1205, - "data_source": "Mono3DRefer", - "scene_name": "006982", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [0.0, 154.17, 47.37, 345.39], \"box_3d\": [2.01, 1.89, 4.91, -8.42, 1.85, 9.15], \"ry\": 0.9}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [226.27, 188.13, 294.7, 236.16], \"box_3d\": [1.52, 1.51, 3.1, -12.19, 2.07, 25.0], \"ry\": -2.23}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [196.89, 189.22, 259.52, 231.3], \"box_3d\": [1.53, 1.58, 3.53, -15.25, 2.19, 28.66], \"ry\": -2.24}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [173.69, 189.21, 209.19, 218.18], \"box_3d\": [1.46, 1.6, 3.71, -22.9, 2.37, 39.43], \"ry\": -2.1}, {\"category\": \"Car\", \"angle\": -1.8, \"box_2d\": [257.76, 187.34, 355.08, 251.34], \"box_3d\": [1.57, 1.72, 3.52, -8.49, 1.98, 19.92], \"ry\": -2.2}]\n```" - } - ] - }, - { - "id": 1206, - "data_source": "Mono3DRefer", - "scene_name": "006988", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [174.69, 177.97, 462.32, 374.0], \"box_3d\": [1.65, 1.75, 3.39, -2.62, 1.71, 7.22], \"ry\": 1.48}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [414.14, 180.5, 529.98, 270.95], \"box_3d\": [1.57, 1.54, 4.01, -2.69, 1.73, 14.65], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 1207, - "data_source": "Mono3DRefer", - "scene_name": "006990", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.27, \"box_2d\": [202.94, 188.19, 469.61, 340.34], \"box_3d\": [1.33, 1.71, 4.09, -2.96, 1.55, 8.67], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [652.68, 165.97, 724.63, 218.92], \"box_3d\": [1.49, 1.56, 4.34, 2.33, 1.33, 22.71], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": -1.4, \"box_2d\": [427.99, 171.41, 521.32, 238.56], \"box_3d\": [1.42, 1.53, 4.12, -3.2, 1.42, 17.54], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.83, \"box_2d\": [710.02, 163.78, 906.93, 304.68], \"box_3d\": [1.59, 1.59, 3.89, 2.5, 1.52, 10.21], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [627.47, 166.29, 659.72, 192.96], \"box_3d\": [1.38, 1.64, 3.51, 1.84, 1.06, 39.42], \"ry\": -1.49}]\n```" - } - ] - }, - { - "id": 1208, - "data_source": "Mono3DRefer", - "scene_name": "007004", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (707.05, 707.05), (cx, cy) = (604.08, 180.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.56, \"box_2d\": [371.3, 160.33, 433.56, 217.12], \"box_3d\": [1.5, 1.59, 3.6, -5.97, 0.96, 20.74], \"ry\": 1.29}, {\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [401.03, 165.53, 446.32, 200.4], \"box_3d\": [1.36, 1.57, 4.06, -7.62, 0.77, 29.93], \"ry\": 1.39}]\n```" - } - ] - }, - { - "id": 1209, - "data_source": "Mono3DRefer", - "scene_name": "007009", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.58, \"box_2d\": [582.93, 173.46, 622.7, 208.95], \"box_3d\": [1.41, 1.59, 4.47, -0.33, 1.47, 31.32], \"ry\": 1.57}]\n```" - } - ] - }, - { - "id": 1210, - "data_source": "Mono3DRefer", - "scene_name": "007014", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [373.84, 179.73, 462.28, 241.62], \"box_3d\": [1.74, 1.65, 4.11, -5.9, 1.96, 22.5], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [552.94, 179.13, 581.72, 199.34], \"box_3d\": [1.41, 1.76, 4.4, -3.12, 1.88, 53.09], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [425.96, 181.56, 489.5, 225.2], \"box_3d\": [1.61, 1.64, 4.0, -6.07, 1.97, 28.99], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [476.59, 182.72, 518.65, 210.61], \"box_3d\": [1.43, 1.7, 3.74, -6.16, 1.99, 39.68], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [639.84, 175.59, 662.02, 194.62], \"box_3d\": [1.55, 1.66, 4.15, 3.43, 1.8, 61.34], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 2.99, \"box_2d\": [695.39, 169.85, 761.18, 190.48], \"box_3d\": [1.47, 1.7, 4.65, 8.82, 1.29, 54.12], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": 2.17, \"box_2d\": [0.0, 207.54, 242.73, 373.33], \"box_3d\": [1.42, 1.54, 3.5, -6.04, 1.9, 8.52], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [724.79, 176.67, 894.42, 286.4], \"box_3d\": [1.51, 1.75, 3.78, 3.04, 1.6, 11.95], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.97, \"box_2d\": [215.54, 198.23, 382.79, 285.37], \"box_3d\": [1.36, 1.51, 4.02, -5.94, 1.92, 14.22], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 1211, - "data_source": "Mono3DRefer", - "scene_name": "007026", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [0.0, 202.65, 158.86, 308.72], \"box_3d\": [1.43, 1.74, 3.63, -9.12, 2.0, 12.44], \"ry\": 1.14}, {\"category\": \"Car\", \"angle\": 1.25, \"box_2d\": [1119.4, 150.91, 1241.0, 251.55], \"box_3d\": [1.52, 1.55, 3.91, 10.41, 1.19, 12.6], \"ry\": 1.93}, {\"category\": \"Car\", \"angle\": 1.35, \"box_2d\": [958.92, 159.6, 1117.64, 252.21], \"box_3d\": [1.62, 1.84, 4.1, 8.51, 1.41, 14.81], \"ry\": 1.86}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [413.09, 182.35, 468.77, 220.01], \"box_3d\": [1.38, 1.73, 3.97, -6.89, 1.78, 29.0], \"ry\": -1.91}, {\"category\": \"Car\", \"angle\": -2.87, \"box_2d\": [584.51, 165.18, 682.88, 207.75], \"box_3d\": [1.74, 1.58, 3.9, 0.99, 1.46, 30.82], \"ry\": -2.83}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [220.2, 184.38, 263.86, 214.76], \"box_3d\": [1.58, 1.68, 3.76, -20.48, 2.24, 40.21], \"ry\": 1.23}, {\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [377.65, 175.08, 401.15, 197.09], \"box_3d\": [1.77, 1.77, 5.1, -18.79, 2.0, 61.28], \"ry\": -1.89}, {\"category\": \"Car\", \"angle\": -1.9, \"box_2d\": [507.77, 190.09, 647.54, 270.48], \"box_3d\": [1.35, 1.64, 