diff --git "a/val/qa_mobj_3dDetect_json.json" "b/val/qa_mobj_3dDetect_json.json" deleted file mode 100644--- "a/val/qa_mobj_3dDetect_json.json" +++ /dev/null @@ -1,2405 +0,0 @@ -[ - { - "dataset": "Mono3DRefer", - "scene_name": "000003", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.55, \"box_2d\": [614.24, 181.78, 727.31, 284.77], \"box_3d\": [1.57, 1.73, 4.15, 1.0, 1.75, 13.22], \"ry\": 1.62}]\n```", - "options": null, - "id": 0 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000011", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [444.29, 171.04, 504.95, 225.82], \"box_3d\": [1.86, 1.57, 3.83, -4.95, 1.83, 26.64], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 2.42, \"box_2d\": [0.0, 217.12, 85.92, 374.0], \"box_3d\": [1.5, 1.46, 3.7, -5.12, 1.85, 4.13], \"ry\": 1.56}]\n```", - "options": null, - "id": 1 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000012", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.84, \"box_2d\": [662.2, 185.85, 690.21, 205.03], \"box_3d\": [1.48, 1.36, 3.51, 5.35, 2.56, 58.84], \"ry\": -1.75}]\n```", - "options": null, - "id": 2 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000014", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [626.59, 176.07, 654.95, 194.24], \"box_3d\": [1.44, 1.68, 4.39, 2.57, 1.72, 59.81], \"ry\": 1.77}]\n```", - "options": null, - "id": 3 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000027", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [590.88, 181.6, 607.07, 194.48], \"box_3d\": [1.26, 1.6, 3.56, -1.15, 2.16, 73.46], \"ry\": -1.59}]\n```", - "options": null, - "id": 4 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000034", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.5, \"box_2d\": [46.17, 196.15, 328.4, 286.09], \"box_3d\": [1.56, 1.71, 4.5, -8.15, 2.06, 14.12], \"ry\": -0.02}, {\"category\": \"Car\", \"angle\": -2.99, \"box_2d\": [466.96, 184.85, 522.12, 207.02], \"box_3d\": [1.44, 1.52, 3.5, -7.84, 2.28, 49.04], \"ry\": 3.13}]\n```", - "options": null, - "id": 5 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000038", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [592.08, 176.5, 637.55, 219.52], \"box_3d\": [1.56, 1.65, 3.69, 0.12, 1.72, 28.28], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.27, \"box_2d\": [895.35, 158.55, 1241.0, 374.0], \"box_3d\": [1.63, 1.61, 3.96, 4.38, 1.54, 6.69], \"ry\": -0.71}, {\"category\": \"Car\", \"angle\": -1.13, \"box_2d\": [817.68, 176.9, 1038.04, 302.31], \"box_3d\": [1.49, 1.63, 3.36, 4.61, 1.57, 10.4], \"ry\": -0.72}, {\"category\": \"Car\", \"angle\": -1.01, \"box_2d\": [749.97, 177.62, 940.82, 262.93], \"box_3d\": [1.41, 1.6, 4.09, 4.63, 1.52, 14.02], \"ry\": -0.7}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [489.07, 180.53, 521.27, 201.03], \"box_3d\": [1.37, 1.64, 3.92, -7.38, 1.92, 50.9], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [302.47, 188.9, 409.18, 250.55], \"box_3d\": [1.38, 1.47, 3.62, -6.49, 1.83, 18.66], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -0.87, \"box_2d\": [682.53, 173.56, 796.81, 220.86], \"box_3d\": [1.49, 1.59, 4.09, 4.47, 1.53, 24.75], \"ry\": -0.7}, {\"category\": \"Car\", \"angle\": -0.94, \"box_2d\": [659.69, 174.94, 745.56, 209.34], \"box_3d\": [1.4, 1.63, 4.05, 4.06, 1.5, 31.46], \"ry\": -0.81}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [434.34, 175.49, 472.4, 199.91], \"box_3d\": [1.67, 1.68, 4.49, -11.26, 1.88, 52.04], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [456.14, 177.0, 487.16, 197.96], \"box_3d\": [1.63, 1.58, 4.2, -11.21, 1.98, 58.61], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [412.71, 177.61, 437.08, 195.86], \"box_3d\": [1.47, 1.29, 2.95, -15.37, 1.87, 59.96], \"ry\": 1.56}]\n```", - "options": null, - "id": 6 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000044", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.11, \"box_2d\": [0.0, 203.44, 365.5, 374.0], \"box_3d\": [1.44, 1.62, 3.67, -3.27, 1.73, 5.22], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [378.01, 183.98, 507.17, 276.44], \"box_3d\": [1.51, 1.62, 3.68, -3.11, 1.75, 13.93], \"ry\": 1.57}]\n```", - "options": null, - "id": 7 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000063", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [875.15, 185.26, 975.98, 253.98], \"box_3d\": [1.52, 1.54, 4.5, 8.0, 1.87, 18.74], \"ry\": -1.35}, {\"category\": \"Car\", \"angle\": -2.55, \"box_2d\": [287.03, 172.92, 514.8, 285.76], \"box_3d\": [1.59, 1.45, 3.15, -3.3, 1.61, 11.32], \"ry\": -2.83}]\n```", - "options": null, - "id": 8 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000067", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.45, \"box_2d\": [671.83, 174.3, 702.61, 195.46], \"box_3d\": [1.32, 1.48, 4.44, 5.12, 1.44, 47.66], \"ry\": -1.34}, {\"category\": \"Car\", \"angle\": -1.45, \"box_2d\": [704.82, 171.69, 744.06, 200.12], \"box_3d\": [1.42, 1.56, 4.16, 6.09, 1.38, 38.29], \"ry\": -1.29}, {\"category\": \"Car\", \"angle\": -1.13, \"box_2d\": [427.15, 179.96, 600.34, 281.01], \"box_3d\": [1.45, 1.59, 3.44, -1.52, 1.59, 12.32], \"ry\": -1.26}, {\"category\": \"Car\", \"angle\": -0.79, \"box_2d\": [76.38, 181.99, 384.72, 306.19], \"box_3d\": [1.51, 1.61, 4.31, -5.57, 1.67, 11.21], \"ry\": -1.24}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [478.81, 182.19, 515.67, 207.52], \"box_3d\": [1.44, 1.35, 3.23, -6.77, 2.02, 43.4], \"ry\": 1.69}]\n```", - "options": null, - "id": 9 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000072", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.26, \"box_2d\": [0.0, 178.54, 117.72, 228.38], \"box_3d\": [1.57, 1.2, 5.24, -19.79, 1.78, 25.33], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [290.87, 187.34, 464.1, 288.52], \"box_3d\": [1.41, 1.69, 4.36, -3.87, 1.69, 12.67], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [590.87, 174.23, 629.98, 213.19], \"box_3d\": [1.62, 1.63, 4.5, -0.01, 1.71, 32.6], \"ry\": -1.57}]\n```", - "options": null, - "id": 10 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000077", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [574.29, 179.23, 616.86, 208.22], \"box_3d\": [1.5, 1.7, 3.77, -0.77, 1.86, 39.62], \"ry\": 1.72}]\n```", - "options": null, - "id": 11 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000092", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.4, \"box_2d\": [654.0, 176.92, 688.53, 201.62], \"box_3d\": [1.54, 1.63, 3.67, 4.03, 1.82, 47.24], \"ry\": -1.31}, {\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [359.95, 187.35, 456.4, 240.78], \"box_3d\": [1.56, 1.59, 3.69, -6.52, 2.06, 23.56], \"ry\": 1.74}, {\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [420.42, 175.29, 496.16, 228.83], \"box_3d\": [1.93, 1.59, 4.01, -5.86, 2.05, 28.17], \"ry\": 1.72}, {\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [686.98, 173.04, 733.38, 211.64], \"box_3d\": [1.58, 1.67, 3.84, 4.32, 1.62, 31.81], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [692.44, 163.91, 724.72, 207.04], \"box_3d\": [2.05, 1.5, 3.83, 5.0, 1.65, 36.49], \"ry\": -1.41}, {\"category\": \"Car\", \"angle\": 2.53, \"box_2d\": [1182.15, 149.81, 1241.0, 197.3], \"box_3d\": [1.39, 1.51, 3.49, 19.65, 0.71, 21.78], \"ry\": -3.03}]\n```", - "options": null, - "id": 12 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000094", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [660.56, 189.93, 747.57, 252.98], \"box_3d\": [1.41, 1.58, 4.36, 2.33, 1.9, 19.16], \"ry\": -1.61}]\n```", - "options": null, - "id": 13 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000105", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.62, \"box_2d\": [531.98, 177.67, 554.49, 195.35], \"box_3d\": [1.51, 1.81, 4.07, -5.96, 1.96, 64.63], \"ry\": 1.52}]\n```", - "options": null, - "id": 14 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000120", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.04, \"box_2d\": [57.05, 202.42, 394.68, 374.0], \"box_3d\": [1.42, 1.67, 3.8, -3.91, 1.82, 8.06], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 2.56, \"box_2d\": [0.0, 210.49, 119.82, 374.0], \"box_3d\": [1.55, 1.71, 4.5, -4.09, 1.78, 2.42], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [346.63, 185.82, 485.85, 278.21], \"box_3d\": [1.54, 1.63, 3.59, -3.68, 1.82, 14.17], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [455.26, 186.71, 532.19, 243.46], \"box_3d\": [1.42, 1.46, 3.63, -3.24, 1.84, 20.44], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [486.97, 186.14, 549.32, 232.01], \"box_3d\": [1.43, 1.66, 4.19, -3.19, 1.93, 25.45], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.5, \"box_2d\": [664.25, 178.14, 721.96, 227.83], \"box_3d\": [1.58, 1.71, 3.75, 2.91, 1.78, 24.95], \"ry\": -1.39}, {\"category\": \"Car\", \"angle\": -0.26, \"box_2d\": [785.76, 164.26, 985.11, 229.86], \"box_3d\": [1.49, 1.57, 4.26, 6.55, 1.33, 17.6], \"ry\": 0.09}, {\"category\": \"Car\", \"angle\": -1.22, \"box_2d\": [679.78, 180.23, 745.05, 220.33], \"box_3d\": [1.51, 1.58, 3.31, 4.17, 1.82, 29.24], \"ry\": -1.09}, {\"category\": \"Car\", \"angle\": -1.0, \"box_2d\": [716.11, 175.23, 783.6, 212.42], \"box_3d\": [1.61, 1.6, 3.11, 6.4, 1.73, 33.03], \"ry\": -0.82}]\n```", - "options": null, - "id": 15 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000133", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.24, \"box_2d\": [468.32, 183.17, 535.36, 203.05], \"box_3d\": [1.36, 1.68, 4.49, -7.78, 2.13, 51.93], \"ry\": 0.09}]\n```", - "options": null, - "id": 16 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000149", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [593.94, 176.33, 617.99, 199.21], \"box_3d\": [1.54, 1.63, 3.67, -0.31, 1.81, 51.09], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [509.06, 180.02, 558.1, 218.52], \"box_3d\": [1.45, 1.59, 3.59, -3.07, 1.75, 29.21], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.87, \"box_2d\": [745.98, 183.65, 941.59, 315.14], \"box_3d\": [1.48, 1.47, 3.7, 3.0, 1.65, 10.17], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [693.29, 175.68, 817.75, 269.33], \"box_3d\": [1.64, 1.64, 3.89, 2.76, 1.72, 14.72], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [679.66, 173.04, 756.77, 244.46], \"box_3d\": [1.69, 1.39, 2.65, 2.69, 1.72, 18.62], \"ry\": -1.65}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [653.18, 177.97, 707.11, 219.58], \"box_3d\": [1.5, 1.59, 3.79, 2.64, 1.72, 28.26], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.61, \"box_2d\": [637.08, 176.4, 666.4, 202.51], \"box_3d\": [1.49, 1.57, 4.07, 2.45, 1.72, 43.55], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [647.16, 176.2, 689.37, 212.64], \"box_3d\": [1.57, 1.59, 3.75, 2.61, 1.75, 33.38], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 0.19, \"box_2d\": [391.75, 183.02, 456.09, 206.9], \"box_3d\": [1.52, 1.51, 3.94, -12.41, 2.22, 48.16], \"ry\": -0.06}, {\"category\": \"Car\", \"angle\": 0.37, \"box_2d\": [304.53, 182.89, 377.06, 207.79], \"box_3d\": [1.58, 1.56, 4.24, -17.92, 2.28, 48.07], \"ry\": 0.01}]\n```", - "options": null, - "id": 17 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000158", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -3.0, \"box_2d\": [1047.18, 177.22, 1179.54, 218.76], \"box_3d\": [1.62, 1.72, 4.35, 21.09, 1.83, 30.37], \"ry\": -2.39}]\n```", - "options": null, - "id": 18 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000164", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.59, \"box_2d\": [10.43, 164.54, 103.22, 201.55], \"box_3d\": [1.72, 1.76, 3.14, -26.97, 1.35, 35.23], \"ry\": -1.24}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [554.53, 172.2, 616.39, 233.86], \"box_3d\": [1.62, 1.63, 4.5, -0.75, 1.64, 21.44], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 2.09, \"box_2d\": [153.97, 167.57, 206.03, 195.3], \"box_3d\": [1.7, 1.64, 2.85, -27.38, 1.39, 45.98], \"ry\": 1.55}]\n```", - "options": null, - "id": 19 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000169", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.96, \"box_2d\": [784.78, 177.12, 1124.16, 374.0], \"box_3d\": [1.62, 1.48, 3.36, 2.76, 1.68, 6.69], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 0.89, \"box_2d\": [0.0, 193.2, 137.39, 374.0], \"box_3d\": [1.55, 1.7, 4.09, -6.19, 1.74, 5.46], \"ry\": 0.07}, {\"category\": \"Car\", \"angle\": 0.63, \"box_2d\": [0.0, 179.47, 272.67, 340.49], \"box_3d\": [1.59, 1.63, 3.75, -6.07, 1.69, 8.05], \"ry\": -0.0}, {\"category\": \"Car\", \"angle\": -2.59, \"box_2d\": [0.0, 159.41, 290.35, 307.83], \"box_3d\": [1.95, 1.67, 3.35, -6.73, 1.79, 10.45], \"ry\": 3.14}, {\"category\": \"Car\", \"angle\": -2.68, \"box_2d\": [111.86, 185.74, 342.81, 272.27], \"box_3d\": [1.42, 1.55, 3.35, -6.84, 1.68, 13.05], \"ry\": 3.12}, {\"category\": \"Car\", \"angle\": 0.36, \"box_2d\": [247.71, 186.65, 435.61, 257.37], \"box_3d\": [1.43, 1.49, 3.56, -5.84, 1.76, 15.82], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": 0.31, \"box_2d\": [302.97, 181.41, 468.47, 247.11], \"box_3d\": [1.53, 1.6, 3.62, -5.56, 1.77, 18.03], \"ry\": 0.02}, {\"category\": \"Car\", \"angle\": 0.34, \"box_2d\": [335.68, 179.85, 462.74, 241.6], \"box_3d\": [1.66, 1.57, 3.11, -5.98, 1.88, 20.52], \"ry\": 0.06}, {\"category\": \"Car\", \"angle\": 0.3, \"box_2d\": [338.33, 174.43, 490.76, 230.08], \"box_3d\": [1.66, 1.71, 4.34, -6.15, 1.74, 22.84], \"ry\": 0.04}, {\"category\": \"Car\", \"angle\": -2.98, \"box_2d\": [394.53, 177.75, 500.3, 219.37], \"box_3d\": [1.51, 1.66, 3.73, -6.22, 1.72, 27.74], \"ry\": 3.09}, {\"category\": \"Car\", \"angle\": 0.16, \"box_2d\": [436.13, 170.1, 526.6, 212.16], \"box_3d\": [2.08, 1.8, 4.37, -6.63, 1.97, 37.37], \"ry\": -0.02}, {\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [677.68, 184.0, 746.78, 238.59], \"box_3d\": [1.38, 1.49, 3.38, 2.77, 1.71, 20.39], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -0.18, \"box_2d\": [731.11, 171.91, 853.44, 221.79], \"box_3d\": [1.67, 1.6, 3.94, 6.33, 1.67, 25.51], \"ry\": 0.06}, {\"category\": \"Car\", \"angle\": -0.17, \"box_2d\": [684.67, 172.21, 757.34, 204.76], \"box_3d\": [1.77, 1.69, 3.82, 6.23, 1.76, 40.91], \"ry\": -0.02}, {\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [633.13, 179.03, 659.07, 198.79], \"box_3d\": [1.41, 1.56, 4.08, 2.66, 1.89, 54.32], \"ry\": -1.61}]\n```", - "options": null, - "id": 20 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000171", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [570.81, 182.88, 644.9, 253.6], \"box_3d\": [1.55, 1.63, 3.32, -0.09, 1.81, 17.72], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.14, \"box_2d\": [58.49, 205.4, 232.86, 270.3], \"box_3d\": [1.4, 1.75, 4.75, -12.63, 2.4, 19.99], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 2.2, \"box_2d\": [0.0, 209.02, 132.78, 259.95], \"box_3d\": [1.46, 1.81, 4.16, -19.04, 2.82, 25.29], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [307.61, 195.31, 361.42, 227.03], \"box_3d\": [1.62, 1.85, 3.57, -15.51, 2.93, 40.72], \"ry\": 1.51}]\n```", - "options": null, - "id": 21 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000182", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.97, \"box_2d\": [78.0, 202.9, 387.08, 374.0], \"box_3d\": [1.39, 1.48, 3.76, -3.72, 1.78, 7.78], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [502.86, 181.97, 555.39, 219.47], \"box_3d\": [1.46, 1.75, 4.23, -3.41, 1.86, 30.77], \"ry\": 1.57}]\n```", - "options": null, - "id": 22 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000194", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [493.46, 180.94, 572.04, 236.25], \"box_3d\": [1.45, 1.62, 3.83, -2.2, 1.7, 21.07], \"ry\": 1.64}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [583.58, 175.32, 620.71, 204.77], \"box_3d\": [1.43, 1.66, 3.67, -0.39, 1.58, 37.37], \"ry\": 1.63}]\n```", - "options": null, - "id": 23 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000197", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.0, \"box_2d\": [164.61, 198.39, 347.27, 297.74], \"box_3d\": [1.42, 1.54, 3.5, -6.11, 1.92, 12.8], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [693.31, 175.02, 802.96, 251.28], \"box_3d\": [1.51, 1.75, 3.78, 2.95, 1.59, 16.37], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [314.77, 192.7, 428.07, 256.06], \"box_3d\": [1.36, 1.51, 4.02, -5.99, 1.91, 18.45], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [412.85, 178.76, 482.89, 230.16], \"box_3d\": [1.74, 1.65, 4.11, -5.93, 1.97, 26.68], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [450.43, 181.8, 503.32, 219.39], \"box_3d\": [1.61, 1.64, 4.0, -6.12, 2.04, 33.36], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -1.61, \"box_2d\": [636.27, 176.17, 656.63, 193.91], \"box_3d\": [1.55, 1.66, 4.15, 3.28, 1.87, 65.65], \"ry\": -1.56}]\n```", - "options": null, - "id": 24 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000198", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [144.26, 188.28, 363.98, 306.89], \"box_3d\": [1.41, 1.72, 4.41, -5.26, 1.69, 11.28], \"ry\": 1.45}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [335.12, 175.34, 437.06, 241.01], \"box_3d\": [1.62, 1.75, 4.3, -6.15, 1.71, 20.17], \"ry\": 1.51}]\n```", - "options": null, - "id": 25 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000204", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [599.29, 178.09, 628.15, 205.88], \"box_3d\": [1.66, 1.73, 3.05, 0.2, 2.0, 45.16], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.05, \"box_2d\": [192.2, 192.26, 290.92, 239.24], \"box_3d\": [1.44, 1.63, 3.43, -12.69, 2.15, 25.0], \"ry\": 1.58}]\n```", - "options": null, - "id": 26 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000216", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [568.38, 180.78, 610.26, 222.22], \"box_3d\": [1.65, 1.67, 3.64, -0.91, 2.0, 30.88], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 2.08, \"box_2d\": [81.08, 190.87, 276.31, 279.87], \"box_3d\": [1.45, 1.62, 3.77, -8.31, 1.85, 14.25], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [297.52, 188.99, 388.66, 243.05], \"box_3d\": [1.53, 1.46, 4.18, -8.57, 2.09, 23.4], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [395.02, 184.59, 444.87, 221.02], \"box_3d\": [1.59, 1.57, 3.39, -8.89, 2.15, 33.83], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [441.49, 187.53, 481.32, 211.29], \"box_3d\": [1.35, 1.63, 4.1, -9.18, 2.29, 44.72], \"ry\": 1.57}]\n```", - "options": null, - "id": 27 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000220", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [528.64, 177.83, 563.79, 197.25], \"box_3d\": [1.45, 1.59, 4.28, -4.96, 1.85, 56.46], \"ry\": 1.77}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [716.97, 175.86, 739.16, 195.51], \"box_3d\": [1.43, 1.59, 4.21, 9.03, 1.68, 55.2], \"ry\": -1.4}]\n```", - "options": null, - "id": 28 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000231", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [559.84, 184.78, 612.94, 227.58], \"box_3d\": [1.26, 1.6, 3.56, -0.82, 1.67, 23.63], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.46, \"box_2d\": [503.61, 181.84, 547.52, 213.63], \"box_3d\": [1.37, 1.63, 3.57, -3.9, 1.8, 33.52], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.99, \"box_2d\": [133.17, 182.4, 244.32, 236.25], \"box_3d\": [1.64, 1.93, 3.52, -14.0, 1.97, 24.16], \"ry\": 1.47}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [311.16, 184.18, 355.65, 210.61], \"box_3d\": [1.42, 1.58, 3.11, -15.8, 2.08, 41.26], \"ry\": 1.48}]\n```", - "options": null, - "id": 29 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000232", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.0, \"box_2d\": [207.17, 187.48, 340.16, 252.73], \"box_3d\": [1.61, 1.67, 4.56, -9.51, 2.06, 20.76], \"ry\": 1.58}]\n```", - "options": null, - "id": 30 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000233", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [527.44, 176.92, 574.06, 214.95], \"box_3d\": [1.53, 1.59, 3.64, -2.66, 1.73, 31.29], \"ry\": -1.77}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [432.49, 173.01, 465.3, 200.99], \"box_3d\": [1.63, 1.6, 3.73, -9.87, 1.67, 44.3], \"ry\": 1.45}]\n```", - "options": null, - "id": 31 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000235", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [589.53, 175.63, 616.49, 202.28], \"box_3d\": [1.65, 1.67, 3.64, -0.5, 1.85, 46.96], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 2.07, \"box_2d\": [60.87, 197.71, 258.76, 287.11], \"box_3d\": [1.44, 1.62, 4.02, -8.89, 2.0, 14.61], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [405.31, 181.1, 450.57, 214.56], \"box_3d\": [1.68, 1.67, 3.87, -9.7, 2.13, 38.59], \"ry\": 1.5}, {\"category\": \"Car\", \"angle\": -1.53, \"box_2d\": [552.27, 177.86, 572.91, 198.02], \"box_3d\": [1.61, 1.51, 4.5, -3.96, 2.05, 60.59], \"ry\": -1.59}]\n```", - "options": null, - "id": 32 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000246", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (718.86, 718.86), (cx, cy) = (607.19, 185.22). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [571.3, 177.37, 609.7, 212.39], \"box_3d\": [1.6, 1.76, 3.84, -0.9, 1.23, 35.13], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -0.93, \"box_2d\": [0.0, 208.25, 277.23, 375.0], \"box_3d\": [1.39, 1.5, 3.4, -3.78, 1.62, 4.78], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -0.82, \"box_2d\": [0.0, 193.79, 97.7, 311.45], \"box_3d\": [1.45, 1.54, 3.59, -8.48, 1.6, 8.98], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -0.91, \"box_2d\": [0.0, 187.92, 121.03, 266.41], \"box_3d\": [1.45, 1.8, 3.73, -12.49, 1.54, 15.21], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.27, \"box_2d\": [323.63, 182.96, 407.85, 230.63], \"box_3d\": [1.48, 1.64, 3.82, -8.24, 1.42, 24.71], \"ry\": -1.58}]\n```", - "options": null, - "id": 33 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000255", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [190.07, 192.94, 450.91, 346.85], \"box_3d\": [1.5, 1.7, 4.21, -3.5, 1.81, 9.53], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [535.35, 181.92, 572.86, 208.46], \"box_3d\": [1.48, 1.87, 4.42, -3.32, 2.04, 43.19], \"ry\": 1.58}]\n```", - "options": null, - "id": 34 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000260", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [581.38, 173.49, 604.08, 191.47], \"box_3d\": [1.26, 1.6, 3.56, -1.3, 1.33, 52.89], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [554.66, 172.93, 574.71, 189.14], \"box_3d\": [1.37, 1.63, 3.57, -3.99, 1.41, 63.68], \"ry\": -1.6}]\n```", - "options": null, - "id": 35 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000261", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.02, \"box_2d\": [278.03, 179.51, 334.17, 206.65], \"box_3d\": [1.58, 1.64, 3.88, -18.63, 2.0, 44.35], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.97, \"box_2d\": [315.16, 180.02, 366.01, 201.87], \"box_3d\": [1.43, 1.81, 4.16, -18.58, 1.93, 49.88], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [360.9, 177.24, 402.39, 196.61], \"box_3d\": [1.47, 1.67, 4.31, -18.15, 1.83, 57.47], \"ry\": 1.64}]\n```", - "options": null, - "id": 36 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000269", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [0.0, 152.03, 86.39, 351.21], \"box_3d\": [2.01, 1.89, 4.91, -7.87, 1.83, 9.1], \"ry\": 0.96}, {\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [53.86, 191.17, 118.05, 244.56], \"box_3d\": [1.65, 1.67, 3.81, -18.11, 2.31, 25.0], \"ry\": -2.16}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [272.75, 187.98, 337.13, 235.6], \"box_3d\": [1.52, 1.51, 3.1, -10.76, 2.07, 25.26], \"ry\": -2.16}, {\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [243.61, 188.82, 302.61, 230.57], \"box_3d\": [1.53, 1.58, 3.53, -13.58, 2.19, 28.92], \"ry\": -2.17}, {\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [223.23, 188.71, 257.16, 217.14], \"box_3d\": [1.46, 1.6, 3.71, -20.62, 2.36, 40.13], \"ry\": -2.06}, {\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [300.41, 186.75, 397.0, 250.73], \"box_3d\": [1.57, 1.72, 3.52, -7.33, 1.96, 19.93], \"ry\": -2.15}]\n```", - "options": null, - "id": 37 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000282", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [36.6, 200.84, 405.7, 374.0], \"box_3d\": [1.41, 1.59, 4.47, -3.36, 1.77, 7.22], \"ry\": 1.49}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [595.17, 178.83, 624.37, 202.42], \"box_3d\": [1.64, 1.69, 3.3, -0.07, 2.09, 52.46], \"ry\": -1.7}]\n```", - "options": null, - "id": 38 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000299", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.57, \"box_2d\": [593.72, 187.39, 613.12, 205.23], \"box_3d\": [1.41, 1.59, 4.47, -0.59, 2.69, 61.86], \"ry\": 1.56}]\n```", - "options": null, - "id": 39 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000318", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [18.98, 155.74, 415.63, 374.0], \"box_3d\": [1.91, 1.66, 4.22, -3.16, 1.82, 6.79], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [361.46, 177.53, 511.19, 302.13], \"box_3d\": [1.83, 1.69, 4.28, -2.91, 1.93, 12.82], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [532.52, 189.9, 599.93, 233.02], \"box_3d\": [1.51, 1.63, 4.03, -1.68, 2.22, 28.45], \"ry\": 1.78}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [579.61, 188.68, 630.28, 224.46], \"box_3d\": [1.57, 1.44, 3.6, -0.22, 2.36, 34.66], \"ry\": 1.86}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [626.67, 189.43, 671.36, 218.24], \"box_3d\": [1.47, 1.51, 3.5, 2.18, 2.42, 39.96], \"ry\": 1.92}]\n```", - "options": null, - "id": 40 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000325", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [316.71, 185.26, 449.47, 279.21], \"box_3d\": [1.47, 1.43, 3.11, -4.03, 1.71, 13.11], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 2.25, \"box_2d\": [0.0, 179.14, 108.62, 229.8], \"box_3d\": [1.71, 1.44, 4.06, -20.67, 1.96, 26.56], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 2.13, \"box_2d\": [108.58, 182.33, 198.02, 218.18], \"box_3d\": [1.53, 1.67, 3.88, -20.99, 1.98, 33.3], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 2.06, \"box_2d\": [196.6, 183.28, 264.35, 214.18], \"box_3d\": [1.53, 1.47, 4.05, -20.21, 2.11, 38.53], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [508.42, 181.34, 550.52, 212.26], \"box_3d\": [1.41, 1.48, 4.93, -3.95, 1.84, 35.77], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [554.46, 177.99, 573.05, 194.52], \"box_3d\": [1.58, 1.6, 4.11, -4.58, 2.11, 71.71], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [581.06, 177.42, 629.34, 225.53], \"box_3d\": [1.62, 1.63, 4.5, -0.19, 1.81, 26.83], \"ry\": -1.57}]\n```", - "options": null, - "id": 41 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000331", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.63, \"box_2d\": [0.0, 204.58, 277.77, 318.09], \"box_3d\": [1.42, 1.62, 3.5, -7.02, 1.94, 10.53], \"ry\": 3.07}, {\"category\": \"Car\", \"angle\": -2.88, \"box_2d\": [315.63, 189.85, 446.85, 238.82], \"box_3d\": [1.41, 1.54, 3.6, -7.08, 1.97, 22.43], \"ry\": 3.1}]\n```", - "options": null, - "id": 42 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000333", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (707.05, 707.05), (cx, cy) = (604.08, 180.51). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.86, \"box_2d\": [0.0, 201.57, 304.1, 369.0], \"box_3d\": [1.5, 1.78, 3.69, -3.16, 1.68, 3.36], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 0.87, \"box_2d\": [503.21, 169.49, 592.05, 206.89], \"box_3d\": [1.44, 1.64, 3.68, -2.42, 1.02, 29.12], \"ry\": 0.79}, {\"category\": \"Car\", \"angle\": 2.15, \"box_2d\": [0.0, 184.62, 123.26, 272.67], \"box_3d\": [1.63, 1.78, 4.54, -12.57, 1.75, 15.09], \"ry\": 1.47}]\n```", - "options": null, - "id": 43 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000338", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (707.05, 707.05), (cx, cy) = (604.08, 180.51). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.06, \"box_2d\": [822.39, 177.95, 1223.0, 369.0], \"box_3d\": [1.52, 1.64, 4.41, 3.65, 1.51, 6.97], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 2.23, \"box_2d\": [46.17, 169.53, 222.76, 248.41], \"box_3d\": [1.85, 1.68, 4.11, -12.27, 1.59, 18.77], \"ry\": 1.66}, {\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [453.92, 175.82, 494.0, 198.36], \"box_3d\": [1.49, 1.58, 3.84, -9.08, 1.18, 49.33], \"ry\": 1.72}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [700.62, 173.49, 768.4, 237.23], \"box_3d\": [1.54, 1.57, 3.52, 3.36, 1.37, 18.95], \"ry\": -1.44}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [684.23, 175.83, 747.85, 224.45], \"box_3d\": [1.5, 1.63, 3.57, 3.65, 1.36, 23.88], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [670.01, 176.47, 714.05, 214.03], \"box_3d\": [1.47, 1.59, 3.64, 3.61, 1.31, 29.9], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [662.81, 175.73, 703.62, 208.13], \"box_3d\": [1.49, 1.74, 4.09, 3.81, 1.27, 35.05], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": -1.61, \"box_2d\": [651.86, 175.27, 678.37, 198.37], \"box_3d\": [1.44, 1.58, 3.28, 3.93, 1.11, 46.43], \"ry\": -1.53}]\n```", - "options": null, - "id": 44 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000340", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [0.0, 201.28, 193.53, 336.11], \"box_3d\": [1.44, 1.58, 3.72, -7.15, 1.91, 10.33], \"ry\": 1.15}, {\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [263.56, 202.7, 302.97, 235.62], \"box_3d\": [1.39, 1.52, 3.56, -15.88, 2.91, 35.12], \"ry\": 1.21}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [163.47, 201.04, 266.81, 277.26], \"box_3d\": [1.46, 1.59, 3.35, -9.06, 2.17, 16.71], \"ry\": 1.23}]\n```", - "options": null, - "id": 45 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000347", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.96, \"box_2d\": [260.83, 191.13, 340.36, 232.37], \"box_3d\": [1.63, 1.78, 4.13, -13.65, 2.48, 32.02], \"ry\": 1.56}]\n```", - "options": null, - "id": 46 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000351", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (718.86, 718.86), (cx, cy) = (607.19, 185.22). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [572.57, 174.58, 622.77, 220.34], \"box_3d\": [1.6, 1.76, 3.84, -0.41, 1.22, 27.32], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.06, \"box_2d\": [100.05, 188.89, 276.08, 273.57], \"box_3d\": [1.47, 1.54, 3.51, -8.43, 1.58, 14.73], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.29, \"box_2d\": [315.05, 170.86, 457.84, 265.27], \"box_3d\": [1.78, 1.81, 4.32, -4.7, 1.5, 15.82], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.28, \"box_2d\": [337.86, 181.63, 416.87, 224.41], \"box_3d\": [1.45, 1.61, 4.3, -8.56, 1.33, 26.97], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.4, \"box_2d\": [449.36, 181.41, 508.49, 221.08], \"box_3d\": [1.45, 1.67, 3.88, -5.07, 1.31, 28.56], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.49, \"box_2d\": [526.46, 176.89, 554.38, 200.81], \"box_3d\": [1.63, 1.67, 3.92, -4.78, 1.05, 51.35], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.16, \"box_2d\": [242.95, 181.37, 326.77, 223.59], \"box_3d\": [1.58, 1.64, 3.96, -13.05, 1.44, 29.24], \"ry\": -1.58}]\n```", - "options": null, - "id": 47 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000358", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.02, \"box_2d\": [288.87, 179.61, 342.91, 205.91], \"box_3d\": [1.58, 1.64, 3.88, -18.58, 2.02, 45.69], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": 1.96, \"box_2d\": [323.85, 180.04, 372.82, 201.28], \"box_3d\": [1.43, 1.81, 4.16, -18.53, 1.95, 51.22], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [367.26, 177.32, 407.49, 196.25], \"box_3d\": [1.47, 1.67, 4.31, -18.09, 1.85, 58.77], \"ry\": 1.64}]\n```", - "options": null, - "id": 48 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000359", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [595.67, 174.11, 623.22, 200.64], \"box_3d\": [1.66, 1.73, 3.05, -0.06, 1.77, 47.24], \"ry\": -1.57}]\n```", - "options": null, - "id": 49 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000402", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [383.59, 188.9, 506.24, 261.26], \"box_3d\": [1.43, 1.75, 4.47, -3.79, 1.85, 17.17], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [527.13, 173.92, 557.45, 198.9], \"box_3d\": [1.39, 1.48, 3.76, -3.97, 1.48, 42.43], \"ry\": 1.55}]\n```", - "options": null, - "id": 50 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000411", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [669.26, 173.4, 701.63, 193.38], \"box_3d\": [1.62, 1.72, 4.35, 6.36, 1.7, 61.37], \"ry\": -1.71}, {\"category\": \"Car\", \"angle\": -2.22, \"box_2d\": [958.05, 197.21, 1241.0, 374.0], \"box_3d\": [1.42, 1.45, 3.2, 3.87, 1.63, 4.94], \"ry\": -1.59}]\n```", - "options": null, - "id": 