3.76, -0.84, 1.73, 14.64], \"ry\": -1.96}]\n```" - } - ] - }, - { - "id": 1212, - "data_source": "Mono3DRefer", - "scene_name": "007034", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [161.43, 220.39, 409.56, 368.42], \"box_3d\": [1.27, 1.54, 3.58, -3.81, 1.97, 9.04], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [357.43, 185.03, 482.16, 276.19], \"box_3d\": [1.63, 1.65, 3.91, -3.89, 1.91, 15.22], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [612.68, 181.34, 652.57, 217.81], \"box_3d\": [1.46, 1.53, 3.17, 0.89, 1.83, 30.87], \"ry\": -1.59}]\n```" - } - ] - }, - { - "id": 1213, - "data_source": "Mono3DRefer", - "scene_name": "007037", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [509.59, 180.41, 557.28, 214.91], \"box_3d\": [1.46, 1.75, 4.23, -3.47, 1.82, 33.08], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [207.04, 198.14, 422.89, 336.58], \"box_3d\": [1.39, 1.48, 3.76, -3.73, 1.78, 9.7], \"ry\": 1.53}]\n```" - } - ] - }, - { - "id": 1214, - "data_source": "Mono3DRefer", - "scene_name": "007043", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.35, \"box_2d\": [674.11, 180.54, 761.99, 241.43], \"box_3d\": [1.52, 1.55, 3.91, 2.87, 1.74, 20.17], \"ry\": 1.49}, {\"category\": \"Car\", \"angle\": 1.58, \"box_2d\": [591.26, 178.45, 659.9, 238.82], \"box_3d\": [1.62, 1.84, 4.1, 0.43, 1.8, 21.56], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": -2.9, \"box_2d\": [275.1, 180.49, 373.77, 220.88], \"box_3d\": [1.74, 1.58, 3.9, -12.96, 2.11, 32.83], \"ry\": 3.01}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [0.0, 188.11, 28.11, 213.8], \"box_3d\": [1.77, 1.77, 5.1, -44.03, 2.91, 52.9], \"ry\": -2.31}, {\"category\": \"Car\", \"angle\": -1.96, \"box_2d\": [208.8, 200.31, 344.03, 266.9], \"box_3d\": [1.35, 1.64, 3.76, -8.16, 2.05, 17.41], \"ry\": -2.4}]\n```" - } - ] - }, - { - "id": 1215, - "data_source": "Mono3DRefer", - "scene_name": "007047", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [662.85, 179.87, 705.94, 211.85], \"box_3d\": [1.41, 1.65, 4.28, 3.46, 1.76, 34.44], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 0.19, \"box_2d\": [414.4, 179.41, 534.39, 217.81], \"box_3d\": [1.44, 1.8, 4.44, -5.38, 1.73, 28.82], \"ry\": 0.0}, {\"category\": \"Car\", \"angle\": 0.3, \"box_2d\": [357.7, 181.17, 505.15, 228.51], \"box_3d\": [1.5, 1.8, 4.5, 0.0, 1.81, 24.4], \"ry\": 0.06}, {\"category\": \"Car\", \"angle\": 0.34, \"box_2d\": [339.9, 178.44, 500.15, 233.26], \"box_3d\": [1.53, 1.8, 4.22, 0.41, 1.72, 21.38], \"ry\": 0.09}, {\"category\": \"Car\", \"angle\": 0.38, \"box_2d\": [318.19, 182.3, 485.11, 242.75], \"box_3d\": [1.5, 1.46, 3.95, -5.46, 1.77, 18.97], \"ry\": 0.1}, {\"category\": \"Car\", \"angle\": 0.34, \"box_2d\": [265.28, 180.44, 448.19, 252.59], \"box_3d\": [1.53, 1.64, 3.6, -5.75, 1.73, 16.53], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": -2.62, \"box_2d\": [54.24, 182.88, 325.48, 289.75], \"box_3d\": [1.53, 1.51, 3.38, -6.5, 1.7, 11.31], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": -2.57, \"box_2d\": [0.0, 193.76, 292.3, 325.75], \"box_3d\": [1.41, 1.63, 3.75, -6.19, 1.71, 8.87], \"ry\": 3.12}]\n```" - } - ] - }, - { - "id": 1216, - "data_source": "Mono3DRefer", - "scene_name": "007058", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [378.97, 183.85, 471.07, 237.74], \"box_3d\": [1.51, 1.81, 4.07, -5.76, 1.88, 22.82], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 1217, - "data_source": "Mono3DRefer", - "scene_name": "007060", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.56, \"box_2d\": [539.8, 184.59, 791.54, 292.9], \"box_3d\": [1.41, 1.62, 3.6, 0.85, 1.62, 11.27], \"ry\": -0.49}, {\"category\": \"Car\", \"angle\": -0.47, \"box_2d\": [440.45, 168.6, 695.64, 278.25], \"box_3d\": [1.77, 1.96, 4.39, -0.79, 1.71, 13.65], \"ry\": -0.53}, {\"category\": \"Car\", \"angle\": 1.42, \"box_2d\": [295.85, 186.38, 349.84, 220.03], \"box_3d\": [1.37, 1.6, 4.5, -12.99, 1.99, 32.39], \"ry\": 1.04}, {\"category\": \"Car\", \"angle\": -0.38, \"box_2d\": [434.19, 187.61, 627.2, 260.38], \"box_3d\": [1.4, 1.51, 3.98, -1.77, 1.76, 16.05], \"ry\": -0.49}]\n```" - } - ] - }, - { - "id": 1218, - "data_source": "Mono3DRefer", - "scene_name": "007063", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [710.59, 184.74, 834.69, 275.52], \"box_3d\": [1.51, 1.56, 3.81, 3.0, 1.76, 14.2], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [681.88, 179.77, 770.96, 245.36], \"box_3d\": [1.57, 1.68, 3.83, 2.98, 1.77, 19.44], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [514.93, 178.12, 552.89, 215.41], \"box_3d\": [1.52, 1.5, 3.85, -3.31, 1.76, 31.6], \"ry\": 1.5}]\n```" - } - ] - }, - { - "id": 1219, - "data_source": "Mono3DRefer", - "scene_name": "007070", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [602.43, 180.87, 668.27, 234.46], \"box_3d\": [1.36, 1.69, 3.38, 0.61, 1.6, 20.33], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.55, \"box_2d\": [990.56, 168.33, 1118.48, 205.24], \"box_3d\": [1.45, 1.67, 4.32, 18.4, 1.29, 30.02], \"ry\": 3.1}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [393.46, 182.52, 439.91, 210.65], \"box_3d\": [1.59, 1.72, 3.86, -11.6, 2.19, 43.42], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [585.58, 176.89, 627.76, 215.38], \"box_3d\": [1.52, 1.67, 3.61, -0.16, 1.7, 30.55], \"ry\": -1.57}]\n```" - } - ] - }, - { - "id": 1220, - "data_source": "Mono3DRefer", - "scene_name": "007076", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [400.75, 175.56, 499.3, 243.96], \"box_3d\": [1.5, 1.62, 