51 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000412", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.96, \"box_2d\": [830.16, 156.08, 1241.0, 374.0], \"box_3d\": [1.61, 1.62, 3.78, 3.01, 1.55, 5.82], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [490.66, 167.2, 536.97, 211.61], \"box_3d\": [1.86, 1.57, 3.83, -4.29, 1.65, 32.4], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 2.05, \"box_2d\": [95.88, 192.42, 352.42, 334.28], \"box_3d\": [1.5, 1.46, 3.7, -4.99, 1.8, 9.84], \"ry\": 1.59}]\n```", - "options": null, - "id": 52 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000416", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [642.28, 179.07, 665.91, 196.85], \"box_3d\": [1.48, 1.36, 3.51, 3.77, 2.03, 62.43], \"ry\": -1.71}]\n```", - "options": null, - "id": 53 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000420", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.15, \"box_2d\": [0.0, 189.0, 351.2, 374.0], \"box_3d\": [1.55, 1.64, 3.7, -3.48, 1.7, 5.34], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -1.38, \"box_2d\": [479.52, 164.58, 599.67, 250.22], \"box_3d\": [1.63, 1.81, 4.54, -1.43, 1.51, 16.11], \"ry\": -1.47}, {\"category\": \"Car\", \"angle\": -1.52, \"box_2d\": [804.28, 168.64, 915.34, 270.42], \"box_3d\": [1.47, 1.54, 3.02, 4.19, 1.43, 12.08], \"ry\": -1.19}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [344.73, 180.38, 464.68, 251.8], \"box_3d\": [1.42, 1.58, 4.08, -4.58, 1.6, 16.53], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.08, \"box_2d\": [844.29, 160.91, 996.56, 233.02], \"box_3d\": [1.5, 1.52, 4.12, 7.31, 1.26, 16.89], \"ry\": -0.68}]\n```", - "options": null, - "id": 54 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000422", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [133.34, 182.71, 364.05, 323.19], \"box_3d\": [1.57, 1.67, 4.14, -4.77, 1.6, 10.34], \"ry\": 1.45}]\n```", - "options": null, - "id": 55 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000451", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [577.4, 188.26, 594.45, 201.66], \"box_3d\": [1.26, 1.6, 3.56, -2.41, 2.83, 72.46], \"ry\": -1.57}]\n```", - "options": null, - "id": 56 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000454", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -3.03, \"box_2d\": [872.5, 178.55, 960.25, 211.59], \"box_3d\": [1.42, 1.58, 3.48, 13.81, 1.7, 32.6], \"ry\": -2.63}, {\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [247.42, 188.77, 311.48, 213.58], \"box_3d\": [1.41, 1.85, 4.7, -20.83, 2.46, 45.63], \"ry\": 1.58}]\n```", - "options": null, - "id": 57 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000464", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.98, \"box_2d\": [224.63, 200.82, 449.61, 310.19], \"box_3d\": [1.37, 1.78, 4.65, -4.36, 1.93, 12.34], \"ry\": 1.65}, {\"category\": \"Car\", \"angle\": 2.25, \"box_2d\": [1140.51, 175.47, 1241.0, 222.49], \"box_3d\": [1.48, 1.51, 4.35, 20.3, 1.58, 24.22], \"ry\": 2.94}]\n```", - "options": null, - "id": 58 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000540", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [581.73, 179.51, 640.18, 227.03], \"box_3d\": [1.36, 1.69, 3.38, -0.02, 1.59, 22.68], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -2.17, \"box_2d\": [240.12, 184.86, 424.16, 265.28], \"box_3d\": [1.45, 1.61, 3.96, -5.87, 1.72, 14.96], \"ry\": -2.53}, {\"category\": \"Car\", \"angle\": -2.19, \"box_2d\": [319.41, 187.06, 464.24, 248.9], \"box_3d\": [1.36, 1.57, 3.73, -5.52, 1.73, 17.92], \"ry\": -2.48}, {\"category\": \"Car\", \"angle\": 1.02, \"box_2d\": [390.98, 179.39, 482.72, 224.32], \"box_3d\": [1.48, 1.57, 3.48, -6.25, 1.73, 25.72], \"ry\": 0.78}, {\"category\": \"Car\", \"angle\": -1.46, \"box_2d\": [499.0, 185.74, 565.16, 228.86], \"box_3d\": [1.25, 1.74, 3.66, -2.49, 1.69, 23.49], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [675.95, 180.05, 765.47, 242.02], \"box_3d\": [1.46, 1.72, 3.95, 2.78, 1.66, 19.2], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [663.14, 183.28, 728.68, 227.67], \"box_3d\": [1.32, 1.7, 4.04, 2.73, 1.69, 24.05], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [657.71, 173.92, 704.53, 214.14], \"box_3d\": [1.56, 1.48, 3.36, 2.86, 1.63, 29.93], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [632.59, 177.89, 660.58, 199.66], \"box_3d\": [1.38, 1.55, 3.6, 2.38, 1.73, 48.07], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 2.12, \"box_2d\": [48.1, 193.81, 234.47, 273.51], \"box_3d\": [1.35, 1.59, 3.41, -9.33, 1.81, 14.63], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [590.29, 176.78, 629.25, 212.34], \"box_3d\": [1.52, 1.67, 3.61, -0.05, 1.71, 32.93], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 2.13, \"box_2d\": [46.51, 185.37, 168.1, 235.1], \"box_3d\": [1.32, 1.43, 3.25, -14.81, 1.71, 21.38], \"ry\": 1.53}]\n```", - "options": null, - "id": 59 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000577", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [354.82, 186.07, 523.76, 310.21], \"box_3d\": [1.56, 1.55, 3.64, -2.46, 1.78, 11.09], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [472.84, 168.44, 574.13, 264.05], \"box_3d\": [1.91, 1.66, 4.22, -1.89, 1.85, 16.59], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [538.72, 174.9, 611.53, 239.72], \"box_3d\": [1.83, 1.69, 4.28, -1.03, 1.91, 22.58], \"ry\": 1.66}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [619.97, 184.4, 668.18, 215.5], \"box_3d\": [1.51, 1.63, 4.03, 1.82, 2.14, 37.99], \"ry\": 1.85}]\n```", - "options": null, - "id": 60 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000619", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [486.63, 171.24, 540.81, 211.21], \"box_3d\": [1.39, 1.52, 3.56, -3.6, 1.36, 27.2], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.98, \"box_2d\": [90.61, 189.93, 399.51, 374.0], \"box_3d\": [1.46, 1.59, 3.35, -3.53, 1.67, 7.58], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 3.1, \"box_2d\": [184.71, 169.18, 362.24, 228.99], \"box_3d\": [1.53, 1.54, 4.35, -9.37, 1.44, 20.19], \"ry\": 2.67}, {\"category\": \"Car\", \"angle\": 2.91, \"box_2d\": [462.11, 165.53, 526.95, 188.3], \"box_3d\": [1.52, 1.41, 4.22, -8.02, 1.04, 50.04], \"ry\": 2.75}]\n```", - "options": null, - "id": 61 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000641", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.39, \"box_2d\": [569.09, 168.49, 635.71, 218.07], \"box_3d\": [1.77, 1.75, 4.67, -0.23, 1.64, 28.21], \"ry\": -1.39}, {\"category\": \"Car\", \"angle\": -2.08, \"box_2d\": [190.06, 211.67, 286.3, 273.1], \"box_3d\": [1.72, 1.39, 3.07, -11.87, 3.01, 22.91], \"ry\": -2.55}, {\"category\": \"Car\", \"angle\": -1.97, \"box_2d\": [38.07, 214.96, 176.69, 294.98], \"box_3d\": [1.46, 1.52, 2.75, -10.91, 2.42, 15.57], \"ry\": -2.57}, {\"category\": \"Car\", \"angle\": -0.51, \"box_2d\": [982.73, 153.48, 1241.0, 241.17], \"box_3d\": [1.48, 1.63, 3.98, 9.22, 1.18, 13.22], \"ry\": 0.09}, {\"category\": \"Car\", \"angle\": -0.36, \"box_2d\": [859.2, 158.51, 1089.82, 227.7], \"box_3d\": [1.47, 1.52, 4.43, 8.12, 1.19, 16.45], \"ry\": 0.09}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [620.64, 160.72, 660.31, 192.23], \"box_3d\": [1.6, 1.68, 3.94, 1.67, 1.0, 38.9], \"ry\": 1.73}, {\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [666.98, 163.66, 687.71, 179.57], \"box_3d\": [1.42, 1.66, 4.11, 6.31, 0.62, 67.39], \"ry\": 1.73}]\n```", - "options": null, - "id": 62 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000666", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.71, \"box_2d\": [173.78, 167.18, 400.59, 270.05], \"box_3d\": [1.86, 1.69, 3.69, -6.25, 1.79, 14.15], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": -2.8, \"box_2d\": [269.86, 172.9, 449.21, 242.06], \"box_3d\": [1.68, 1.69, 4.07, -6.46, 1.71, 18.79], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": -1.85, \"box_2d\": [821.06, 150.49, 899.04, 199.95], \"box_3d\": [2.04, 1.99, 4.82, 11.08, 1.14, 32.42], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": -1.94, \"box_2d\": [878.52, 160.37, 947.02, 195.14], \"box_3d\": [1.49, 1.56, 3.88, 13.8, 0.98, 33.18], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -2.88, \"box_2d\": [343.4, 177.76, 481.47, 225.44], \"box_3d\": [1.48, 1.56, 4.12, -6.48, 1.66, 23.8], \"ry\": 3.14}, {\"category\": \"Car\", \"angle\": -2.92, \"box_2d\": [372.01, 176.32, 478.23, 219.9], \"box_3d\": [1.51, 1.59, 3.5, -6.75, 1.66, 26.44], \"ry\": 3.11}, {\"category\": \"Car\", \"angle\": -2.91, \"box_2d\": [382.36, 169.07, 506.8, 215.16], \"box_3d\": [1.74, 1.75, 4.57, -6.59, 1.63, 28.89], \"ry\": -3.14}]\n```", - "options": null, - "id": 63 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000676", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.44, \"box_2d\": [507.41, 173.44, 566.33, 218.27], \"box_3d\": [1.47, 1.59, 4.03, -2.25, 1.2, 25.79], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.49, \"box_2d\": [544.06, 172.23, 579.87, 201.6], \"box_3d\": [1.41, 1.54, 3.36, -1.96, 0.96, 36.46], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [606.23, 169.19, 639.9, 199.8], \"box_3d\": [1.52, 1.67, 4.38, 1.21, 0.91, 38.35], \"ry\": -1.52}, {\"category\": \"Car\", \"angle\": -1.5, \"box_2d\": [575.0, 168.3, 602.37, 192.51], \"box_3d\": [1.56, 1.59, 3.65, -0.81, 0.7, 48.67], \"ry\": -1.52}]\n```", - "options": null, - "id": 64 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000685", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.61, \"box_2d\": [904.51, 172.8, 1241.0, 374.0], \"box_3d\": [1.51, 1.69, 4.04, 3.45, 1.53, 5.88], \"ry\": -1.1}, {\"category\": \"Car\", \"angle\": -1.93, \"box_2d\": [645.61, 179.48, 866.65, 314.27], \"box_3d\": [1.46, 1.58, 3.93, 1.71, 1.57, 9.99], \"ry\": -1.77}, {\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [62.36, 196.22, 282.69, 314.79], \"box_3d\": [1.5, 1.68, 4.03, -6.93, 1.94, 11.88], \"ry\": 1.42}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [576.83, 171.33, 681.87, 249.04], \"box_3d\": [1.6, 1.7, 3.56, 0.3, 1.59, 16.85], \"ry\": -1.76}, {\"category\": \"Car\", \"angle\": -1.77, \"box_2d\": [542.59, 175.03, 625.23, 231.38], \"box_3d\": [1.52, 1.73, 3.89, -0.92, 1.62, 21.8], \"ry\": -1.81}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [280.34, 191.81, 352.98, 246.44], \"box_3d\": [1.44, 1.6, 3.79, -8.9, 2.06, 22.04], \"ry\": 1.31}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [311.61, 186.46, 366.3, 227.54], \"box_3d\": [1.42, 1.51, 3.71, -10.36, 1.96, 27.66], \"ry\": 1.33}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [527.21, 177.53, 581.5, 217.35], \"box_3d\": [1.37, 1.51, 3.49, -2.17, 1.56, 26.87], \"ry\": -1.79}]\n```", - "options": null, - "id": 65 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000690", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.42, \"box_2d\": [0.0, 221.34, 256.06, 374.0], \"box_3d\": [1.52, 1.73, 4.28, -3.39, 1.84, 2.84], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [480.17, 186.33, 545.85, 232.53], \"box_3d\": [1.44, 1.7, 3.65, -3.33, 1.94, 25.1], \"ry\": 1.6}]\n```", - "options": null, - "id": 66 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000694", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.09, \"box_2d\": [0.0, 77.93, 357.76, 374.0], \"box_3d\": [1.86, 1.63, 4.2, -2.91, 1.59, 4.09], \"ry\": 1.5}, {\"category\": \"Car\", \"angle\": -1.93, \"box_2d\": [771.65, 190.59, 1079.84, 360.67], \"box_3d\": [1.37, 1.61, 4.26, 3.12, 1.61, 8.27], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [468.61, 168.99, 526.47, 215.75], \"box_3d\": [1.51, 1.55, 3.92, -3.94, 1.42, 25.56], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [233.61, 172.09, 444.72, 316.21], \"box_3d\": [1.51, 1.55, 3.71, -3.35, 1.53, 9.52], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": -1.83, \"box_2d\": [720.78, 181.29, 854.87, 275.26], \"box_3d\": [1.5, 1.53, 3.37, 3.1, 1.67, 13.39], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [678.58, 174.86, 759.86, 233.2], \"box_3d\": [1.47, 1.56, 3.16, 2.91, 1.55, 20.05], \"ry\": -1.64}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [651.85, 172.28, 713.66, 219.73], \"box_3d\": [1.48, 1.56, 3.62, 2.36, 1.49, 24.51], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [638.41, 173.59, 686.63, 209.12], \"box_3d\": [1.39, 1.54, 4.6, 2.13, 1.45, 30.81], \"ry\": -1.61}]\n```", - "options": null, - "id": 67 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000748", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [572.36, 180.57, 612.06, 218.83], \"box_3d\": [1.66, 1.73, 3.05, -0.82, 2.03, 33.24], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.06, \"box_2d\": [123.85, 195.95, 258.43, 252.79], \"box_3d\": [1.47, 1.71, 4.36, -12.75, 2.24, 22.24], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [339.07, 187.41, 395.64, 218.12], \"box_3d\": [1.52, 1.71, 3.89, -13.04, 2.33, 38.9], \"ry\": 1.58}]\n```", - "options": null, - "id": 68 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000767", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [299.74, 185.98, 357.96, 211.54], \"box_3d\": [1.39, 1.82, 4.11, -16.58, 2.19, 42.69], \"ry\": 1.58}]\n```", - "options": null, - "id": 69 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000774", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.57, \"box_2d\": [633.13, 148.32, 661.2, 174.94], \"box_3d\": [1.5, 1.62, 4.19, 2.24, 0.1, 44.11], \"ry\": 1.62}]\n```", - "options": null, - "id": 70 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000777", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [576.83, 176.73, 605.63, 203.16], \"box_3d\": [1.43, 1.55, 3.91, -1.06, 1.67, 41.39], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.97, \"box_2d\": [648.3, 173.95, 728.05, 202.25], \"box_3d\": [1.51, 1.65, 4.21, 4.32, 1.6, 40.5], \"ry\": 3.07}]\n```", - "options": null, - "id": 71 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000864", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [575.5, 187.27, 607.31, 217.05], \"box_3d\": [1.43, 1.55, 3.91, -1.05, 2.21, 37.86], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": 2.71, \"box_2d\": [773.3, 180.08, 1052.03, 275.37], \"box_3d\": [1.51, 1.65, 4.21, 5.15, 1.66, 12.69], \"ry\": 3.09}, {\"category\": \"Car\", \"angle\": -0.28, \"box_2d\": [756.01, 181.73, 978.59, 257.21], \"box_3d\": [1.48, 1.49, 4.16, 5.27, 1.69, 15.25], \"ry\": 0.05}, {\"category\": \"Car\", \"angle\": -0.1, \"box_2d\": [710.58, 182.05, 860.14, 232.53], \"box_3d\": [1.51, 1.69, 4.42, 5.4, 1.82, 22.83], \"ry\": 0.12}, {\"category\": \"Car\", \"angle\": 2.89, \"box_2d\": [744.87, 179.44, 920.32, 247.77], \"box_3d\": [1.56, 1.5, 3.8, 5.28, 1.74, 17.55], \"ry\": -3.1}]\n```", - "options": null, - "id": 72 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000889", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (707.05, 707.05), (cx, cy) = (604.08, 180.51). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [581.94, 155.52, 635.25, 212.46], \"box_3d\": [1.91, 1.65, 4.4, 0.04, 1.06, 26.14], \"ry\": -1.63}]\n```", - "options": null, - "id": 73 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000914", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [757.39, 165.28, 832.71, 230.8], \"box_3d\": [1.56, 1.65, 3.69, 4.76, 1.4, 19.12], \"ry\": -1.39}, {\"category\": \"Car\", \"angle\": 2.53, \"box_2d\": [0.0, 188.83, 92.48, 328.13], \"box_3d\": [1.56, 1.62, 4.19, -7.91, 1.76, 7.43], \"ry\": 1.73}, {\"category\": \"Car\", \"angle\": 2.11, \"box_2d\": [369.3, 181.3, 510.84, 240.84], \"box_3d\": [1.42, 1.67, 4.53, -4.54, 1.67, 19.88], \"ry\": 1.89}, {\"category\": \"Car\", \"angle\": -0.48, \"box_2d\": [979.45, 157.16, 1076.51, 190.33], \"box_3d\": [1.54, 1.63, 3.74, 20.29, 0.82, 35.17], \"ry\": 0.04}]\n```", - "options": null, - "id": 74 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000917", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.92, \"box_2d\": [278.11, 167.1, 465.46, 228.82], \"box_3d\": [1.49, 1.67, 4.5, -6.36, 1.36, 19.3], \"ry\": 2.61}, {\"category\": \"Car\", \"angle\": -1.1, \"box_2d\": [909.2, 155.06, 1093.76, 252.73], \"box_3d\": [1.65, 1.67, 3.64, 7.56, 1.36, 13.91], \"ry\": -0.62}, {\"category\": \"Car\", \"angle\": 2.29, \"box_2d\": [1203.77, 149.81, 1241.0, 196.35], \"box_3d\": [1.59, 1.55, 3.93, 23.46, 0.79, 25.49], \"ry\": 3.03}]\n```", - "options": null, - "id": 75 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000929", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [611.39, 174.25, 665.48, 226.45], \"box_3d\": [1.61, 1.66, 3.2, 0.96, 1.68, 24.03], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [496.2, 179.78, 553.94, 219.1], \"box_3d\": [1.44, 1.61, 3.66, -3.31, 1.72, 28.46], \"ry\": 1.64}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [448.7, 181.06, 504.38, 214.7], \"box_3d\": [1.41, 1.53, 3.37, -5.93, 1.78, 32.23], \"ry\": 1.68}]\n```", - "options": null, - "id": 76 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000936", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [587.25, 183.18, 641.94, 228.06], \"box_3d\": [1.36, 1.69, 3.38, 0.1, 1.73, 24.13], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [0.0, 188.45, 213.17, 374.0], \"box_3d\": [1.46, 1.49, 3.4, -4.81, 1.6, 6.04], \"ry\": -2.4}, {\"category\": \"Car\", \"angle\": -1.95, \"box_2d\": [128.23, 181.84, 376.54, 325.12], \"box_3d\": [1.43, 1.62, 3.47, -4.63, 1.56, 9.03], \"ry\": -2.41}, {\"category\": \"Car\", \"angle\": -2.15, \"box_2d\": [331.58, 182.93, 487.4, 250.01], \"box_3d\": [1.37, 1.55, 3.96, -4.78, 1.62, 16.77], \"ry\": -2.43}, {\"category\": \"Car\", \"angle\": -2.12, \"box_2d\": [383.08, 180.87, 503.98, 239.54], \"box_3d\": [1.44, 1.59, 3.57, -4.66, 1.68, 19.75], \"ry\": -2.34}, {\"category\": \"Car\", \"angle\": -2.18, \"box_2d\": [431.15, 180.09, 525.14, 220.16], \"box_3d\": [1.4, 1.57, 3.83, -5.08, 1.7, 27.34], \"ry\": -2.37}, {\"category\": \"Car\", \"angle\": -2.29, \"box_2d\": [970.53, 158.15, 1241.0, 374.0], \"box_3d\": [1.72, 1.79, 3.94, 4.03, 1.64, 4.34], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -0.65, \"box_2d\": [914.78, 175.76, 1134.12, 258.29], \"box_3d\": [1.7, 1.77, 4.06, 9.27, 1.79, 16.36], \"ry\": -0.14}, {\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [592.26, 181.15, 626.59, 212.56], \"box_3d\": [1.52, 1.67, 3.61, -0.03, 1.95, 37.12], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": 2.22, \"box_2d\": [0.0, 186.11, 148.97, 302.58], \"box_3d\": [1.4, 1.44, 3.45, -8.25, 1.6, 10.03], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 2.05, \"box_2d\": [165.82, 173.97, 273.62, 220.95], \"box_3d\": [1.58, 1.73, 4.22, -14.33, 1.65, 26.71], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 2.04, \"box_2d\": [255.62, 181.18, 324.02, 211.0], \"box_3d\": [1.26, 1.46, 3.32, -14.41, 1.65, 32.57], \"ry\": 1.63}]\n```", - "options": null, - "id": 77 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000940", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.27, \"box_2d\": [0.0, 195.3, 85.4, 251.29], \"box_3d\": [1.34, 1.65, 4.44, -16.18, 1.98, 18.82], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [308.74, 184.77, 365.76, 212.68], \"box_3d\": [1.49, 1.76, 4.01, -15.67, 2.2, 41.57], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [380.6, 184.6, 418.22, 203.97], \"box_3d\": [1.38, 1.8, 3.41, -15.87, 2.28, 54.44], \"ry\": 1.58}]\n```", - "options": null, - "id": 78 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000984", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [373.48, 182.4, 419.25, 203.92], \"box_3d\": [1.49, 1.88, 4.41, -15.75, 2.22, 53.34], \"ry\": 1.62}]\n```", - "options": null, - "id": 79 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000992", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [694.52, 174.87, 805.75, 261.39], \"box_3d\": [1.57, 1.54, 3.22, 2.72, 1.63, 14.84], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [663.76, 167.23, 742.28, 237.48], \"box_3d\": [1.74, 1.6, 3.86, 2.43, 1.62, 19.93], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [650.71, 165.91, 711.69, 223.14], \"box_3d\": [1.83, 1.69, 4.44, 2.38, 1.63, 25.46], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [532.27, 169.06, 556.4, 193.64], \"box_3d\": [1.86, 1.57, 3.83, -5.17, 1.6, 56.98], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [471.58, 176.36, 514.64, 209.91], \"box_3d\": [1.5, 1.46, 3.7, -5.54, 1.68, 34.36], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [131.34, 193.24, 373.7, 318.48], \"box_3d\": [1.54, 1.7, 4.32, -5.41, 1.91, 11.54], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [314.39, 184.58, 435.22, 261.54], \"box_3d\": [1.54, 1.54, 3.72, -5.31, 1.83, 16.62], \"ry\": 1.58}]\n```", - "options": null, - "id": 80 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000994", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.13, \"box_2d\": [107.18, 182.53, 202.23, 222.36], \"box_3d\": [1.58, 1.64, 3.88, -19.51, 2.01, 31.06], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 2.09, \"box_2d\": [184.58, 183.63, 266.34, 214.37], \"box_3d\": [1.43, 1.81, 4.16, -19.37, 1.99, 36.49], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 2.03, \"box_2d\": [263.65, 181.86, 324.64, 207.6], \"box_3d\": [1.47, 1.67, 4.31, -19.23, 2.03, 44.07], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.98, \"box_2d\": [310.71, 179.87, 353.94, 202.42], \"box_3d\": [1.48, 1.44, 3.59, -19.06, 1.97, 49.59], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [343.58, 179.95, 382.96, 199.47], \"box_3d\": [1.42, 1.68, 3.38, -18.74, 1.97, 54.88], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [395.53, 178.45, 425.07, 196.36], \"box_3d\": [1.57, 1.47, 3.35, -18.16, 2.09, 65.67], \"ry\": 1.67}]\n```", - "options": null, - "id": 81 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "000995", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.19, \"box_2d\": [140.71, 191.0, 426.86, 366.73], \"box_3d\": [1.41, 1.54, 3.36, -3.12, 1.54, 7.71], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": -1.53, \"box_2d\": [509.13, 171.24, 587.2, 242.43], \"box_3d\": [1.52, 1.67, 4.38, -1.26, 1.31, 17.76], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.45, \"box_2d\": [502.86, 169.93, 564.05, 220.56], \"box_3d\": [1.56, 1.59, 3.65, -2.23, 1.21, 24.21], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [598.21, 170.06, 635.98, 207.95], \"box_3d\": [1.48, 1.47, 2.23, 0.61, 1.03, 29.4], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [462.18, 174.02, 503.05, 202.49], \"box_3d\": [1.42, 1.44, 3.19, -6.21, 1.05, 37.82], \"ry\": 1.63}]\n```", - "options": null, - "id": 82 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001007", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.4, \"box_2d\": [0.0, 174.69, 166.19, 339.8], \"box_3d\": [1.58, 1.54, 3.3, -6.92, 1.62, 7.72], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": 0.56, \"box_2d\": [0.0, 169.68, 326.87, 300.25], \"box_3d\": [1.68, 1.71, 4.07, -6.48, 1.67, 10.34], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": -2.73, \"box_2d\": [86.7, 177.52, 384.07, 274.23], \"box_3d\": [1.53, 1.7, 4.32, -6.43, 1.63, 12.67], \"ry\": 3.09}, {\"category\": \"Car\", \"angle\": 0.33, \"box_2d\": [274.75, 173.98, 464.49, 239.81], \"box_3d\": [1.54, 1.78, 4.2, -6.02, 1.6, 18.27], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": 0.16, \"box_2d\": [392.93, 175.04, 516.27, 216.83], \"box_3d\": [1.5, 1.66, 4.4, -5.9, 1.61, 27.61], \"ry\": -0.05}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [656.64, 180.27, 714.11, 222.95], \"box_3d\": [1.41, 1.62, 4.19, 2.62, 1.69, 26.25], \"ry\": -1.57}]\n```", - "options": null, - "id": 83 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001079", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [641.66, 178.79, 689.8, 216.72], \"box_3d\": [1.64, 1.69, 3.3, 2.49, 1.93, 33.26], \"ry\": -1.65}]\n```", - "options": null, - "id": 84 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001090", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.54, \"box_2d\": [540.86, 187.59, 572.24, 219.89], \"box_3d\": [1.68, 1.63, 3.9, -3.09, 2.54, 40.81], \"ry\": 1.46}]\n```", - "options": null, - "id": 85 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001098", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [563.63, 174.29, 618.67, 227.47], \"box_3d\": [1.6, 1.66, 3.55, -0.71, 1.67, 23.73], \"ry\": -1.63}]\n```", - "options": null, - "id": 86 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001117", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [630.8, 175.46, 653.85, 195.24], \"box_3d\": [1.42, 1.45, 3.2, 2.37, 1.63, 53.83], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [667.61, 180.68, 729.33, 226.81], \"box_3d\": [1.38, 1.49, 3.32, 2.77, 1.64, 23.48], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.91, \"box_2d\": [763.02, 182.19, 977.84, 310.17], \"box_3d\": [1.5, 1.62, 3.89, 3.49, 1.65, 10.59], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.85, \"box_2d\": [729.1, 175.97, 783.84, 207.71], \"box_3d\": [1.44, 1.56, 3.96, 7.03, 1.61, 35.12], \"ry\": -1.65}]\n```", - "options": null, - "id": 87 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001153", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.85, \"box_2d\": [383.38, 172.55, 745.19, 374.0], \"box_3d\": [1.6, 1.57, 3.23, -0.28, 1.62, 6.68], \"ry\": -0.9}, {\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [688.1, 174.33, 907.0, 314.2], \"box_3d\": [1.57, 1.5, 3.68, 2.69, 1.6, 9.99], \"ry\": 2.26}, {\"category\": \"Car\", \"angle\": -1.29, \"box_2d\": [886.11, 181.32, 1018.67, 262.02], \"box_3d\": [1.47, 1.6, 3.66, 7.29, 1.67, 15.28], \"ry\": -0.86}, {\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [1048.69, 175.86, 1123.25, 219.59], \"box_3d\": [1.7, 1.63, 4.08, 20.0, 1.85, 30.29], \"ry\": 2.41}, {\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [1233.88, 190.35, 1241.0, 259.75], \"box_3d\": [1.59, 1.59, 2.47, 16.14, 2.03, 17.38], \"ry\": -0.81}]\n```", - "options": null, - "id": 88 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001155", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [649.12, 183.94, 703.84, 225.62], \"box_3d\": [1.41, 1.58, 4.36, 2.39, 1.85, 27.23], \"ry\": -1.59}]\n```", - "options": null, - "id": 89 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001166", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [222.1, 191.45, 467.15, 344.94], \"box_3d\": [1.42, 1.66, 4.11, -3.06, 1.7, 9.13], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.6, \"box_2d\": [550.36, 157.67, 577.34, 185.05], \"box_3d\": [1.97, 1.87, 4.53, -3.47, 0.9, 54.69], \"ry\": 1.54}]\n```", - "options": null, - "id": 90 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001178", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.8, \"box_2d\": [754.39, 172.19, 900.54, 263.26], \"box_3d\": [1.38, 1.64, 3.51, 3.62, 1.39, 12.76], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [740.73, 169.42, 833.35, 238.84], \"box_3d\": [1.52, 1.53, 4.15, 4.25, 1.48, 18.03], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [739.34, 157.62, 804.81, 219.36], \"box_3d\": [1.8, 1.74, 4.96, 5.15, 1.38, 23.6], \"ry\": -1.42}]\n```", - "options": null, - "id": 91 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001202", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [550.41, 180.18, 577.8, 199.51], \"box_3d\": [1.44, 1.72, 4.14, -3.58, 2.03, 56.62], \"ry\": 1.61}]\n```", - "options": null, - "id": 92 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001212", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.04, \"box_2d\": [203.15, 196.65, 286.16, 232.69], \"box_3d\": [1.48, 1.79, 4.15, -17.13, 2.66, 34.0], \"ry\": 1.58}]\n```", - "options": null, - "id": 93 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001219", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [543.31, 179.15, 565.2, 197.83], \"box_3d\": [1.45, 1.53, 3.98, -4.53, 1.98, 58.78], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [501.25, 178.63, 544.45, 215.09], \"box_3d\": [1.65, 1.56, 3.98, -4.19, 1.94, 34.88], \"ry\": 1.58}]\n```", - "options": null, - "id": 94 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001261", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.25, \"box_2d\": [765.66, 148.65, 1005.7, 317.65], \"box_3d\": [1.77, 1.63, 4.1, 3.32, 1.55, 9.69], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [678.23, 165.78, 743.0, 218.81], \"box_3d\": [1.69, 1.64, 4.21, 3.41, 1.5, 25.39], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [502.89, 176.96, 552.87, 215.63], \"box_3d\": [1.55, 1.64, 3.7, -3.51, 1.74, 31.06], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -1.44, \"box_2d\": [566.92, 173.52, 608.01, 203.47], \"box_3d\": [1.63, 1.81, 4.54, -1.28, 1.7, 41.98], \"ry\": -1.47}]\n```", - "options": null, - "id": 95 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001312", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.62, \"box_2d\": [547.03, 176.7, 565.88, 193.04], \"box_3d\": [1.45, 1.53, 3.98, -4.95, 1.83, 66.91], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [518.79, 175.59, 552.12, 204.6], \"box_3d\": [1.65, 1.56, 3.98, -4.46, 1.83, 43.34], \"ry\": 1.57}]\n```", - "options": null, - "id": 96 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001324", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [406.78, 176.5, 482.13, 246.64], \"box_3d\": [1.68, 1.63, 3.9, -4.41, 1.8, 19.51], \"ry\": 1.44}, {\"category\": \"Car\", \"angle\": 1.49, \"box_2d\": [436.52, 177.89, 481.38, 211.8], \"box_3d\": [1.47, 1.7, 4.39, -7.18, 1.73, 33.88], \"ry\": 1.28}]\n```", - "options": null, - "id": 97 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001345", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [399.07, 181.42, 434.44, 200.9], \"box_3d\": [1.49, 1.76, 4.01, -15.5, 2.19, 58.0], \"ry\": 1.57}]\n```", - "options": null, - "id": 98 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001358", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (707.05, 707.05), (cx, cy) = (604.08, 180.51). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.61, \"box_2d\": [914.2, 168.07, 1223.0, 301.71], \"box_3d\": [1.31, 1.61, 3.78, 5.67, 1.19, 7.92], \"ry\": -0.01}, {\"category\": \"Car\", \"angle\": -0.22, \"box_2d\": [712.41, 168.74, 841.24, 208.98], \"box_3d\": [1.41, 1.63, 4.42, 6.35, 1.0, 26.51], \"ry\": 0.01}]\n```", - "options": null, - "id": 99 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001381", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [333.06, 176.16, 474.57, 269.44], \"box_3d\": [1.5, 1.62, 4.19, -3.8, 1.59, 13.86], \"ry\": 1.54}]\n```", - "options": null, - "id": 100 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001396", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [597.14, 180.9, 653.65, 226.91], \"box_3d\": [1.36, 1.69, 3.38, 0.42, 1.64, 23.41], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [419.2, 178.15, 458.67, 201.71], \"box_3d\": [1.59, 1.72, 3.86, -12.08, 1.98, 51.07], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [582.15, 175.79, 621.23, 211.45], \"box_3d\": [1.52, 1.67, 3.61, -0.37, 1.67, 32.82], \"ry\": -1.56}]\n```", - "options": null, - "id": 101 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001440", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.9, \"box_2d\": [369.84, 177.56, 464.63, 213.63], \"box_3d\": [1.49, 1.64, 3.7, -8.39, 1.72, 31.48], \"ry\": 3.13}]\n```", - "options": null, - "id": 102 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001460", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [672.82, 187.92, 776.98, 263.78], \"box_3d\": [1.41, 1.64, 3.77, 2.35, 1.77, 15.87], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [490.91, 188.09, 532.97, 215.78], \"box_3d\": [1.43, 1.76, 3.97, -5.53, 2.33, 40.91], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [365.46, 192.89, 463.17, 255.43], \"box_3d\": [1.49, 1.61, 3.94, -5.36, 2.09, 20.1], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [415.65, 190.1, 492.02, 238.07], \"box_3d\": [1.48, 1.66, 3.9, -5.4, 2.12, 25.26], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [150.29, 207.66, 370.2, 315.14], \"box_3d\": [1.31, 1.58, 4.23, -5.58, 1.98, 12.07], \"ry\": 1.58}]\n```", - "options": null, - "id": 103 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001501", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [367.39, 186.45, 413.18, 209.71], \"box_3d\": [1.46, 1.86, 4.01, -14.83, 2.4, 48.81], \"ry\": 1.57}]\n```", - "options": null, - "id": 104 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001510", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [629.16, 174.41, 656.27, 196.53], \"box_3d\": [1.5, 1.62, 3.89, 2.28, 1.64, 51.54], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.74, \"box_2d\": [669.41, 171.16, 732.52, 222.54], \"box_3d\": [1.69, 1.58, 3.95, 3.16, 1.66, 25.98], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [719.28, 182.28, 821.79, 252.43], \"box_3d\": [1.41, 1.57, 3.23, 3.48, 1.64, 16.36], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -2.02, \"box_2d\": [868.9, 179.82, 1033.59, 254.84], \"box_3d\": [1.45, 1.61, 