4.19, -3.92, 1.59, 18.12], \"ry\": 1.55}]\n```" - } - ] - }, - { - "id": 1221, - "data_source": "Mono3DRefer", - "scene_name": "007082", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [647.05, 173.15, 715.63, 222.32], \"box_3d\": [1.41, 1.58, 4.36, 2.15, 1.45, 23.11], \"ry\": -1.61}]\n```" - } - ] - }, - { - "id": 1222, - "data_source": "Mono3DRefer", - "scene_name": "007090", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.36, \"box_2d\": [321.76, 185.1, 451.62, 222.61], \"box_3d\": [1.36, 1.68, 4.49, -8.62, 1.86, 27.94], \"ry\": 0.07}]\n```" - } - ] - }, - { - "id": 1223, - "data_source": "Mono3DRefer", - "scene_name": "007091", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.32, \"box_2d\": [740.73, 177.82, 940.06, 322.53], \"box_3d\": [1.62, 1.58, 3.89, 2.9, 1.7, 10.07], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [420.67, 186.62, 533.0, 261.45], \"box_3d\": [1.45, 1.67, 4.43, -2.96, 1.8, 16.72], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [688.78, 183.98, 790.47, 255.97], \"box_3d\": [1.47, 1.59, 3.62, 2.87, 1.74, 16.88], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": 1.46, \"box_2d\": [669.56, 182.26, 740.81, 235.07], \"box_3d\": [1.44, 1.64, 4.5, 2.8, 1.75, 22.35], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [525.45, 181.69, 565.56, 217.32], \"box_3d\": [1.46, 1.51, 3.24, -2.81, 1.86, 31.6], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.5, \"box_2d\": [661.79, 181.65, 699.6, 211.06], \"box_3d\": [1.38, 1.59, 4.19, 3.49, 1.83, 36.45], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 1224, - "data_source": "Mono3DRefer", - "scene_name": "007093", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.08, \"box_2d\": [0.0, 206.62, 418.13, 374.0], \"box_3d\": [1.51, 1.63, 4.03, -3.21, 1.89, 6.42], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [371.98, 195.37, 531.85, 306.84], \"box_3d\": [1.57, 1.44, 3.6, -2.62, 2.01, 12.6], \"ry\": 1.75}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [514.7, 196.2, 613.34, 264.42], \"box_3d\": [1.47, 1.51, 3.5, -1.1, 2.1, 18.26], \"ry\": 1.81}, {\"category\": \"Car\", \"angle\": -1.34, \"box_2d\": [764.38, 176.37, 810.22, 206.39], \"box_3d\": [1.77, 1.75, 4.67, 11.08, 2.0, 44.96], \"ry\": -1.1}, {\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [805.88, 173.99, 838.14, 195.93], \"box_3d\": [1.6, 1.68, 3.94, 16.06, 1.7, 54.61], \"ry\": 2.03}]\n```" - } - ] - }, - { - "id": 1225, - "data_source": "Mono3DRefer", - "scene_name": "007095", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.26, \"box_2d\": [985.18, 189.76, 1241.0, 374.0], \"box_3d\": [1.66, 1.56, 3.42, 2.81, 1.75, 2.5], \"ry\": -1.48}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [500.14, 174.51, 571.73, 227.35], \"box_3d\": [1.45, 1.64, 4.21, -2.21, 1.53, 22.12], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.48, \"box_2d\": [694.99, 175.59, 739.41, 214.5], \"box_3d\": [1.62, 1.58, 3.89, 4.69, 1.76, 32.17], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [573.43, 175.64, 614.26, 204.28], \"box_3d\": [1.45, 1.67, 4.43, -0.84, 1.62, 39.13], \"ry\": 1.67}]\n```" - } - ] - }, - { - "id": 1226, - "data_source": "Mono3DRefer", - "scene_name": "007097", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.4, \"box_2d\": [453.56, 176.5, 546.48, 244.0], \"box_3d\": [1.47, 1.59, 4.03, -2.41, 1.36, 17.8], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.48, \"box_2d\": [519.98, 175.51, 567.11, 213.85], \"box_3d\": [1.41, 1.54, 3.36, -2.24, 1.19, 28.33], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [601.2, 169.86, 640.92, 206.07], \"box_3d\": [1.52, 1.67, 4.38, 0.92, 1.03, 32.75], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": -1.5, \"box_2d\": [565.62, 168.18, 597.05, 195.91], \"box_3d\": [1.56, 1.59, 3.65, -1.15, 0.81, 42.73], \"ry\": -1.53}]\n```" - } - ] - }, - { - "id": 1227, - "data_source": "Mono3DRefer", - "scene_name": "007123", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.86, \"box_2d\": [0.0, 218.09, 253.72, 374.0], \"box_3d\": [1.38, 1.66, 4.38, -5.03, 1.75, 4.88], \"ry\": 0.09}, {\"category\": \"Car\", \"angle\": -0.87, \"box_2d\": [1007.79, 137.15, 1241.0, 374.0], \"box_3d\": [1.74, 1.66, 4.37, 4.65, 1.58, 3.7], \"ry\": -0.01}, {\"category\": \"Car\", \"angle\": -2.71, \"box_2d\": [237.41, 187.55, 465.25, 261.96], \"box_3d\": [1.47, 1.67, 4.28, -5.48, 1.81, 15.43], \"ry\": -3.04}, {\"category\": \"Car\", \"angle\": -2.99, \"box_2d\": [339.93, 187.76, 505.31, 240.79], \"box_3d\": [1.38, 1.67, 4.37, -5.29, 1.84, 20.66], \"ry\": 3.04}, {\"category\": \"Car\", \"angle\": -0.44, \"box_2d\": [844.33, 176.17, 1115.5, 290.36], \"box_3d\": [1.54, 1.58, 3.2, 5.3, 1.61, 10.72], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": -0.56, \"box_2d\": [884.18, 176.73, 1241.0, 325.48], \"box_3d\": [1.51, 1.72, 4.03, 5.48, 1.57, 8.34], \"ry\": 0.01}]\n```" - } - ] - }, - { - "id": 1228, - "data_source": "Mono3DRefer", - "scene_name": "007140", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.32, \"box_2d\": [402.43, 172.74, 540.07, 270.03], \"box_3d\": [1.77, 1.75, 4.67, -2.81, 1.8, 15.52], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": -0.96, \"box_2d\": [1218.42, 136.54, 1241.0, 337.64], \"box_3d\": [1.47, 1.52, 4.43, 6.78, 1.17, 4.73], \"ry\": -0.03}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [518.73, 173.61, 579.46, 221.23], \"box_3d\": [1.6, 1.68, 3.94, -2.18, 1.65, 26.38], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [588.48, 172.62, 613.33, 192.1], \"box_3d\": [1.42, 1.66, 4.11, -0.68, 1.43, 55.3], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": -0.04, \"box_2d\": [949.04, 125.87, 1186.15, 230.18], \"box_3d\": [1.66, 1.15, 3.39, 7.89, 0.94, 12.78], \"ry\": 0.51}]\n```" - } - ] - }, - { - "id": 1229, - "data_source": "Mono3DRefer", - "scene_name": "007147", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (707.05, 707.05), (cx, cy) = (604.08, 180.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.88, \"box_2d\": [510.73, 157.13, 746.01, 251.35], \"box_3d\": [1.46, 1.53, 3.8, 0.32, 1.09, 12.29], \"ry\": 2.9}, {\"category\": \"Car\", \"angle\": -3.02, \"box_2d\": [195.37, 146.41, 471.42, 252.08], \"box_3d\": [1.53, 1.4, 3.94, -4.24, 1.02, 11.45], \"ry\": 2.92}, {\"category\": \"Car\", \"angle\": -2.6, \"box_2d\": [0.0, 148.79, 85.45, 256.99], \"box_3d\": [1.48, 1.55, 3.29, -9.75, 1.06, 10.14], \"ry\": 2.93}]\n```" - } - ] - }, - { - "id": 1230, - "data_source": "Mono3DRefer", - "scene_name": "007153", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [508.19, 186.84, 591.36, 239.55], \"box_3d\": [1.51, 1.63, 4.03, -1.89, 1.99, 23.42], \"ry\": 1.77}, {\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [564.76, 185.22, 627.34, 227.09], \"box_3d\": [1.57, 1.44, 3.6, -0.54, 2.1, 29.63], \"ry\": 1.89}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [619.51, 186.3, 670.03, 219.05], \"box_3d\": [1.47, 1.51, 3.5, 1.71, 2.14, 35.06], \"ry\": 1.91}, {\"category\": \"Car\", \"angle\": -1.26, \"box_2d\": [794.28, 170.48, 833.21, 192.57], \"box_3d\": [1.77, 1.75, 4.67, 17.02, 1.6, 60.17], \"ry\": -0.98}]\n```" - } - ] - }, - { - "id": 1231, - "data_source": "Mono3DRefer", - "scene_name": "007155", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [178.21, 181.46, 498.15, 374.0], \"box_3d\": [1.57, 1.54, 4.01, -2.36, 1.67, 7.29], \"ry\": 1.62}]\n```" - } - ] - }, - { - "id": 1232, - "data_source": "Mono3DRefer", - "scene_name": "007158", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [653.74, 162.26, 681.96, 185.85], \"box_3d\": [1.47, 1.55, 3.67, 3.74, 0.83, 47.36], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -0.32, \"box_2d\": [795.32, 162.17, 921.1, 211.61], \"box_3d\": [1.62, 1.58, 3.78, 8.48, 1.29, 24.94], \"ry\": 0.0}]\n```" - } - ] - }, - { - "id": 1233, - "data_source": "Mono3DRefer", - "scene_name": "007175", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.58, \"box_2d\": [545.53, 178.79, 573.49, 203.72], \"box_3d\": [1.41, 1.59, 4.47, -3.07, 1.79, 43.68], \"ry\": 1.51}]\n```" - } - ] - }, - { - "id": 1234, - "data_source": "Mono3DRefer", - "scene_name": "007177", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.03, \"box_2d\": [0.0, 184.72, 332.54, 374.0], \"box_3d\": [1.72, 1.8, 4.16, -4.58, 1.86, 7.34], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [388.6, 182.09, 477.6, 243.86], \"box_3d\": [1.62, 1.66, 3.81, -5.11, 1.9, 21.14], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [475.83, 181.26, 522.71, 212.0], \"box_3d\": [1.46, 1.67, 4.92, -5.68, 1.91, 37.33], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [630.84, 175.57, 654.71, 195.21], \"box_3d\": [1.51, 1.84, 4.1, 2.63, 1.75, 58.17], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [670.66, 172.47, 762.8, 245.88], \"box_3d\": [1.61, 1.67, 3.62, 2.48, 1.63, 17.79], \"ry\": -1.58}]\n```" - } - ] - }, - { - "id": 1235, - "data_source": "Mono3DRefer", - "scene_name": "007184", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [770.41, 167.35, 846.24, 232.58], \"box_3d\": [1.56, 1.65, 3.69, 5.14, 1.45, 19.2], \"ry\": -1.38}, {\"category\": \"Car\", \"angle\": 2.12, \"box_2d\": [351.24, 182.24, 504.69, 246.91], \"box_3d\": [1.42, 1.67, 4.53, -4.52, 1.68, 18.55], \"ry\": 1.88}, {\"category\": \"Car\", \"angle\": -0.42, \"box_2d\": [1031.94, 158.52, 1137.75, 193.41], \"box_3d\": [1.54, 1.63, 3.74, 21.94, 0.92, 33.5], \"ry\": 0.15}]\n```" - } - ] - }, - { - "id": 1236, - "data_source": "Mono3DRefer", - "scene_name": "007185", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.07, \"box_2d\": [14.94, 196.05, 413.8, 374.0], \"box_3d\": [1.43, 1.7, 3.95, -3.55, 1.72, 7.34], \"ry\": 1.64}, {\"category\": \"Car\", \"angle\": -1.84, \"box_2d\": [775.65, 182.89, 921.72, 280.65], \"box_3d\": [1.51, 1.6, 3.24, 4.07, 1.7, 12.93], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [742.95, 177.39, 836.96, 241.2], \"box_3d\": [1.45, 1.74, 4.1, 4.44, 1.58, 18.53], \"ry\": -1.47}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [448.0, 184.37, 543.14, 249.39], \"box_3d\": [1.54, 1.68, 3.79, -3.0, 1.87, 19.41], \"ry\": 1.65}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [504.62, 181.35, 564.98, 228.52], \"box_3d\": [1.49, 1.52, 3.35, -2.55, 1.8, 24.8], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [715.5, 180.15, 782.84, 228.22], \"box_3d\": [1.53, 1.65, 4.37, 4.75, 1.79, 25.36], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [535.46, 178.04, 584.6, 220.02], \"box_3d\": [1.74, 1.65, 4.18, -2.2, 1.99, 32.27], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [578.42, 178.34, 626.62, 210.14], \"box_3d\": [1.64, 1.45, 3.48, -0.39, 1.94, 39.07], \"ry\": 1.92}, {\"category\": \"Car\", \"angle\": 2.05, \"box_2d\": [664.79, 175.77, 714.07, 199.43], \"box_3d\": [1.54, 1.64, 4.08, 5.44, 1.75, 49.25], \"ry\": 2.16}]\n```" - } - ] - }, - { - "id": 1237, - "data_source": "Mono3DRefer", - "scene_name": "007190", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.94, \"box_2d\": [807.23, 176.66, 928.12, 237.24], \"box_3d\": [1.4, 1.55, 4.06, 6.57, 1.52, 18.98], \"ry\": -1.62}]\n```" - } - ] - }, - { - "id": 