4.2, 7.39, 1.62, 16.21], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -2.48, \"box_2d\": [809.78, 175.2, 877.81, 202.99], \"box_3d\": [1.48, 1.36, 3.51, 13.02, 1.63, 40.41], \"ry\": -2.17}]\n```", - "options": null, - "id": 105 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001531", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.7, \"box_2d\": [0.0, 196.48, 304.15, 316.19], \"box_3d\": [1.38, 1.67, 4.37, -6.69, 1.74, 9.71], \"ry\": 2.99}, {\"category\": \"Car\", \"angle\": -2.23, \"box_2d\": [1075.99, 150.49, 1241.0, 279.95], \"box_3d\": [1.54, 1.44, 2.54, 7.88, 1.28, 9.86], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 0.4, \"box_2d\": [233.99, 184.3, 443.47, 250.6], \"box_3d\": [1.42, 1.41, 4.3, -6.2, 1.71, 16.59], \"ry\": 0.05}]\n```", - "options": null, - "id": 106 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001546", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [581.6, 176.42, 599.71, 190.77], \"box_3d\": [1.26, 1.6, 3.56, -1.79, 1.6, 65.84], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [340.44, 182.08, 427.1, 233.16], \"box_3d\": [1.48, 1.64, 3.59, -7.17, 1.79, 23.08], \"ry\": 1.57}]\n```", - "options": null, - "id": 107 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001594", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.24, \"box_2d\": [0.0, 208.3, 315.98, 374.0], \"box_3d\": [1.38, 1.69, 3.74, -3.05, 1.63, 3.47], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [682.99, 184.25, 756.21, 242.35], \"box_3d\": [1.51, 1.56, 3.81, 3.07, 1.86, 21.09], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [668.57, 182.33, 729.87, 229.62], \"box_3d\": [1.57, 1.68, 3.83, 3.14, 1.93, 26.36], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.59, \"box_2d\": [536.31, 178.36, 566.16, 208.44], \"box_3d\": [1.52, 1.5, 3.85, -3.14, 1.83, 38.71], \"ry\": 1.51}]\n```", - "options": null, - "id": 108 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001608", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [602.8, 176.56, 670.99, 231.97], \"box_3d\": [1.36, 1.69, 3.38, 0.63, 1.48, 19.69], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.52, \"box_2d\": [1016.5, 166.63, 1154.68, 206.08], \"box_3d\": [1.45, 1.67, 4.32, 18.44, 1.24, 28.13], \"ry\": 3.1}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [385.0, 177.62, 434.12, 206.82], \"box_3d\": [1.59, 1.72, 3.86, -11.53, 1.88, 41.62], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [584.99, 172.99, 627.78, 212.04], \"box_3d\": [1.52, 1.67, 3.61, -0.19, 1.55, 30.14], \"ry\": -1.58}]\n```", - "options": null, - "id": 109 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001613", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [431.65, 182.37, 464.52, 201.12], \"box_3d\": [1.55, 1.88, 4.59, -14.11, 2.39, 63.05], \"ry\": 1.56}]\n```", - "options": null, - "id": 110 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001614", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.16, \"box_2d\": [583.44, 186.99, 762.22, 282.79], \"box_3d\": [1.48, 1.71, 4.65, 1.38, 1.79, 14.03], \"ry\": -1.07}, {\"category\": \"Car\", \"angle\": -0.94, \"box_2d\": [447.69, 162.94, 635.32, 272.52], \"box_3d\": [2.04, 1.66, 4.62, -1.34, 1.87, 15.83], \"ry\": -1.03}, {\"category\": \"Car\", \"angle\": -1.49, \"box_2d\": [872.24, 167.19, 958.81, 232.79], \"box_3d\": [1.42, 1.64, 3.95, 7.57, 1.32, 17.77], \"ry\": -1.09}, {\"category\": \"Car\", \"angle\": -0.74, \"box_2d\": [272.24, 190.82, 460.84, 262.76], \"box_3d\": [1.53, 1.61, 4.73, -6.12, 2.04, 18.66], \"ry\": -1.05}, {\"category\": \"Car\", \"angle\": -0.64, \"box_2d\": [198.02, 196.91, 339.01, 252.77], \"box_3d\": [1.4, 1.56, 3.48, -9.93, 2.15, 21.13], \"ry\": -1.07}]\n```", - "options": null, - "id": 111 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001636", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.37, \"box_2d\": [729.46, 180.61, 775.58, 207.91], \"box_3d\": [1.48, 1.8, 4.31, 8.34, 1.96, 42.07], \"ry\": -1.18}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [655.01, 180.19, 688.71, 206.82], \"box_3d\": [1.69, 1.88, 4.5, 4.19, 2.2, 48.52], \"ry\": 1.74}]\n```", - "options": null, - "id": 112 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001642", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.28, \"box_2d\": [415.51, 183.18, 503.66, 209.05], \"box_3d\": [1.36, 1.68, 4.49, -8.34, 1.96, 40.07], \"ry\": 0.08}]\n```", - "options": null, - "id": 113 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001669", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.25, \"box_2d\": [0.0, 202.0, 330.76, 374.0], \"box_3d\": [1.6, 1.68, 3.94, -3.01, 1.8, 3.47], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [519.33, 186.05, 565.04, 221.09], \"box_3d\": [1.42, 1.66, 4.11, -3.01, 2.04, 32.34], \"ry\": 1.57}]\n```", - "options": null, - "id": 114 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001720", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.27, \"box_2d\": [0.0, 201.56, 142.82, 320.82], \"box_3d\": [1.53, 1.66, 4.01, -9.13, 2.02, 10.72], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 2.09, \"box_2d\": [100.22, 193.47, 277.74, 274.71], \"box_3d\": [1.48, 1.6, 3.9, -9.08, 1.98, 15.89], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [395.03, 181.1, 454.37, 216.33], \"box_3d\": [1.61, 1.67, 4.56, -9.1, 2.03, 35.68], \"ry\": 1.59}]\n```", - "options": null, - "id": 115 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001769", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.27, \"box_2d\": [0.0, 181.64, 109.17, 316.75], \"box_3d\": [1.8, 1.61, 3.83, -9.01, 1.93, 9.89], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 2.2, \"box_2d\": [79.62, 194.97, 261.27, 278.46], \"box_3d\": [1.45, 1.45, 3.4, -8.91, 1.94, 14.87], \"ry\": 1.67}, {\"category\": \"Car\", \"angle\": 2.0, \"box_2d\": [212.59, 179.61, 348.88, 252.31], \"box_3d\": [1.75, 1.71, 4.31, -8.84, 1.95, 19.72], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [365.06, 182.58, 436.45, 223.51], \"box_3d\": [1.53, 1.66, 4.01, -8.47, 1.94, 29.43], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [410.82, 181.17, 466.89, 214.36], \"box_3d\": [1.48, 1.6, 3.9, -8.18, 1.89, 34.64], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [494.52, 177.24, 527.6, 199.73], \"box_3d\": [1.61, 1.67, 4.56, -7.43, 1.96, 54.42], \"ry\": 1.61}]\n```", - "options": null, - "id": 116 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "001793", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [475.76, 183.59, 549.41, 232.12], \"box_3d\": [1.48, 1.87, 4.42, -3.27, 1.87, 24.78], \"ry\": 1.58}]\n```", - "options": null, - "id": 117 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002032", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [579.01, 173.56, 604.92, 188.61], \"box_3d\": [1.45, 1.59, 4.28, -1.78, 1.55, 72.36], \"ry\": 1.79}, {\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [719.29, 173.26, 739.12, 190.02], \"box_3d\": [1.43, 1.59, 4.21, 10.66, 1.49, 64.43], \"ry\": -1.37}]\n```", - "options": null, - "id": 118 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002033", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.96, \"box_2d\": [577.56, 173.61, 830.23, 259.64], \"box_3d\": [1.47, 1.77, 4.49, 1.61, 1.51, 13.72], \"ry\": 3.08}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [547.9, 184.44, 575.2, 208.59], \"box_3d\": [1.52, 1.62, 4.5, -3.34, 2.32, 48.74], \"ry\": -1.68}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [453.42, 187.92, 487.62, 212.19], \"box_3d\": [1.42, 1.66, 4.5, -8.93, 2.42, 46.35], \"ry\": 1.49}]\n```", - "options": null, - "id": 119 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002038", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.37, \"box_2d\": [1096.87, 174.29, 1233.96, 260.59], \"box_3d\": [1.65, 1.67, 3.64, 12.09, 1.7, 15.68], \"ry\": -0.72}, {\"category\": \"Car\", \"angle\": 2.3, \"box_2d\": [623.24, 179.23, 782.74, 255.14], \"box_3d\": [1.59, 1.55, 3.93, 2.26, 1.77, 17.27], \"ry\": 2.43}, {\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [877.05, 180.37, 979.81, 227.64], \"box_3d\": [1.47, 1.57, 4.24, 11.02, 1.75, 24.92], \"ry\": 2.43}, {\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [1006.5, 180.5, 1084.61, 214.51], \"box_3d\": [1.4, 1.68, 4.45, 19.67, 1.77, 32.55], \"ry\": 2.43}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [1078.58, 179.52, 1139.87, 209.96], \"box_3d\": [1.45, 1.62, 3.77, 25.55, 1.81, 36.93], \"ry\": 2.44}]\n```", - "options": null, - "id": 120 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002041", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [199.46, 176.48, 399.01, 323.96], \"box_3d\": [1.7, 1.5, 3.34, -4.16, 1.77, 10.08], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 1.34, \"box_2d\": [319.58, 188.12, 450.36, 273.67], \"box_3d\": [1.48, 1.71, 3.93, -4.82, 1.81, 14.93], \"ry\": 1.04}, {\"category\": \"Car\", \"angle\": -0.75, \"box_2d\": [842.79, 160.35, 908.02, 182.4], \"box_3d\": [1.6, 1.75, 4.63, 19.92, 0.69, 54.19], \"ry\": -0.4}, {\"category\": \"Car\", \"angle\": -1.85, \"box_2d\": [1034.91, 189.97, 1241.0, 374.0], \"box_3d\": [1.51, 1.58, 3.31, 3.73, 1.64, 4.39], \"ry\": -1.18}, {\"category\": \"Car\", \"angle\": -1.45, \"box_2d\": [990.1, 166.12, 1188.15, 336.63], \"box_3d\": [1.61, 1.6, 3.11, 5.71, 1.56, 8.51], \"ry\": -0.88}, {\"category\": \"Car\", \"angle\": 3.11, \"box_2d\": [726.45, 160.98, 910.69, 229.88], \"box_3d\": [1.79, 1.69, 4.72, 5.68, 1.52, 20.15], \"ry\": -2.91}, {\"category\": \"Car\", \"angle\": 2.13, \"box_2d\": [1038.64, 168.01, 1171.3, 215.63], \"box_3d\": [1.43, 1.7, 3.95, 16.02, 1.3, 23.4], \"ry\": 2.73}]\n```", - "options": null, - "id": 121 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002064", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [188.09, 182.39, 420.91, 323.83], \"box_3d\": [1.59, 1.63, 3.83, -4.05, 1.74, 10.17], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [659.9, 168.48, 695.33, 201.93], \"box_3d\": [1.62, 1.7, 3.29, 3.4, 1.44, 37.05], \"ry\": -1.51}]\n```", - "options": null, - "id": 122 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002094", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.21, \"box_2d\": [0.0, 187.8, 140.31, 257.61], \"box_3d\": [1.58, 1.82, 3.87, -14.4, 2.0, 18.74], \"ry\": 1.56}]\n```", - "options": null, - "id": 123 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002240", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.54, \"box_2d\": [18.14, 191.26, 280.39, 284.87], \"box_3d\": [1.52, 1.64, 3.7, -8.26, 1.89, 13.09], \"ry\": -0.02}, {\"category\": \"Car\", \"angle\": -2.62, \"box_2d\": [191.47, 191.18, 365.98, 254.81], \"box_3d\": [1.5, 1.62, 3.69, -8.35, 1.99, 18.25], \"ry\": -3.04}, {\"category\": \"Car\", \"angle\": 1.29, \"box_2d\": [945.16, 160.32, 1002.18, 188.75], \"box_3d\": [1.51, 1.7, 4.45, 20.53, 0.87, 40.96], \"ry\": 1.76}]\n```", - "options": null, - "id": 124 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002242", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [466.82, 185.0, 539.35, 234.47], \"box_3d\": [1.45, 1.77, 4.04, -3.47, 1.88, 23.84], \"ry\": 1.58}]\n```", - "options": null, - "id": 125 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002244", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.62, \"box_2d\": [546.18, 178.81, 571.58, 202.28], \"box_3d\": [1.7, 1.72, 4.11, -3.87, 2.17, 54.89], \"ry\": 1.55}]\n```", - "options": null, - "id": 126 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002320", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.57, \"box_2d\": [619.6, 163.41, 648.29, 190.08], \"box_3d\": [1.5, 1.62, 4.19, 1.4, 0.99, 43.18], \"ry\": 1.6}]\n```", - "options": null, - "id": 127 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002355", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.06, \"box_2d\": [127.46, 199.34, 245.49, 252.76], \"box_3d\": [1.35, 1.5, 3.64, -12.57, 2.2, 21.6], \"ry\": 1.53}]\n```", - "options": null, - "id": 128 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002361", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.07, \"box_2d\": [782.22, 192.06, 1093.38, 373.49], \"box_3d\": [1.39, 1.59, 3.17, 3.39, 1.62, 7.39], \"ry\": -0.66}, {\"category\": \"Car\", \"angle\": -0.77, \"box_2d\": [634.58, 189.0, 958.28, 330.22], \"box_3d\": [1.46, 1.55, 4.28, 2.51, 1.71, 9.56], \"ry\": -0.52}, {\"category\": \"Car\", \"angle\": -0.48, \"box_2d\": [426.71, 190.75, 635.79, 273.73], \"box_3d\": [1.49, 1.66, 4.16, -1.67, 1.92, 15.51], \"ry\": -0.59}, {\"category\": \"Car\", \"angle\": 2.8, \"box_2d\": [340.71, 179.15, 520.23, 257.22], \"box_3d\": [1.85, 1.68, 4.34, -4.78, 2.04, 19.3], \"ry\": 2.56}]\n```", - "options": null, - "id": 129 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002417", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [559.04, 178.05, 580.68, 194.46], \"box_3d\": [1.45, 1.77, 4.04, -3.69, 1.95, 66.65], \"ry\": 1.57}]\n```", - "options": null, - "id": 130 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002434", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.21, \"box_2d\": [0.0, 190.8, 153.16, 248.0], \"box_3d\": [1.49, 1.76, 4.01, -15.95, 2.07, 21.71], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 2.0, \"box_2d\": [241.18, 188.53, 311.16, 219.87], \"box_3d\": [1.38, 1.8, 3.41, -16.03, 2.16, 34.74], \"ry\": 1.57}]\n```", - "options": null, - "id": 131 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002469", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [533.1, 174.6, 555.89, 195.22], \"box_3d\": [1.65, 1.56, 3.98, -5.44, 1.82, 60.15], \"ry\": 1.56}]\n```", - "options": null, - "id": 132 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002515", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [375.39, 183.56, 500.35, 293.98], \"box_3d\": [1.65, 1.75, 3.39, -2.93, 1.85, 12.74], \"ry\": 1.45}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [453.71, 184.16, 532.47, 247.8], \"box_3d\": [1.57, 1.54, 4.01, -3.21, 1.9, 20.21], \"ry\": 1.57}]\n```", - "options": null, - "id": 133 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002518", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [417.4, 181.38, 498.76, 242.92], \"box_3d\": [1.7, 1.72, 4.11, -4.61, 1.98, 22.27], \"ry\": 1.55}]\n```", - "options": null, - "id": 134 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002550", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [43.01, 176.47, 439.29, 374.0], \"box_3d\": [1.65, 1.75, 3.39, -2.59, 1.7, 5.9], \"ry\": 1.48}, {\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [393.84, 180.67, 526.04, 281.44], \"box_3d\": [1.57, 1.54, 4.01, -2.64, 1.72, 13.36], \"ry\": 1.59}]\n```", - "options": null, - "id": 135 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002608", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.6, \"box_2d\": [452.34, 191.44, 519.08, 255.75], \"box_3d\": [1.65, 1.75, 3.39, -3.6, 2.22, 21.05], \"ry\": 1.43}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [479.21, 191.26, 531.19, 236.15], \"box_3d\": [1.57, 1.54, 4.01, -4.1, 2.33, 28.46], \"ry\": 1.55}]\n```", - "options": null, - "id": 136 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002660", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.55, \"box_2d\": [104.88, 194.63, 336.49, 259.53], \"box_3d\": [1.36, 1.68, 4.49, -8.93, 1.9, 16.66], \"ry\": 0.06}]\n```", - "options": null, - "id": 137 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002672", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.48, \"box_2d\": [690.91, 183.35, 721.75, 208.7], \"box_3d\": [1.54, 1.63, 3.67, 6.24, 2.24, 46.65], \"ry\": -1.34}, {\"category\": \"Car\", \"angle\": 2.32, \"box_2d\": [0.0, 194.35, 228.35, 356.77], \"box_3d\": [1.56, 1.59, 3.69, -6.76, 1.87, 9.03], \"ry\": 1.69}, {\"category\": \"Car\", \"angle\": 2.11, \"box_2d\": [165.68, 169.46, 369.08, 290.18], \"box_3d\": [1.93, 1.59, 4.01, -6.25, 1.91, 13.63], \"ry\": 1.69}, {\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [722.74, 182.95, 823.6, 257.29], \"box_3d\": [1.58, 1.67, 3.84, 3.8, 1.85, 17.57], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [629.0, 183.08, 666.36, 208.79], \"box_3d\": [1.53, 1.67, 3.78, 2.4, 2.19, 45.55], \"ry\": 1.81}, {\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [723.29, 167.42, 784.98, 240.46], \"box_3d\": [2.05, 1.5, 3.83, 4.34, 1.93, 22.3], \"ry\": -1.47}]\n```", - "options": null, - "id": 138 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002750", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [802.5, 178.94, 928.66, 246.61], \"box_3d\": [1.48, 