1238, - "data_source": "Mono3DRefer", - "scene_name": "007191", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [388.4, 165.29, 514.03, 266.35], \"box_3d\": [1.97, 1.87, 4.53, -3.47, 1.86, 16.48], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [465.31, 184.16, 533.37, 232.19], \"box_3d\": [1.47, 1.62, 3.79, -3.71, 1.87, 24.53], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [524.78, 176.74, 559.45, 208.76], \"box_3d\": [1.7, 1.53, 3.82, -3.81, 1.94, 40.65], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 1239, - "data_source": "Mono3DRefer", - "scene_name": "007198", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.52, \"box_2d\": [888.44, 171.07, 976.99, 257.7], \"box_3d\": [1.66, 1.56, 3.42, 7.03, 1.64, 15.67], \"ry\": -1.11}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [770.05, 175.69, 817.47, 206.72], \"box_3d\": [1.45, 1.64, 4.21, 9.19, 1.61, 35.98], \"ry\": 1.98}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [819.43, 174.63, 853.92, 196.1], \"box_3d\": [1.45, 1.67, 4.43, 16.07, 1.59, 51.1], \"ry\": 2.03}]\n```" - } - ] - }, - { - "id": 1240, - "data_source": "Mono3DRefer", - "scene_name": "007218", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [297.44, 185.31, 362.1, 235.01], \"box_3d\": [1.78, 1.72, 4.0, -11.03, 2.29, 28.55], \"ry\": 1.37}, {\"category\": \"Car\", \"angle\": 2.08, \"box_2d\": [0.0, 211.23, 47.75, 313.0], \"box_3d\": [1.47, 1.69, 3.74, -9.9, 2.1, 10.16], \"ry\": 1.32}, {\"category\": \"Car\", \"angle\": 0.25, \"box_2d\": [17.42, 201.73, 117.53, 229.68], \"box_3d\": [1.43, 1.55, 4.37, -31.24, 3.16, 41.65], \"ry\": -0.39}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [790.47, 178.49, 836.03, 204.57], \"box_3d\": [1.48, 1.61, 3.61, 12.17, 1.83, 43.18], \"ry\": 2.15}]\n```" - } - ] - }, - { - "id": 1241, - "data_source": "Mono3DRefer", - "scene_name": "007222", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [610.18, 178.41, 684.16, 243.6], \"box_3d\": [1.37, 1.56, 3.54, 0.79, 1.52, 17.11], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -1.44, \"box_2d\": [517.02, 168.84, 589.51, 226.16], \"box_3d\": [1.61, 1.69, 4.26, -1.71, 1.53, 22.58], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [629.23, 168.02, 679.46, 214.94], \"box_3d\": [1.53, 1.63, 4.5, 1.6, 1.4, 26.0], \"ry\": -1.49}, {\"category\": \"Car\", \"angle\": -0.98, \"box_2d\": [1159.5, 132.98, 1241.0, 201.04], \"box_3d\": [1.48, 1.6, 3.84, 15.12, 0.61, 16.89], \"ry\": -0.25}, {\"category\": \"Car\", \"angle\": -0.87, \"box_2d\": [1060.06, 136.58, 1218.96, 195.48], \"box_3d\": [1.55, 1.61, 3.64, 14.74, 0.59, 20.17], \"ry\": -0.25}]\n```" - } - ] - }, - { - "id": 1242, - "data_source": "Mono3DRefer", - "scene_name": "007246", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [610.31, 163.17, 648.43, 185.96], \"box_3d\": [1.43, 1.68, 4.19, 1.31, 0.84, 47.82], \"ry\": 1.81}, {\"category\": \"Car\", \"angle\": 2.03, \"box_2d\": [0.0, 181.98, 397.62, 374.0], \"box_3d\": [1.7, 1.82, 4.45, -3.84, 1.81, 7.56], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.48, \"box_2d\": [665.93, 163.22, 705.17, 198.89], \"box_3d\": [1.57, 1.56, 2.62, 3.5, 1.17, 33.4], \"ry\": -1.38}]\n```" - } - ] - }, - { - "id": 1243, - "data_source": "Mono3DRefer", - "scene_name": "007251", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.59, \"box_2d\": [433.95, 185.71, 479.57, 228.55], \"box_3d\": [1.65, 1.75, 3.39, -6.4, 2.21, 30.22], \"ry\": 1.39}]\n```" - } - ] - }, - { - "id": 1244, - "data_source": "Mono3DRefer", - "scene_name": "007257", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [981.82, 181.55, 1022.75, 215.19], \"box_3d\": [1.43, 1.55, 3.91, 18.08, 1.85, 33.26], \"ry\": -1.06}]\n```" - } - ] - }, - { - "id": 1245, - "data_source": "Mono3DRefer", - "scene_name": "007259", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.36, \"box_2d\": [547.0, 159.84, 621.55, 213.16], \"box_3d\": [1.77, 1.75, 4.67, -0.87, 1.37, 26.42], \"ry\": -1.39}, {\"category\": \"Car\", \"angle\": -2.05, \"box_2d\": [131.83, 206.07, 237.75, 272.98], \"box_3d\": [1.72, 1.39, 3.07, -12.36, 2.73, 20.87], \"ry\": -2.58}, {\"category\": \"Car\", \"angle\": -1.91, \"box_2d\": [0.0, 210.13, 98.54, 301.26], \"box_3d\": [1.46, 1.52, 2.75, -11.23, 2.21, 13.62], \"ry\": -2.59}, {\"category\": \"Car\", \"angle\": -0.57, \"box_2d\": [1009.87, 147.95, 1241.0, 247.18], \"box_3d\": [1.48, 1.63, 3.98, 8.91, 1.12, 11.69], \"ry\": 0.07}, {\"category\": \"Car\", \"angle\": -0.4, \"box_2d\": [868.59, 150.73, 1126.22, 227.72], \"box_3d\": [1.47, 1.52, 4.43, 7.79, 1.07, 14.89], \"ry\": 0.08}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [603.02, 157.28, 644.41, 190.3], \"box_3d\": [1.6, 1.68, 3.94, 0.73, 0.86, 37.23], \"ry\": 1.7}]\n```" - } - ] - }, - { - "id": 1246, - "data_source": "Mono3DRefer", - "scene_name": "007261", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [674.01, 179.01, 748.32, 237.65], \"box_3d\": [1.6, 1.53, 4.18, 2.93, 1.8, 21.95], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [659.13, 178.33, 716.76, 222.87], \"box_3d\": [1.56, 1.57, 3.98, 2.87, 1.79, 27.63], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [404.83, 182.62, 532.29, 276.72], \"box_3d\": [1.55, 1.68, 4.42, -2.66, 1.76, 14.33], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 1247, - "data_source": "Mono3DRefer", - "scene_name": "007267", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [696.62, 177.91, 846.09, 283.55], \"box_3d\": [1.49, 1.56, 4.34, 2.52, 1.59, 12.47], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.23, \"box_2d\": [117.65, 187.11, 437.14, 374.0], \"box_3d\": [1.42, 1.53, 4.12, -3.04, 1.6, 7.49], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [645.28, 177.85, 688.58, 214.17], \"box_3d\": [1.38, 1.64, 3.51, 2.25, 1.59, 29.37], \"ry\": -1.51}]\n```" - } - ] - }, - { - "id": 1248, - "data_source": "Mono3DRefer", - "scene_name": "007268", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [0.0, 202.87, 40.72, 254.42], \"box_3d\": [1.48, 1.65, 3.6, -19.6, 2.48, 22.58], \"ry\": 1.24}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [573.4, 185.2, 629.82, 220.35], \"box_3d\": [1.48, 1.61, 3.61, -0.36, 2.07, 32.98], \"ry\": 1.85}]\n```" - } - ] - }, - { - "id": 1249, - "data_source": "Mono3DRefer", - "scene_name": "007270", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.57, \"box_2d\": [469.38, 176.82, 513.34, 221.93], \"box_3d\": [1.68, 1.63, 3.9, -4.83, 1.86, 29.21], \"ry\": 1.41}, {\"category\": \"Car\", \"angle\": 1.45, \"box_2d\": [456.02, 181.67, 493.33, 207.72], \"box_3d\": [1.47, 1.7, 4.39, -8.23, 2.01, 43.47], \"ry\": 1.27}]\n```" - } - ] - }, - { - "id": 1250, - "data_source": "Mono3DRefer", - "scene_name": "007275", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [313.12, 191.96, 530.01, 344.58], \"box_3d\": [1.44, 1.67, 3.88, -2.14, 1.72, 9.1], \"ry\": 1.58}]\n```" - } - ] - }, - { - "id": 1251, - "data_source": "Mono3DRefer", - "scene_name": "007277", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.99, \"box_2d\": [790.84, 188.38, 1217.1, 374.0], \"box_3d\": [1.47, 1.55, 3.37, 2.63, 1.62, 5.8], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.19, \"box_2d\": [179.72, 182.04, 413.81, 321.05], \"box_3d\": [1.72, 1.8, 4.16, -4.54, 1.87, 11.13], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [429.88, 181.41, 499.82, 232.66], \"box_3d\": [1.62, 1.66, 3.81, -4.98, 1.93, 25.04], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [491.89, 182.21, 532.91, 210.09], \"box_3d\": [1.46, 1.67, 4.92, -5.52, 2.01, 41.09], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [634.07, 176.03, 656.42, 194.41], \"box_3d\": [1.51, 1.84, 4.1, 3.01, 1.8, 62.0], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [664.46, 173.89, 738.05, 233.13], \"box_3d\": [1.61, 1.67, 3.62, 2.61, 1.66, 21.62], \"ry\": -1.59}]\n```" - } - ] - }, - { - "id": 1252, - "data_source": "Mono3DRefer", - "scene_name": "007285", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [326.42, 191.75, 486.14, 298.44], \"box_3d\": [1.45, 1.53, 3.98, -3.31, 1.82, 12.34], \"ry\": 1.59}]\n```" - } - ] - }, - { - "id": 1253, - "data_source": "Mono3DRefer", - "scene_name": "007294", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [398.77, 187.68, 515.77, 278.39], \"box_3d\": [1.46, 1.6, 3.33, -2.81, 1.77, 13.69], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [686.14, 176.06, 809.76, 272.45], \"box_3d\": [1.64, 1.7, 3.54, 2.52, 1.72, 14.19], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [670.83, 182.69, 750.42, 242.08], \"box_3d\": [1.37, 1.51, 3.39, 2.47, 1.64, 18.64], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [531.45, 179.43, 569.09, 210.82], \"box_3d\": [1.55, 1.68, 4.42, -3.14, 1.91, 38.25], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 1254, - "data_source": "Mono3DRefer", - "scene_name": "007297", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [368.27, 179.33, 495.2, 279.79], \"box_3d\": [1.7, 1.53, 3.82, -3.4, 1.84, 14.25], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [489.97, 181.58, 544.87, 228.93], \"box_3d\": [1.5, 1.53, 3.38, -3.17, 1.81, 24.92], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [540.92, 175.01, 589.08, 204.94], \"box_3d\": [1.56, 1.56, 3.6, -2.45, 1.7, 39.58], \"ry\": 1.83}]\n```" - } - ] - }, - { - "id": 1255, - "data_source": "Mono3DRefer", - "scene_name": "007299", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.39, \"box_2d\": [0.0, 183.13, 211.25, 374.0], \"box_3d\": [1.73, 1.65, 3.96, -3.47, 1.78, 2.78], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": -1.99, \"box_2d\": [747.57, 140.38, 1100.83, 352.13], \"box_3d\": [1.78, 1.89, 4.37, 3.07, 1.53, 8.34], \"ry\": -1.65}, {\"category\": \"Car\", \"angle\": 1.96, \"box_2d\": [88.26, 180.73, 389.2, 374.0], \"box_3d\": [1.68, 1.51, 3.29, -3.54, 1.77, 7.45], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [322.67, 194.03, 459.28, 279.15], \"box_3d\": [1.39, 1.66, 4.04, -4.28, 1.86, 14.59], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [471.6, 177.14, 519.42, 218.54], \"box_3d\": [1.86, 1.67, 4.13, -5.47, 2.08, 34.71], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 1256, - "data_source": "Mono3DRefer", - "scene_name": "007317", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [11.96, 190.55, 131.99, 264.28], \"box_3d\": [1.78, 1.72, 4.0, -14.94, 2.31, 20.22], \"ry\": 1.19}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [90.66, 200.3, 174.24, 246.73], \"box_3d\": [1.48, 1.65, 3.6, -17.6, 2.55, 26.69], \"ry\": 1.27}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [644.68, 183.08, 695.94, 214.75], \"box_3d\": [1.48, 1.61, 3.61, 3.05, 2.01, 36.23], \"ry\": 1.94}]\n```" - } - ] - }, - { - "id": 1257, - "data_source": "Mono3DRefer", - "scene_name": "007325", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [635.46, 183.97, 678.15, 214.59], \"box_3d\": [1.32, 1.63, 4.1, 2.11, 1.86, 34.01], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 0.23, \"box_2d\": [378.56, 180.62, 500.16, 224.56], \"box_3d\": [1.58, 1.56, 4.25, -6.46, 1.9, 27.43], \"ry\": -0.0}, {\"category\": \"Car\", \"angle\": -2.83, \"box_2d\": [278.63, 