1.51, 4.35, 6.49, 1.65, 18.13], \"ry\": 2.28}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [1109.25, 174.78, 1157.64, 203.81], \"box_3d\": [1.46, 1.74, 3.99, 27.87, 1.57, 38.4], \"ry\": 2.28}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [884.31, 181.36, 985.4, 233.28], \"box_3d\": [1.42, 1.68, 4.29, 10.11, 1.71, 22.36], \"ry\": 2.28}, {\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [968.03, 179.76, 1032.64, 217.66], \"box_3d\": [1.36, 1.4, 3.8, 15.27, 1.65, 28.19], \"ry\": 2.29}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [1044.25, 176.96, 1086.28, 205.39], \"box_3d\": [1.44, 1.65, 2.96, 24.28, 1.68, 38.49], \"ry\": 2.21}]\n```", - "options": null, - "id": 139 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002766", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.62, \"box_2d\": [533.4, 172.66, 566.78, 205.81], \"box_3d\": [1.97, 1.87, 4.53, -3.76, 1.99, 45.62], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [558.53, 175.33, 577.46, 193.62], \"box_3d\": [1.7, 1.53, 3.82, -4.04, 1.96, 69.73], \"ry\": 1.59}]\n```", - "options": null, - "id": 140 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002770", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.97, \"box_2d\": [0.0, 187.51, 376.85, 374.0], \"box_3d\": [1.77, 1.75, 4.67, -3.18, 1.89, 4.52], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 1.78, \"box_2d\": [404.06, 182.31, 525.19, 270.5], \"box_3d\": [1.6, 1.68, 3.94, -2.95, 1.81, 15.24], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [553.1, 179.8, 584.72, 204.39], \"box_3d\": [1.42, 1.66, 4.11, -2.5, 1.86, 44.25], \"ry\": 1.58}]\n```", - "options": null, - "id": 141 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002784", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.53, \"box_2d\": [534.85, 183.46, 562.66, 211.04], \"box_3d\": [1.68, 1.63, 3.9, -4.03, 2.39, 46.83], \"ry\": 1.45}]\n```", - "options": null, - "id": 142 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002787", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.21, \"box_2d\": [733.38, 136.2, 945.33, 272.19], \"box_3d\": [1.79, 1.72, 4.35, 3.39, 1.33, 11.77], \"ry\": 1.48}, {\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [278.8, 188.44, 434.64, 291.61], \"box_3d\": [1.45, 1.49, 4.07, -4.25, 1.75, 12.61], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [421.28, 185.27, 490.74, 234.56], \"box_3d\": [1.27, 1.54, 3.58, -4.43, 1.65, 20.97], \"ry\": 1.48}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [455.58, 172.48, 511.73, 219.43], \"box_3d\": [1.63, 1.65, 3.91, -4.75, 1.65, 27.34], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [588.29, 169.6, 616.89, 195.28], \"box_3d\": [1.46, 1.53, 3.17, -0.5, 1.3, 43.04], \"ry\": -1.63}]\n```", - "options": null, - "id": 143 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002791", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.43, \"box_2d\": [607.66, 185.09, 652.84, 216.23], \"box_3d\": [1.44, 1.66, 4.53, 1.06, 2.08, 36.56], \"ry\": -1.4}]\n```", - "options": null, - "id": 144 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002815", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.89, \"box_2d\": [748.26, 150.35, 1071.41, 373.69], \"box_3d\": [1.76, 1.71, 4.35, 2.76, 1.61, 7.91], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 0.85, \"box_2d\": [0.0, 190.41, 119.86, 374.0], \"box_3d\": [1.52, 1.71, 4.49, -6.36, 1.68, 5.09], \"ry\": -0.02}, {\"category\": \"Car\", \"angle\": -2.41, \"box_2d\": [0.0, 191.67, 230.81, 357.11], \"box_3d\": [1.49, 1.61, 3.87, -6.35, 1.72, 7.5], \"ry\": -3.1}, {\"category\": \"Car\", \"angle\": -2.56, \"box_2d\": [0.0, 178.57, 298.64, 297.28], \"box_3d\": [1.5, 1.58, 3.47, -6.48, 1.6, 10.07], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": 0.44, \"box_2d\": [158.71, 155.71, 421.66, 260.61], \"box_3d\": [1.95, 1.64, 4.57, -6.34, 1.66, 14.5], \"ry\": 0.03}, {\"category\": \"Car\", \"angle\": -2.67, \"box_2d\": [116.38, 171.13, 359.58, 277.79], \"box_3d\": [1.66, 1.58, 3.31, -6.24, 1.66, 12.25], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": -1.82, \"box_2d\": [707.66, 177.92, 846.09, 268.3], \"box_3d\": [1.44, 1.66, 3.25, 2.87, 1.55, 13.26], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -2.77, \"box_2d\": [233.2, 160.04, 413.3, 246.41], \"box_3d\": [1.87, 1.61, 3.53, -6.6, 1.62, 16.74], \"ry\": -3.14}, {\"category\": \"Car\", \"angle\": 0.44, \"box_2d\": [295.98, 178.28, 450.83, 242.44], \"box_3d\": [1.59, 1.56, 3.5, -6.16, 1.75, 18.83], \"ry\": 0.13}, {\"category\": \"Car\", \"angle\": -2.8, \"box_2d\": [303.43, 170.14, 466.27, 231.43], \"box_3d\": [1.72, 1.65, 4.33, -6.68, 1.67, 21.53], \"ry\": -3.1}, {\"category\": \"Car\", \"angle\": 0.29, \"box_2d\": [392.91, 167.84, 503.97, 208.6], \"box_3d\": [1.57, 1.54, 4.1, -6.53, 1.4, 29.2], \"ry\": 0.07}, {\"category\": \"Car\", \"angle\": -2.95, \"box_2d\": [429.36, 162.4, 518.15, 204.26], \"box_3d\": [1.89, 1.59, 3.87, -6.42, 1.44, 34.07], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [651.53, 173.33, 698.51, 209.93], \"box_3d\": [1.45, 1.59, 3.59, 2.68, 1.5, 30.78], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [682.5, 174.41, 770.03, 240.81], \"box_3d\": [1.54, 1.67, 3.72, 2.87, 1.6, 18.77], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [639.21, 166.09, 673.62, 196.58], \"box_3d\": [1.66, 1.74, 3.65, 2.6, 1.32, 41.53], \"ry\": -1.57}]\n```", - "options": null, - "id": 145 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002829", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.81, \"box_2d\": [271.73, 201.21, 499.32, 374.0], \"box_3d\": [1.38, 1.43, 3.64, -2.21, 1.75, 7.87], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": -0.59, \"box_2d\": [668.47, 172.86, 841.71, 251.4], \"box_3d\": [1.62, 1.47, 3.66, 3.27, 1.64, 16.44], \"ry\": -0.4}]\n```", - "options": null, - "id": 146 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002835", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [546.08, 179.86, 574.66, 200.58], \"box_3d\": [1.44, 1.72, 4.14, -3.62, 1.97, 52.96], \"ry\": 1.59}]\n```", - "options": null, - "id": 147 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002836", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 3.07, \"box_2d\": [899.37, 176.51, 974.98, 204.6], \"box_3d\": [1.42, 1.58, 3.48, 17.03, 1.63, 37.76], \"ry\": -2.79}, {\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [339.83, 188.91, 381.97, 207.39], \"box_3d\": [1.41, 1.85, 4.7, -20.83, 2.8, 60.48], \"ry\": 1.58}]\n```", - "options": null, - "id": 148 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002879", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [471.83, 176.1, 553.23, 240.39], \"box_3d\": [1.46, 1.63, 4.47, -2.47, 1.57, 18.88], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [526.42, 165.78, 588.41, 209.98], \"box_3d\": [1.7, 1.82, 4.45, -2.14, 1.45, 30.18], \"ry\": 1.68}, {\"category\": \"Car\", \"angle\": -0.72, \"box_2d\": [612.77, 167.65, 664.07, 195.27], \"box_3d\": [1.57, 1.56, 2.62, 1.66, 1.29, 42.49], \"ry\": -0.68}, {\"category\": \"Car\", \"angle\": -2.43, \"box_2d\": [0.0, 179.91, 193.71, 374.0], \"box_3d\": [1.47, 1.68, 4.26, -6.11, 1.55, 5.93], \"ry\": 3.08}]\n```", - "options": null, - "id": 149 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002910", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [648.07, 182.34, 742.98, 247.07], \"box_3d\": [1.38, 1.51, 4.1, 1.94, 1.63, 17.77], \"ry\": -1.66}, {\"category\": \"Car\", \"angle\": -2.09, \"box_2d\": [797.24, 158.03, 1219.69, 374.0], \"box_3d\": [1.6, 1.66, 3.55, 3.12, 1.52, 6.62], \"ry\": -1.67}]\n```", - "options": null, - "id": 150 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002941", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.55, \"box_2d\": [520.77, 179.88, 555.6, 215.69], \"box_3d\": [1.68, 1.63, 3.9, -3.68, 2.05, 36.27], \"ry\": 1.45}, {\"category\": \"Car\", \"angle\": 1.45, \"box_2d\": [493.8, 181.79, 524.34, 202.72], \"box_3d\": [1.47, 1.7, 4.39, -7.55, 2.14, 53.51], \"ry\": 1.31}]\n```", - "options": null, - "id": 151 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002958", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.87, \"box_2d\": [744.32, 181.71, 998.35, 331.31], \"box_3d\": [1.41, 1.58, 4.36, 2.87, 1.53, 9.09], \"ry\": -1.58}]\n```", - "options": null, - "id": 152 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "002959", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.26, \"box_2d\": [930.55, 208.82, 1241.0, 374.0], \"box_3d\": [1.51, 1.56, 3.81, 3.07, 1.76, 3.33], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.91, \"box_2d\": [763.29, 186.68, 1039.96, 362.67], \"box_3d\": [1.57, 1.68, 3.83, 3.06, 1.76, 8.58], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [466.46, 177.36, 530.7, 235.99], \"box_3d\": [1.52, 1.5, 3.85, -3.17, 1.66, 20.82], \"ry\": 1.5}, {\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [667.75, 181.66, 723.09, 223.38], \"box_3d\": [1.43, 1.59, 3.68, 3.08, 1.77, 26.9], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [516.02, 178.58, 556.48, 208.48], \"box_3d\": [1.29, 1.59, 4.12, -3.42, 1.58, 33.72], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [538.99, 174.37, 569.44, 202.15], \"box_3d\": [1.52, 1.61, 3.59, -3.21, 1.63, 41.63], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.49, \"box_2d\": [645.22, 177.13, 679.86, 204.39], \"box_3d\": [1.51, 1.65, 4.28, 3.02, 1.78, 42.52], \"ry\": 1.56}]\n```", - "options": null, - "id": 153 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003015", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.45, \"box_2d\": [733.66, 181.69, 901.64, 326.55], \"box_3d\": [1.51, 1.69, 4.04, 2.96, 1.65, 9.81], \"ry\": -1.17}, {\"category\": \"Car\", \"angle\": -1.9, \"box_2d\": [596.47, 186.05, 744.91, 278.47], \"box_3d\": [1.46, 1.58, 3.93, 0.94, 1.73, 13.81], \"ry\": -1.83}, {\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [153.59, 199.58, 301.06, 290.06], \"box_3d\": [1.5, 1.68, 4.03, -7.83, 2.12, 15.08], \"ry\": 1.36}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [538.9, 182.4, 624.39, 245.02], \"box_3d\": [1.6, 1.7, 3.56, -0.94, 1.88, 20.54], \"ry\": -1.82}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [514.33, 182.75, 586.7, 230.7], \"box_3d\": [1.52, 1.73, 3.89, -2.21, 1.88, 25.34], \"ry\": -1.87}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [278.75, 190.9, 340.5, 238.02], \"box_3d\": [1.44, 1.6, 3.79, -10.42, 2.11, 25.14], \"ry\": 1.28}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [491.73, 188.42, 540.0, 223.95], \"box_3d\": [1.37, 1.51, 3.49, -4.07, 2.05, 30.57], \"ry\": -1.85}]\n```", - "options": null, - "id": 154 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003049", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.42, \"box_2d\": [254.34, 187.95, 415.36, 234.01], \"box_3d\": [1.36, 1.68, 4.49, -8.73, 1.87, 22.98], \"ry\": 0.06}]\n```", - "options": null, - "id": 155 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003062", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [434.48, 188.96, 531.68, 258.17], \"box_3d\": [1.46, 1.75, 4.23, -3.08, 1.9, 18.08], \"ry\": 1.55}]\n```", - "options": null, - "id": 156 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003076", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [397.99, 183.14, 436.43, 212.75], \"box_3d\": [1.45, 1.62, 3.83, -10.13, 2.0, 37.98], \"ry\": 1.4}]\n```", - "options": null, - "id": 157 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003103", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.76, \"box_2d\": [406.36, 172.59, 523.27, 266.09], \"box_3d\": [1.7, 1.63, 4.08, -2.96, 1.73, 15.3], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": -2.21, \"box_2d\": [896.53, 186.21, 1241.0, 374.0], \"box_3d\": [1.59, 1.59, 2.47, 2.43, 1.66, 2.97], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [749.06, 196.18, 990.51, 366.58], \"box_3d\": [1.37, 1.59, 3.22, 2.47, 1.67, 7.76], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": -1.4, \"box_2d\": [489.2, 183.11, 561.01, 231.47], \"box_3d\": [1.4, 1.67, 3.72, -2.68, 1.75, 23.22], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": 1.5, \"box_2d\": [641.29, 177.5, 675.14, 205.82], \"box_3d\": [1.49, 1.61, 3.74, 2.61, 1.76, 40.14], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -2.48, \"box_2d\": [0.0, 162.71, 249.7, 304.9], \"box_3d\": [1.59, 1.63, 4.43, -7.17, 1.51, 9.0], \"ry\": -3.14}]\n```", - "options": null, - "id": 158 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003111", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.8, \"box_2d\": [668.92, 182.7, 778.1, 253.41], \"box_3d\": [1.38, 1.51, 4.1, 2.4, 1.62, 16.47], \"ry\": -1.66}, {\"category\": \"Car\", \"angle\": -2.18, \"box_2d\": [855.7, 155.55, 1241.0, 374.0], \"box_3d\": [1.6, 1.66, 3.55, 3.49, 1.52, 5.76], \"ry\": -1.66}]\n```", - "options": null, - "id": 159 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003117", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.51, \"box_2d\": [362.99, 194.12, 401.02, 227.64], \"box_3d\": [1.65, 1.75, 3.39, -12.39, 2.84, 39.01], \"ry\": 1.21}, {\"category\": \"Car\", \"angle\": 1.94, \"box_2d\": [0.0, 200.53, 416.4, 374.0], \"box_3d\": [1.47, 1.7, 4.39, -3.05, 1.78, 6.31], \"ry\": 1.5}]\n```", - "options": null, - "id": 160 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003153", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [498.21, 184.79, 561.16, 228.71], \"box_3d\": [1.5, 1.7, 3.77, -2.99, 1.98, 27.23], \"ry\": 1.64}]\n```", - "options": null, - "id": 161 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003158", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [425.55, 188.26, 520.1, 255.88], \"box_3d\": [1.45, 1.53, 3.98, -3.36, 1.87, 18.13], \"ry\": 1.59}]\n```", - "options": null, - "id": 162 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003257", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [803.68, 187.19, 1126.09, 374.0], \"box_3d\": [1.46, 1.53, 3.17, 2.58, 1.61, 6.1], \"ry\": -1.43}, {\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [764.1, 179.0, 825.31, 229.29], \"box_3d\": [1.52, 1.54, 4.5, 6.05, 1.74, 24.15], \"ry\": -1.42}, {\"category\": \"Car\", \"angle\": -2.63, \"box_2d\": [321.86, 189.7, 484.31, 267.41], \"box_3d\": [1.59, 1.45, 3.15, -4.6, 1.99, 16.02], \"ry\": -2.91}]\n```", - "options": null, - "id": 163 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003271", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [275.05, 179.5, 461.66, 317.96], \"box_3d\": [1.7, 1.53, 3.82, -3.44, 1.81, 10.87], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [465.84, 182.44, 532.29, 238.12], \"box_3d\": [1.5, 1.53, 3.38, -3.27, 1.8, 21.5], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [529.99, 176.6, 582.88, 209.72], \"box_3d\": [1.56, 1.56, 3.6, -2.64, 1.76, 35.92], \"ry\": 1.82}]\n```", - "options": null, - "id": 164 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003366", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.22, \"box_2d\": [0.0, 198.46, 118.5, 255.16], \"box_3d\": [1.67, 1.82, 4.42, -19.5, 2.63, 25.02], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [231.67, 192.69, 305.58, 230.48], \"box_3d\": [1.63, 1.87, 3.47, -16.23, 2.62, 34.41], \"ry\": 1.57}]\n```", - "options": null, - "id": 165 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003422", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.32, \"box_2d\": [0.0, 197.5, 28.95, 263.13], \"box_3d\": [1.63, 1.87, 3.47, -16.36, 2.28, 17.36], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [319.42, 189.24, 370.27, 212.66], \"box_3d\": [1.33, 1.72, 3.63, -16.37, 2.37, 44.63], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [466.08, 180.71, 490.01, 196.81], \"box_3d\": [1.6, 1.79, 3.91, -13.61, 2.42, 74.57], \"ry\": 1.56}]\n```", - "options": null, - "id": 166 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003432", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [478.81, 182.48, 536.38, 225.85], \"box_3d\": [1.39, 1.48, 3.76, -3.57, 1.74, 25.42], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [550.16, 172.71, 578.53, 195.23], \"box_3d\": [1.46, 1.75, 4.23, -3.12, 1.48, 49.54], \"ry\": 1.55}]\n```", - "options": null, - "id": 167 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003439", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.18, \"box_2d\": [0.0, 181.99, 315.92, 374.0], \"box_3d\": [1.68, 1.63, 3.9, -3.07, 1.74, 3.68], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": 1.44, \"box_2d\": [353.47, 187.42, 388.36, 214.08], \"box_3d\": [1.65, 1.75, 3.39, -15.8, 2.63, 47.53], \"ry\": 1.12}, {\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [377.84, 188.69, 477.26, 265.0], \"box_3d\": [1.47, 1.7, 4.39, -4.12, 1.87, 16.76], \"ry\": 1.45}]\n```", - "options": null, - "id": 168 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003446", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.6, \"box_2d\": [539.16, 175.02, 555.13, 191.02], \"box_3d\": [1.65, 1.56, 3.98, -6.69, 1.9, 77.0], \"ry\": 1.51}]\n```", - "options": null, - "id": 169 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003549", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [579.3, 176.78, 620.59, 216.07], \"box_3d\": [1.54, 1.63, 3.67, -0.44, 1.73, 30.52], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [597.11, 172.55, 630.62, 207.49], \"box_3d\": [1.77, 1.69, 4.32, 0.17, 1.78, 39.05], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -2.14, \"box_2d\": [914.77, 183.78, 1241.0, 374.0], \"box_3d\": [1.54, 1.55, 3.51, 3.02, 1.62, 3.85], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [686.46, 167.41, 783.22, 249.56], \"box_3d\": [1.74, 1.62, 4.13, 2.85, 1.66, 17.51], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 2.27, \"box_2d\": [0.0, 194.95, 152.6, 332.71], \"box_3d\": [1.51, 1.65, 3.94, -7.81, 1.83, 8.96], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [669.77, 176.42, 724.38, 223.04], \"box_3d\": [1.53, 1.46, 3.26, 2.98, 1.67, 25.49], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.64, \"box_2d\": [645.57, 173.93, 676.76, 202.72], \"box_3d\": [1.6, 1.56, 3.45, 2.93, 1.68, 42.18], \"ry\": -1.57}]\n```", - "options": null, - "id": 170 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003593", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.95, \"box_2d\": [432.04, 178.12, 503.9, 206.01], \"box_3d\": [1.49, 1.64, 3.7, -7.95, 1.81, 40.44], \"ry\": 3.14}]\n```", - "options": null, - "id": 171 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003596", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.62, \"box_2d\": [0.0, 199.87, 239.79, 296.97], \"box_3d\": [1.41, 1.62, 3.6, -8.97, 1.96, 13.25], \"ry\": -1.21}]\n```", - "options": null, - "id": 172 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003604", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [430.35, 181.03, 531.59, 248.73], \"box_3d\": [1.5, 1.7, 3.77, -3.15, 1.72, 18.12], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": -1.53, \"box_2d\": [539.22, 170.31, 564.58, 194.69], \"box_3d\": [1.57, 1.55, 3.32, -3.91, 1.43, 48.6], \"ry\": -1.61}]\n```", - "options": null, - "id": 173 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003651", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [640.81, 173.37, 710.29, 238.53], \"box_3d\": [1.56, 1.65, 3.69, 1.71, 1.6, 19.22], \"ry\": -1.48}, {\"category\": \"Car\", \"angle\": -1.32, \"box_2d\": [942.89, 171.72, 1241.0, 374.0], \"box_3d\": [1.39, 1.61, 3.65, 4.64, 1.39, 6.64], \"ry\": -0.73}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [490.71, 170.73, 556.88, 219.05], \"box_3d\": [1.76, 1.58, 3.68, -3.33, 1.71, 28.23], \"ry\": 1.74}, {\"category\": \"Car\", \"angle\": 2.22, \"box_2d\": [122.96, 184.57, 309.89, 262.38], \"box_3d\": [1.56, 1.62, 4.19, -9.06, 1.86, 16.98], \"ry\": 1.74}, {\"category\": \"Car\", \"angle\": 2.03, \"box_2d\": [403.93, 182.68, 492.14, 220.91], \"box_3d\": [1.42, 1.67, 4.53, -6.58, 1.85, 29.7], \"ry\": 1.82}, {\"category\": \"Car\", \"angle\": -0.93, \"box_2d\": [715.32, 168.61, 774.58, 194.94], \"box_3d\": [1.54, 1.63, 3.74, 8.24, 1.3, 44.04], \"ry\": -0.75}, {\"category\": \"Car\", \"angle\": 2.4, \"box_2d\": [0.0, 191.05, 128.62, 263.7], \"box_3d\": [1.45, 1.5, 4.53, -13.28, 1.9, 16.1], \"ry\": 1.72}]\n```", - "options": null, - "id": 174 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003696", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.31, \"box_2d\": [0.0, 216.47, 68.51, 374.0], \"box_3d\": [1.41, 1.54, 3.6, -6.64, 1.8, 5.59], \"ry\": 3.12}]\n```", - "options": null, - "id": 175 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003747", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 3.11, \"box_2d\": [952.52, 179.27, 1134.28, 239.94], \"box_3d\": [1.54, 1.68, 4.16, 11.86, 1.74, 20.08], \"ry\": -2.64}, {\"category\": \"Car\", \"angle\": -0.79, \"box_2d\": [706.52, 198.71, 788.43, 229.63], \"box_3d\": [1.37, 1.56, 4.19, 6.97, 2.74, 36.55], \"ry\": -0.6}]\n```", - "options": null, - "id": 176 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003748", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [321.4, 180.77, 372.19, 204.2], \"box_3d\": [1.34, 1.73, 3.98, -16.0, 1.84, 43.97], \"ry\": 1.54}]\n```", - "options": null, - "id": 177 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003767", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.25, \"box_2d\": [0.0, 175.11, 95.47, 233.27], \"box_3d\": [1.92, 1.58, 3.93, -20.7, 2.03, 25.89], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.62, \"box_2d\": [553.27, 177.98, 575.0, 197.56], \"box_3d\": [1.66, 1.67, 4.49, -4.05, 2.13, 64.13], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.28, \"box_2d\": [375.24, 180.97, 404.22, 198.49], \"box_3d\": [1.56, 1.72, 3.09, -20.33, 2.32, 66.64], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [584.36, 176.74, 622.1, 214.33], \"box_3d\": [1.62, 1.63, 4.5, -0.33, 1.82, 33.71], \"ry\": -1.57}]\n```", - "options": null, - "id": 178 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003769", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.04, \"box_2d\": [193.63, 191.59, 288.76, 231.91], \"box_3d\": [1.56, 1.91, 4.44, -16.1, 2.43, 31.71], \"ry\": 1.57}]\n```", - "options": null, - "id": 179 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003808", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.03, \"box_2d\": [0.0, 210.77, 372.12, 374.0], \"box_3d\": [1.33, 1.71, 4.09, -2.94, 1.66, 4.29], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [664.81, 177.13, 756.89, 245.0], \"box_3d\": [1.49, 1.56, 4.34, 2.37, 1.62, 18.21], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.37, \"box_2d\": [361.74, 187.54, 495.6, 283.92], \"box_3d\": [1.42, 1.53, 4.12, -3.16, 1.72, 13.17], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.99, \"box_2d\": [768.17, 179.46, 1241.0, 374.0], \"box_3d\": [1.59, 1.59, 3.89, 2.53, 1.66, 5.83], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [632.49, 178.09, 668.23, 208.12], \"box_3d\": [1.38, 1.64, 3.51, 1.96, 1.64, 35.16], \"ry\": -1.5}]\n```", - "options": null, - "id": 180 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003904", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.01, \"box_2d\": [0.0, 202.66, 366.21, 374.0], \"box_3d\": [1.54, 1.5, 3.48, -2.49, 1.74, 3.4], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [242.62, 197.41, 481.3, 364.6], \"box_3d\": [1.44, 1.63, 3.32, -2.63, 1.78, 8.31], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.95, \"box_2d\": [752.67, 190.14, 1023.32, 374.0], \"box_3d\": [1.55, 1.55, 3.27, 2.72, 1.77, 7.95], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [688.05, 188.11, 822.26, 287.92], \"box_3d\": [1.48, 1.6, 3.79, 2.4, 1.79, 13.03], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [497.33, 185.82, 560.97, 238.62], \"box_3d\": [1.56, 1.55, 3.64, -2.61, 2.0, 23.67], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [533.48, 174.06, 584.33, 225.26], \"box_3d\": [1.91, 1.66, 4.22, -2.03, 1.98, 29.19], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [563.63, 176.24, 608.05, 216.44], \"box_3d\": [1.83, 1.69, 4.28, -1.14, 2.01, 35.17], \"ry\": 1.65}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [615.51, 182.74, 652.5, 205.62], \"box_3d\": [1.51, 1.63, 4.03, 1.71, 2.22, 50.61], \"ry\": 1.85}]\n```", - "options": null, - "id": 181 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003928", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [490.6, 188.74, 553.02, 235.61], \"box_3d\": [1.42, 1.66, 4.11, -3.0, 2.0, 24.93], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.59, \"box_2d\": [568.94, 169.48, 589.32, 190.55], \"box_3d\": [1.97, 1.87, 4.53, -2.99, 1.69, 70.43], \"ry\": 1.55}]\n```", - "options": null, - "id": 182 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "003978", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.42, \"box_2d\": [0.0, 219.75, 117.82, 374.0], \"box_3d\": [1.42, 1.62, 3.5, -6.89, 1.92, 6.61], \"ry\": 3.07}, {\"category\": \"Car\", \"angle\": -2.83, \"box_2d\": [257.22, 191.93, 419.89, 251.91], \"box_3d\": [1.41, 1.54, 3.6, -6.92, 1.93, 18.53], \"ry\": 3.1}]\n```", - "options": null, - "id": 183 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "004034", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.15, \"box_2d\": [0.0, 215.06, 295.43, 374.0], \"box_3d\": [1.39, 1.48, 3.76, -3.69, 1.77, 4.9], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [490.47, 184.76, 552.0, 227.9], \"box_3d\": [1.46, 1.75, 4.23, -3.31, 1.94, 27.3], \"ry\": 1.58}]\n```", - "options": null, - "id": 184 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "004043", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.62, \"box_2d\": [434.96, 183.48, 491.75, 240.68], \"box_3d\": [1.68, 1.63, 3.9, -4.76, 2.05, 23.63], \"ry\": 1.42}, {\"category\": \"Car\", \"angle\": 1.46, \"box_2d\": [446.17, 184.9, 488.04, 215.15], \"box_3d\": [1.47, 1.7, 4.39, -7.63, 2.12, 38.11], \"ry\": 1.27}]\n```", - "options": null, - "id": 185 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "004060", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.6, \"box_2d\": [753.33, 190.34, 1207.55, 344.89], \"box_3d\": [1.45, 1.68, 4.53, 4.08, 1.69, 8.33], \"ry\": -0.16}]\n```", - "options": null, - "id": 186 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "004080", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [770.49, 172.76, 822.76, 218.36], \"box_3d\": [1.56, 1.65, 3.69, 6.78, 1.58, 26.62], \"ry\": -1.37}]\n```", - "options": null, - "id": 187 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "004160", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.9, \"box_2d\": [739.31, 179.35, 1006.22, 357.08], \"box_3d\": [1.6, 1.53, 4.18, 2.73, 1.69, 8.66], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [682.77, 180.48, 808.7, 272.05], \"box_3d\": [1.56, 1.57, 3.98, 2.51, 1.73, 14.48], \"ry\": -1.62}, {\"category\": \"Car\", \"angle\": -2.89, \"box_2d\": [402.27, 177.63, 475.77, 210.56], \"box_3d\": [1.61, 1.57, 3.35, -8.72, 1.88, 36.83], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": 1.4, \"box_2d\": [666.86, 168.21, 702.05, 195.27], \"box_3d\": [1.79, 1.72, 4.35, 5.12, 1.51, 50.31], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 1.59, \"box_2d\": [562.64, 176.5, 584.89, 197.74], \"box_3d\": [1.45, 1.49, 4.07, -2.59, 1.73, 51.75], \"ry\": 1.54}]\n```", - "options": null, - "id": 188 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "004183", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [480.23, 195.42, 548.26, 247.84], \"box_3d\": [1.42, 1.66, 4.11, -3.0, 2.19, 23.04], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.59, \"box_2d\": [567.57, 175.67, 588.56, 197.33], \"box_3d\": [1.97, 1.87, 4.53, -3.01, 2.26, 68.6], \"ry\": 1.55}]\n```", - "options": null, - "id": 189 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "004258", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.58, \"box_2d\": [423.25, 188.9, 466.98, 230.21], \"box_3d\": [1.65, 1.75, 3.39, -7.23, 2.38, 31.55], \"ry\": 1.35}]\n```", - "options": null, - "id": 190 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "004272", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -0.59, \"box_2d\": [698.55, 185.81, 1021.06, 319.05], \"box_3d\": [1.41, 1.62, 3.6, 3.08, 1.61, 9.21], \"ry\": -0.28}, {\"category\": \"Car\", \"angle\": -0.47, \"box_2d\": [589.5, 163.98, 888.45, 290.14], \"box_3d\": [1.77, 1.96, 4.39, 2.01, 1.66, 11.82], \"ry\": -0.31}, {\"category\": \"Car\", \"angle\": 1.43, \"box_2d\": [446.21, 185.06, 496.14, 218.49], \"box_3d\": [1.37, 1.6, 4.5, -6.39, 1.94, 32.67], \"ry\": 1.24}, {\"category\": \"Car\", \"angle\": -0.4, \"box_2d\": [585.76, 186.44, 801.44, 266.6], \"box_3d\": [1.4, 1.51, 3.98, 1.6, 1.7, 14.37], \"ry\": -0.29}]\n```", - "options": null, - "id": 191 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "004296", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.57, \"box_2d\": [570.39, 177.69, 620.62, 226.32], \"box_3d\": [1.61, 1.66, 3.2, -0.55, 1.8, 25.75], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 2.12, \"box_2d\": [0.0, 202.41, 327.38, 374.0], \"box_3d\": [1.46, 1.66, 4.05, -3.96, 1.77, 5.92], \"ry\": 1.55}]\n```", - "options": null, - "id": 192 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "004376", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.54, \"box_2d\": [664.57, 178.43, 690.24, 201.85], \"box_3d\": [1.54, 1.63, 3.67, 4.67, 1.94, 49.84], \"ry\": -1.44}, {\"category\": \"Car\", \"angle\": 2.34, \"box_2d\": [0.0, 199.5, 297.3, 374.0], \"box_3d\": [1.53, 1.67, 3.78, -2.99, 1.69, 2.9], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -2.24, \"box_2d\": [939.78, 65.8, 1241.0, 374.0], \"box_3d\": [2.05, 1.5, 3.83, 3.02, 1.67, 3.23], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [736.06, 169.26, 824.07, 249.29], \"box_3d\": [1.88, 1.57, 3.71, 4.32, 1.83, 18.96], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [763.66, 182.97, 884.37, 278.9], \"box_3d\": [1.5, 1.56, 3.69, 3.75, 1.7, 13.34], \"ry\": -1.44}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [717.72, 176.72, 769.97, 222.88], \"box_3d\": [1.7, 1.62, 3.76, 5.22, 1.87, 28.69], \"ry\": -1.5}, {\"category\": \"Car\", \"angle\": -1.61, \"box_2d\": [711.84, 181.56, 742.54, 208.31], \"box_3d\": [1.4, 1.53, 3.56, 6.46, 1.9, 40.16], \"ry\": -1.45}]\n```", - "options": null, - "id": 193 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "004425", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.62, \"box_2d\": [565.33, 157.41, 594.97, 182.13], \"box_3d\": [1.49, 1.68, 4.08, -1.3, 0.02, 45.81], \"ry\": 1.59}]\n```", - "options": null, - "id": 194 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "004444", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.98, \"box_2d\": [770.57, 182.29, 1170.67, 374.0], \"box_3d\": [1.6, 1.53, 4.18, 2.76, 1.7, 6.75], \"ry\": -1.6}, {\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [695.18, 182.16, 845.61, 290.2], \"box_3d\": [1.56, 1.57, 3.98, 2.56, 1.74, 12.63], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -2.88, \"box_2d\": [393.7, 178.52, 471.46, 213.25], \"box_3d\": [1.61, 1.57, 3.35, -8.59, 1.91, 34.95], \"ry\": -3.12}, {\"category\": \"Car\", \"angle\": 1.4, \"box_2d\": [670.89, 168.19, 707.72, 196.36], \"box_3d\": [1.79, 1.72, 4.35, 5.25, 1.52, 48.43], \"ry\": 1.51}, {\"category\": \"Car\", \"angle\": 1.59, \"box_2d\": [563.7, 178.0, 586.51, 200.11], \"box_3d\": [1.45, 1.49, 4.07, -2.4, 1.82, 49.79], \"ry\": 1.54}]\n```", - "options": null, - "id": 195 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "004575", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.1, \"box_2d\": [105.7, 199.36, 201.5, 238.65], \"box_3d\": [1.42, 1.7, 4.14, -19.2, 2.6, 30.52], \"ry\": -1.66}, {\"category\": \"Car\", \"angle\": -1.01, \"box_2d\": [33.03, 198.18, 140.42, 240.21], \"box_3d\": [1.52, 1.71, 4.11, -21.86, 2.65, 30.3], \"ry\": -1.63}]\n```", - "options": null, - "id": 196 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "004760", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [318.9, 171.86, 464.74, 288.42], \"box_3d\": [1.68, 1.63, 3.9, -3.63, 1.7, 12.52], \"ry\": 1.49}, {\"category\": \"Car\", \"angle\": 1.54, \"box_2d\": [433.55, 182.46, 484.78, 226.21], \"box_3d\": [1.47, 1.7, 4.39, -5.72, 1.83, 26.84], \"ry\": 1.33}]\n```", - "options": null, - "id": 197 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "004823", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.5, \"box_2d\": [685.73, 175.8, 714.09, 199.84], \"box_3d\": [1.54, 1.63, 3.67, 6.06, 1.76, 48.57], \"ry\": -1.38}, {\"category\": \"Car\", \"angle\": -2.08, \"box_2d\": [834.93, 181.39, 1241.0, 374.0], \"box_3d\": [1.58, 1.67, 3.84, 3.1, 1.66, 5.38], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [477.95, 182.27, 574.62, 250.53], \"box_3d\": [1.53, 1.67, 3.78, -2.04, 1.78, 18.26], \"ry\": 1.66}, {\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [777.42, 139.72, 968.06, 321.21], \"box_3d\": [2.05, 1.5, 3.83, 3.39, 1.71, 10.13], \"ry\": -1.5}]\n```", - "options": null, - "id": 198 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "004873", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.81, \"box_2d\": [712.09, 