184.95, 445.57, 246.03], \"box_3d\": [1.5, 1.58, 3.88, -6.52, 1.85, 19.12], \"ry\": 3.13}, {\"category\": \"Car\", \"angle\": -0.32, \"box_2d\": [742.07, 181.76, 905.18, 241.82], \"box_3d\": [1.5, 1.62, 3.88, 5.63, 1.76, 19.43], \"ry\": -0.04}, {\"category\": \"Car\", \"angle\": -2.79, \"box_2d\": [141.34, 170.72, 392.99, 269.89], \"box_3d\": [1.77, 1.68, 4.12, -6.59, 1.76, 14.13], \"ry\": 3.07}, {\"category\": \"Car\", \"angle\": 0.52, \"box_2d\": [55.13, 184.89, 375.37, 289.04], \"box_3d\": [1.51, 1.59, 4.28, -6.21, 1.73, 11.54], \"ry\": 0.04}, {\"category\": \"Car\", \"angle\": 0.6, \"box_2d\": [0.0, 191.59, 332.09, 323.56], \"box_3d\": [1.5, 1.49, 3.86, -5.8, 1.76, 9.16], \"ry\": 0.05}, {\"category\": \"Car\", \"angle\": -0.63, \"box_2d\": [925.43, 182.86, 1241.0, 313.4], \"box_3d\": [1.55, 1.58, 3.51, 6.31, 1.71, 9.71], \"ry\": -0.07}, {\"category\": \"Car\", \"angle\": -0.79, \"box_2d\": [1023.87, 193.53, 1241.0, 356.16], \"box_3d\": [1.4, 1.59, 3.49, 6.4, 1.63, 7.41], \"ry\": -0.09}]\n```" - } - ] - }, - { - "id": 1258, - "data_source": "Mono3DRefer", - "scene_name": "007331", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [590.81, 171.47, 622.3, 202.59], \"box_3d\": [1.65, 1.67, 3.64, -0.18, 1.6, 40.46], \"ry\": -1.55}]\n```" - } - ] - }, - { - "id": 1259, - "data_source": "Mono3DRefer", - "scene_name": "007358", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.22, \"box_2d\": [0.0, 206.79, 428.52, 374.0], \"box_3d\": [1.56, 1.56, 3.6, -2.34, 1.8, 3.57], \"ry\": 1.67}, {\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [212.98, 191.66, 423.53, 358.97], \"box_3d\": [1.7, 1.43, 3.2, -3.54, 1.97, 9.2], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.51, \"box_2d\": [502.23, 182.13, 526.77, 205.16], \"box_3d\": [1.68, 1.63, 3.9, -7.38, 2.4, 55.39], \"ry\": 1.38}]\n```" - } - ] - }, - { - "id": 1260, - "data_source": "Mono3DRefer", - "scene_name": "007370", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [363.97, 169.91, 471.82, 261.87], \"box_3d\": [1.68, 1.63, 3.9, -3.98, 1.66, 15.34], \"ry\": 1.46}, {\"category\": \"Car\", \"angle\": 1.52, \"box_2d\": [432.28, 176.86, 480.7, 215.96], \"box_3d\": [1.47, 1.7, 4.39, -6.41, 1.66, 29.67], \"ry\": 1.31}]\n```" - } - ] - }, - { - "id": 1261, - "data_source": "Mono3DRefer", - "scene_name": "007383", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [392.94, 173.22, 508.3, 265.09], \"box_3d\": [1.73, 1.65, 3.96, -3.35, 1.76, 15.69], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [682.46, 163.67, 778.87, 231.12], \"box_3d\": [1.78, 1.89, 4.37, 3.4, 1.56, 21.42], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [459.28, 174.09, 528.76, 239.04], \"box_3d\": [1.68, 1.51, 3.29, -3.25, 1.73, 20.46], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 2.03, \"box_2d\": [0.0, 160.03, 396.67, 374.0], \"box_3d\": [1.77, 1.71, 4.28, -3.49, 1.72, 6.62], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.82, \"box_2d\": [721.89, 178.59, 853.34, 257.92], \"box_3d\": [1.45, 1.71, 3.93, 3.55, 1.59, 15.33], \"ry\": -1.6}]\n```" - } - ] - }, - { - "id": 1262, - "data_source": "Mono3DRefer", - "scene_name": "007390", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.28, \"box_2d\": [0.0, 172.09, 121.09, 240.89], \"box_3d\": [1.78, 1.6, 4.29, -16.48, 1.79, 20.81], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 2.2, \"box_2d\": [110.09, 182.54, 218.35, 227.85], \"box_3d\": [1.5, 1.47, 3.62, -16.06, 1.87, 26.13], \"ry\": 1.65}]\n```" - } - ] - }, - { - "id": 1263, - "data_source": "Mono3DRefer", - "scene_name": "007406", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [569.63, 185.18, 653.77, 266.1], \"box_3d\": [1.56, 1.65, 3.69, -0.04, 1.86, 16.12], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 0.61, \"box_2d\": [0.0, 203.84, 26.88, 243.05], \"box_3d\": [1.43, 1.45, 3.5, -24.82, 2.67, 27.69], \"ry\": -0.11}]\n```" - } - ] - }, - { - "id": 1264, - "data_source": "Mono3DRefer", - "scene_name": "007413", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.39, \"box_2d\": [806.49, 166.38, 970.04, 230.39], \"box_3d\": [1.62, 1.58, 3.78, 7.39, 1.47, 19.48], \"ry\": -0.04}, {\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [620.65, 169.37, 653.02, 196.32], \"box_3d\": [1.47, 1.55, 3.67, 1.48, 1.31, 41.68], \"ry\": -1.62}]\n```" - } - ] - }, - { - "id": 1265, - "data_source": "Mono3DRefer", - "scene_name": "007414", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [1085.21, 174.48, 1241.0, 374.0], \"box_3d\": [1.51, 1.69, 4.04, 3.66, 1.51, 3.81], \"ry\": -1.06}, {\"category\": \"Car\", \"angle\": -1.98, \"box_2d\": [693.54, 182.08, 997.29, 361.76], \"box_3d\": [1.46, 1.58, 3.93, 2.18, 1.57, 8.01], \"ry\": -1.72}, {\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [0.0, 197.7, 275.08, 339.99], \"box_3d\": [1.5, 1.68, 4.03, -6.41, 1.91, 10.26], \"ry\": 1.46}, {\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [607.4, 171.22, 728.3, 260.59], \"box_3d\": [1.6, 1.7, 3.56, 1.01, 1.59, 14.89], \"ry\": -1.72}, {\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [572.05, 177.5, 662.59, 239.73], \"box_3d\": [1.52, 1.73, 3.89, 0.05, 1.67, 19.94], \"ry\": -1.77}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [287.25, 192.18, 370.56, 251.45], \"box_3d\": [1.44, 1.6, 3.79, -7.91, 2.03, 20.5], \"ry\": 1.37}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [326.11, 188.28, 387.35, 232.24], \"box_3d\": [1.42, 1.51, 3.71, -9.14, 2.01, 26.15], \"ry\": 1.39}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [554.11, 181.02, 611.83, 223.82], \"box_3d\": [1.37, 1.51, 3.49, -1.04, 1.66, 25.15], \"ry\": -1.75}]\n```" - } - ] - }, - { - "id": 1266, - "data_source": "Mono3DRefer", - "scene_name": "007430", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.1, \"box_2d\": [833.92, 195.6, 1241.0, 374.0], \"box_3d\": [1.49, 1.65, 4.06, 2.77, 1.68, 4.36], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -0.97, \"box_2d\": [0.0, 211.75, 356.51, 374.0], \"box_3d\": [1.48, 1.58, 3.98, -3.09, 1.8, 4.33], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.89, \"box_2d\": [735.6, 187.04, 957.45, 327.04], \"box_3d\": [1.44, 1.56, 3.76, 2.82, 1.66, 9.6], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [686.93, 180.01, 806.83, 266.74], \"box_3d\": [1.53, 1.56, 3.68, 2.61, 1.69, 14.76], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": -1.44, \"box_2d\": [442.69, 183.23, 531.84, 255.08], \"box_3d\": [1.48, 1.54, 4.05, -2.85, 1.75, 17.25], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [509.9, 178.27, 561.58, 219.64], \"box_3d\": [1.48, 1.67, 4.3, -2.86, 1.7, 28.2], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [640.87, 175.63, 675.51, 205.25], \"box_3d\": [1.46, 1.54, 3.62, 2.45, 1.62, 37.74], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [558.92, 174.18, 582.09, 195.11], \"box_3d\": [1.52, 1.57, 4.42, -3.0, 1.65, 55.04], \"ry\": 1.56}]\n```" - } - ] - }, - { - "id": 1267, - "data_source": "Mono3DRefer", - "scene_name": "007435", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.84, \"box_2d\": [740.28, 181.43, 850.02, 258.37], \"box_3d\": [1.59, 1.59, 3.34, 4.15, 1.8, 16.8], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [718.78, 184.15, 790.61, 238.12], \"box_3d\": [1.47, 1.6, 3.7, 4.28, 1.84, 22.0], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [703.36, 182.58, 759.58, 225.27], \"box_3d\": [1.49, 1.64, 3.52, 4.51, 1.88, 27.41], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [567.89, 182.11, 594.89, 202.49], \"box_3d\": [1.45, 1.6, 3.2, -2.13, 2.16, 53.95], \"ry\": 1.66}, {\"category\": \"Car\", \"angle\": -2.19, \"box_2d\": [903.69, 157.83, 1241.0, 374.0], \"box_3d\": [1.77, 1.77, 5.1, 3.87, 1.71, 4.96], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [469.88, 186.68, 534.41, 230.15], \"box_3d\": [1.47, 1.66, 3.76, -3.99, 2.01, 27.02], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": -1.86, \"box_2d\": [790.57, 193.43, 987.08, 309.95], \"box_3d\": [1.41, 1.66, 3.86, 4.0, 1.77, 11.14], \"ry\": -1.53}]\n```" - } - ] - }, - { - "id": 1268, - "data_source": "Mono3DRefer", - "scene_name": "007438", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.59, \"box_2d\": [553.48, 179.23, 598.53, 219.44], \"box_3d\": [1.41, 1.59, 4.47, -1.32, 1.67, 27.93], \"ry\": 1.54}]\n```" - } - ] - }, - { - "id": 1269, - "data_source": "Mono3DRefer", - "scene_name": "007469", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.62, \"box_2d\": [572.94, 158.83, 604.12, 184.67], \"box_3d\": [1.49, 1.68, 4.08, -0.73, 0.17, 43.92], \"ry\": 1.6}]\n```" - } - ] - }, - { - "id": 1270, - "data_source": "Mono3DRefer", - "scene_name": "007471", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [531.9, 174.39, 561.02, 201.7], \"box_3d\": [1.7, 1.72, 4.11, -4.16, 1.83, 47.48], \"ry\": 1.53}]\n```" - } - ] - }, - { - "id": 1271, - "data_source": "Mono3DRefer", - "scene_name": "007475", - "question_type": "mobj_3dDetect_json", - "conversations": [ - { - "from": "human", - "value": "Please identify all Car objects in this scene for KITTI 3D Object Detection Benchmark evaluation. Ignore objects that are too small or too far away to be recognized. Given the camera intrinsic parameters: (fx, fy) = (718.86, 718.86), (cx, cy) = (607.19, 185.22). Based on the provided information, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle on the image plane (-π to π). When the 2D object faces the image center, angle is 0; facing left is -π/2, facing right is π/2.\n- box_2d: 2D bounding box in image coordinates [x1, y1, x2, y2] in pixels.\n- box_3d: 3D object size and location [height, width, length, x, y, z], which position (x, y, z) is the coordinates of center point of the object's upper plane in the real-world coordinate system, where x is horizontal axis (positive right), y is vertical ground axis, and z is depth axis (distance from the camera).\n- ry: rotation around Y-axis (-π to π). When object front faces away from camera (-π/2), ry is 0 and ranges from 0 to -π (counter-clockwise). When object front faces towards camera (π/2), ry is 0 and ranges from 0 to π (counter-clockwise).\n\nNoting the response JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```" - }, - { - "from": "gpt", - "value": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.52, \"box_2d\": [674.88, 180.14, 737.11, 235.66], \"box_3d\": [1.6, 1.76, 3.84, 3.16, 1.45, 22.87], \"ry\": -1.38}, {\"category\": \"Car\", \"angle\": -1.11, \"box_2d\": [0.0, 196.03, 257.49, 339.49], \"box_3d\": [1.55, 1.63, 3.92, -6.67, 1.76, 10.28], \"ry\": -1.67}, {\"category\": \"Car\", \"angle\": -1.2, \"box_2d\": [155.47, 196.44, 271.13, 256.81], \"box_3d\": [1.51, 1.58, 4.12, -11.62, 1.9, 21.43], \"ry\": -1.69}, {\"category\": \"Car\", \"angle\": -1.3, \"box_2d\": [271.08, 193.61, 375.66, 255.15], \"box_3d\": [1.38, 1.56, 3.83, -7.46, 1.65, 19.13], \"ry\": -1.66}, {\"category\": \"Car\", \"angle\": -1.31, \"box_2d\": [763.24, 183.51, 798.35, 204.53], \"box_3d\": [1.39, 1.56, 3.45, 12.13, 1.29, 50.33], \"ry\": -1.08}]\n```" - } - ] - } -] \ No newline at end of file