185.45, 879.29, 292.9], \"box_3d\": [1.41, 1.58, 4.36, 2.78, 1.64, 11.96], \"ry\": -1.58}]\n```", - "options": null, - "id": 199 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "004886", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [491.83, 181.76, 569.36, 232.09], \"box_3d\": [1.44, 1.68, 4.39, -2.47, 1.74, 23.14], \"ry\": 1.65}]\n```", - "options": null, - "id": 200 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "004890", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.58, \"box_2d\": [575.94, 174.75, 617.13, 211.51], \"box_3d\": [1.41, 1.59, 4.47, -0.59, 1.52, 30.32], \"ry\": 1.56}]\n```", - "options": null, - "id": 201 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "005036", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [430.5, 182.6, 536.18, 264.85], \"box_3d\": [1.46, 1.63, 4.47, -2.58, 1.69, 15.37], \"ry\": 1.53}, {\"category\": \"Car\", \"angle\": 1.74, \"box_2d\": [512.4, 171.14, 581.72, 221.59], \"box_3d\": [1.7, 1.82, 4.45, -2.26, 1.67, 26.7], \"ry\": 1.65}, {\"category\": \"Car\", \"angle\": -0.95, \"box_2d\": [629.27, 173.17, 679.67, 202.38], \"box_3d\": [1.57, 1.56, 2.62, 2.48, 1.6, 40.34], \"ry\": -0.89}]\n```", - "options": null, - "id": 202 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "005048", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [805.07, 172.46, 879.03, 235.74], \"box_3d\": [1.56, 1.65, 3.69, 6.21, 1.57, 19.75], \"ry\": -1.33}, {\"category\": \"Car\", \"angle\": 2.17, \"box_2d\": [263.48, 186.52, 473.27, 274.26], \"box_3d\": [1.42, 1.67, 4.53, -4.63, 1.73, 14.51], \"ry\": 1.87}, {\"category\": \"Car\", \"angle\": -0.44, \"box_2d\": [1223.73, 163.63, 1241.0, 202.89], \"box_3d\": [1.54, 1.63, 3.74, 26.64, 1.16, 28.18], \"ry\": 0.32}]\n```", - "options": null, - "id": 203 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "005144", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.75, \"box_2d\": [700.79, 184.53, 806.65, 263.98], \"box_3d\": [1.51, 1.56, 3.81, 3.0, 1.79, 15.95], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [676.66, 179.43, 756.56, 239.12], \"box_3d\": [1.57, 1.68, 3.83, 2.99, 1.78, 21.17], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 1.6, \"box_2d\": [520.17, 178.79, 555.81, 213.98], \"box_3d\": [1.52, 1.5, 3.85, -3.32, 1.8, 33.38], \"ry\": 1.5}]\n```", - "options": null, - "id": 204 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "005309", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [591.63, 172.71, 623.07, 203.78], \"box_3d\": [1.65, 1.67, 3.64, -0.14, 1.67, 40.51], \"ry\": -1.55}]\n```", - "options": null, - "id": 205 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "005345", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.23, \"box_2d\": [748.14, 194.28, 1077.63, 374.0], \"box_3d\": [1.42, 1.59, 4.19, 2.64, 1.69, 7.53], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.93, \"box_2d\": [145.69, 196.52, 425.53, 366.97], \"box_3d\": [1.42, 1.64, 3.63, -3.41, 1.74, 8.25], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [692.97, 186.37, 817.83, 273.58], \"box_3d\": [1.38, 1.58, 3.66, 2.54, 1.66, 13.61], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.43, \"box_2d\": [669.59, 180.35, 755.48, 243.0], \"box_3d\": [1.42, 1.6, 3.78, 2.47, 1.62, 18.42], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 2.73, \"box_2d\": [700.35, 171.72, 805.98, 210.76], \"box_3d\": [1.57, 1.65, 4.1, 6.08, 1.55, 30.92], \"ry\": 2.93}]\n```", - "options": null, - "id": 206 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "005346", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [323.32, 185.03, 499.81, 307.44], \"box_3d\": [1.6, 1.68, 3.94, -3.0, 1.81, 11.61], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [542.45, 181.48, 577.43, 208.56], \"box_3d\": [1.42, 1.66, 4.11, -2.79, 1.92, 40.54], \"ry\": 1.57}]\n```", - "options": null, - "id": 207 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "005397", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.34, \"box_2d\": [346.51, 184.1, 464.99, 218.44], \"box_3d\": [1.36, 1.68, 4.49, -8.59, 1.86, 30.41], \"ry\": 0.07}]\n```", - "options": null, - "id": 208 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "005444", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (718.86, 718.86), (cx, cy) = (607.19, 185.22). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [716.43, 170.65, 759.72, 209.19], \"box_3d\": [1.6, 1.76, 3.84, 5.71, 0.99, 31.98], \"ry\": -1.42}, {\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [701.02, 175.78, 724.03, 195.77], \"box_3d\": [1.39, 1.56, 3.45, 7.54, 0.72, 52.04], \"ry\": -1.45}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [691.65, 175.99, 713.34, 189.93], \"box_3d\": [1.41, 1.66, 3.76, 9.92, 0.46, 75.47], \"ry\": -1.6}]\n```", - "options": null, - "id": 209 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "005528", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [600.65, 174.79, 628.08, 201.73], \"box_3d\": [1.65, 1.67, 3.64, 0.22, 1.8, 46.47], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": 2.12, \"box_2d\": [83.56, 199.55, 254.53, 271.12], \"box_3d\": [1.32, 1.6, 3.63, -9.75, 1.98, 16.23], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [407.53, 186.04, 455.28, 213.06], \"box_3d\": [1.33, 1.46, 3.87, -9.54, 2.06, 38.68], \"ry\": 1.61}]\n```", - "options": null, - "id": 210 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "005577", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.36, \"box_2d\": [0.0, 203.63, 255.73, 374.0], \"box_3d\": [1.5, 1.7, 3.77, -3.47, 1.71, 3.28], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -1.49, \"box_2d\": [478.3, 182.25, 517.89, 218.36], \"box_3d\": [1.57, 1.55, 3.32, -5.19, 2.02, 33.52], \"ry\": -1.65}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [488.51, 180.44, 526.53, 210.16], \"box_3d\": [1.56, 1.68, 3.91, -5.68, 1.99, 40.19], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": -3.02, \"box_2d\": [288.91, 173.66, 421.48, 224.01], \"box_3d\": [1.77, 1.73, 4.53, -9.55, 1.83, 27.27], \"ry\": 2.93}]\n```", - "options": null, - "id": 211 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "005672", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [11.25, 183.09, 419.37, 374.0], \"box_3d\": [1.46, 1.63, 4.47, -3.05, 1.57, 6.65], \"ry\": 1.49}, {\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [411.66, 160.09, 527.36, 239.65], \"box_3d\": [1.7, 1.82, 4.45, -3.33, 1.45, 17.76], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": -1.27, \"box_2d\": [646.39, 161.48, 691.89, 194.02], \"box_3d\": [1.57, 1.56, 2.62, 2.99, 1.04, 36.45], \"ry\": -1.19}]\n```", - "options": null, - "id": 212 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "005677", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.04, \"box_2d\": [0.0, 202.76, 405.34, 374.0], \"box_3d\": [1.46, 1.6, 3.33, -2.65, 1.72, 4.76], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": -2.04, \"box_2d\": [809.69, 177.53, 1241.0, 374.0], \"box_3d\": [1.64, 1.7, 3.54, 2.65, 1.68, 4.85], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -1.83, \"box_2d\": [733.25, 194.21, 928.23, 327.91], \"box_3d\": [1.37, 1.51, 3.39, 2.64, 1.69, 9.51], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [518.57, 181.98, 569.39, 224.02], \"box_3d\": [1.55, 1.68, 4.42, -2.64, 1.93, 29.27], \"ry\": 1.56}]\n```", - "options": null, - "id": 213 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "005741", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.98, \"box_2d\": [758.74, 171.72, 1083.69, 371.34], \"box_3d\": [1.72, 1.67, 4.42, 3.16, 1.74, 8.54], \"ry\": -1.64}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [391.59, 195.67, 512.77, 293.08], \"box_3d\": [1.44, 1.56, 3.84, -2.8, 1.91, 13.32], \"ry\": 1.52}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [515.35, 185.06, 555.71, 224.89], \"box_3d\": [1.5, 1.55, 4.21, -3.11, 2.03, 30.11], \"ry\": -1.66}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [636.74, 176.51, 686.07, 223.35], \"box_3d\": [1.82, 1.65, 3.67, 2.05, 1.99, 30.12], \"ry\": -1.61}]\n```", - "options": null, - "id": 214 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "005756", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [610.64, 175.32, 641.39, 196.16], \"box_3d\": [1.44, 1.68, 4.39, 1.19, 1.64, 52.43], \"ry\": 1.72}]\n```", - "options": null, - "id": 215 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "005767", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [793.15, 180.84, 825.73, 209.76], \"box_3d\": [1.43, 1.55, 3.91, 10.44, 1.86, 38.02], \"ry\": -1.32}]\n```", - "options": null, - "id": 216 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "005793", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.01, \"box_2d\": [53.01, 224.61, 381.44, 374.0], \"box_3d\": [1.27, 1.54, 3.58, -3.77, 1.93, 7.58], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [330.26, 181.69, 473.38, 283.18], \"box_3d\": [1.63, 1.65, 3.91, -3.82, 1.81, 13.74], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [614.42, 176.53, 656.13, 214.75], \"box_3d\": [1.46, 1.53, 3.17, 0.95, 1.63, 29.45], \"ry\": -1.59}]\n```", - "options": null, - "id": 217 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "005794", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.76, \"box_2d\": [689.79, 176.05, 822.8, 270.98], \"box_3d\": [1.49, 1.56, 4.34, 2.52, 1.57, 13.63], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": -1.28, \"box_2d\": [203.73, 186.15, 455.62, 348.28], \"box_3d\": [1.42, 1.53, 4.12, -3.04, 1.61, 8.64], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [644.18, 179.61, 685.83, 214.54], \"box_3d\": [1.38, 1.64, 3.51, 2.26, 1.67, 30.47], \"ry\": -1.51}]\n```", - "options": null, - "id": 218 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "005802", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [631.14, 178.37, 676.2, 218.01], \"box_3d\": [1.37, 1.56, 3.54, 1.6, 1.59, 26.99], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": -1.46, \"box_2d\": [574.03, 170.1, 620.21, 206.84], \"box_3d\": [1.61, 1.69, 4.26, -0.57, 1.52, 34.04], \"ry\": -1.47}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [645.36, 170.51, 679.59, 202.47], \"box_3d\": [1.53, 1.63, 4.5, 2.69, 1.45, 37.14], \"ry\": -1.49}, {\"category\": \"Car\", \"angle\": -0.74, \"box_2d\": [971.36, 151.1, 1090.23, 191.53], \"box_3d\": [1.48, 1.6, 3.84, 16.05, 0.68, 27.63], \"ry\": -0.21}, {\"category\": \"Car\", \"angle\": -0.67, \"box_2d\": [926.55, 153.32, 1028.08, 191.19], \"box_3d\": [1.55, 1.61, 3.64, 15.64, 0.75, 30.83], \"ry\": -0.21}]\n```", - "options": null, - "id": 219 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "005875", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [470.47, 175.29, 527.4, 209.79], \"box_3d\": [1.42, 1.78, 3.62, -4.47, 1.17, 31.75], \"ry\": 1.63}]\n```", - "options": null, - "id": 220 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "005889", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.62, \"box_2d\": [825.14, 172.9, 937.58, 284.1], \"box_3d\": [1.66, 1.56, 3.42, 4.56, 1.68, 12.61], \"ry\": -1.29}, {\"category\": \"Car\", \"angle\": 1.71, \"box_2d\": [657.52, 178.62, 707.1, 213.07], \"box_3d\": [1.45, 1.64, 4.21, 3.3, 1.72, 32.64], \"ry\": 1.81}, {\"category\": \"Car\", \"angle\": 1.54, \"box_2d\": [804.04, 176.09, 835.85, 206.06], \"box_3d\": [1.62, 1.58, 3.89, 11.87, 1.81, 41.02], \"ry\": 1.82}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [706.05, 178.34, 739.05, 200.77], \"box_3d\": [1.45, 1.67, 4.43, 7.69, 1.83, 49.16], \"ry\": 1.85}]\n```", - "options": null, - "id": 221 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "005890", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 0.6, \"box_2d\": [0.0, 198.63, 400.32, 333.83], \"box_3d\": [1.38, 1.66, 4.38, -5.0, 1.73, 8.64], \"ry\": 0.09}, {\"category\": \"Car\", \"angle\": -0.76, \"box_2d\": [982.15, 171.24, 1241.0, 374.0], \"box_3d\": [1.48, 1.86, 4.18, 4.68, 1.46, 4.14], \"ry\": 0.06}, {\"category\": \"Car\", \"angle\": -0.57, \"box_2d\": [838.98, 148.44, 1241.0, 342.79], \"box_3d\": [1.74, 1.66, 4.37, 4.78, 1.53, 7.44], \"ry\": -0.02}, {\"category\": \"Car\", \"angle\": -2.77, \"box_2d\": [320.03, 181.75, 500.06, 240.79], \"box_3d\": [1.47, 1.67, 4.28, -5.27, 1.73, 19.13], \"ry\": -3.04}, {\"category\": \"Car\", \"angle\": -3.03, \"box_2d\": [393.18, 180.36, 529.92, 224.1], \"box_3d\": [1.38, 1.67, 4.37, -4.98, 1.65, 24.47], \"ry\": 3.05}, {\"category\": \"Car\", \"angle\": -0.34, \"box_2d\": [793.96, 168.65, 985.13, 251.75], \"box_3d\": [1.54, 1.58, 3.2, 5.45, 1.49, 14.44], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": -0.41, \"box_2d\": [808.13, 168.59, 1095.68, 266.97], \"box_3d\": [1.51, 1.72, 4.03, 5.57, 1.47, 12.22], \"ry\": 0.01}]\n```", - "options": null, - "id": 222 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "005895", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [489.62, 180.04, 560.35, 215.68], \"box_3d\": [1.37, 1.78, 4.65, -3.51, 1.69, 30.43], \"ry\": 1.72}]\n```", - "options": null, - "id": 223 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "005951", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [499.38, 184.13, 556.94, 226.74], \"box_3d\": [1.42, 1.66, 4.11, -2.99, 1.86, 26.77], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.59, \"box_2d\": [570.0, 166.37, 589.81, 186.89], \"box_3d\": [1.97, 1.87, 4.53, -2.99, 1.38, 72.26], \"ry\": 1.55}]\n```", - "options": null, - "id": 224 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "005996", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.39, \"box_2d\": [1098.87, 213.82, 1241.0, 374.0], \"box_3d\": [1.34, 1.52, 3.99, 6.42, 1.69, 5.76], \"ry\": -3.08}, {\"category\": \"Car\", \"angle\": -0.53, \"box_2d\": [985.57, 177.76, 1241.0, 334.86], \"box_3d\": [1.63, 1.53, 3.45, 6.41, 1.69, 8.28], \"ry\": 0.11}, {\"category\": \"Car\", \"angle\": 2.74, \"box_2d\": [790.76, 167.67, 1019.61, 253.08], \"box_3d\": [1.76, 1.71, 4.4, 6.44, 1.68, 16.14], \"ry\": 3.12}, {\"category\": \"Car\", \"angle\": 0.49, \"box_2d\": [63.48, 135.07, 354.91, 275.99], \"box_3d\": [2.22, 1.68, 4.08, -6.78, 1.64, 12.42], \"ry\": -0.0}, {\"category\": \"Car\", \"angle\": -2.72, \"box_2d\": [178.96, 179.73, 368.76, 252.76], \"box_3d\": [1.4, 1.46, 3.23, -6.89, 1.56, 14.9], \"ry\": 3.13}, {\"category\": \"Car\", \"angle\": -2.86, \"box_2d\": [312.45, 177.97, 469.92, 229.3], \"box_3d\": [1.46, 1.61, 4.31, -6.62, 1.64, 21.99], \"ry\": 3.14}, {\"category\": \"Car\", \"angle\": -3.04, \"box_2d\": [381.5, 167.74, 506.8, 218.95], \"box_3d\": [1.8, 1.77, 4.41, -6.16, 1.65, 27.05], \"ry\": 3.02}]\n```", - "options": null, - "id": 225 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006016", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [388.85, 187.64, 529.68, 273.48], \"box_3d\": [1.44, 1.64, 3.78, -2.89, 1.77, 14.43], \"ry\": 1.69}, {\"category\": \"Car\", \"angle\": 1.82, \"box_2d\": [461.82, 173.54, 568.58, 251.13], \"box_3d\": [1.77, 1.71, 4.28, -2.35, 1.82, 18.71], \"ry\": 1.69}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [726.15, 179.91, 788.23, 222.09], \"box_3d\": [1.45, 1.71, 3.93, 5.39, 1.72, 27.01], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [544.58, 175.61, 603.34, 223.9], \"box_3d\": [1.73, 1.65, 3.96, -1.35, 1.85, 27.96], \"ry\": 1.68}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [713.5, 172.02, 766.47, 213.79], \"box_3d\": [1.78, 1.89, 4.37, 5.86, 1.77, 33.17], \"ry\": -1.51}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [569.32, 175.96, 611.75, 214.98], \"box_3d\": [1.68, 1.51, 3.29, -0.87, 1.83, 32.86], \"ry\": 1.67}, {\"category\": \"Car\", \"angle\": -2.41, \"box_2d\": [0.0, 180.52, 79.86, 302.92], \"box_3d\": [1.5, 1.63, 4.14, -9.63, 1.62, 9.35], \"ry\": 3.09}]\n```", - "options": null, - "id": 226 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006019", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [529.17, 180.26, 559.49, 205.28], \"box_3d\": [1.45, 1.53, 3.98, -4.02, 1.92, 44.38], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.77, \"box_2d\": [429.96, 179.92, 513.19, 243.78], \"box_3d\": [1.65, 1.56, 3.98, -3.91, 1.86, 20.76], \"ry\": 1.59}]\n```", - "options": null, - "id": 227 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006025", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [708.34, 188.54, 881.63, 305.68], \"box_3d\": [1.41, 1.64, 3.77, 2.54, 1.68, 10.96], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [487.22, 186.35, 534.47, 217.72], \"box_3d\": [1.43, 1.76, 3.97, -4.93, 2.14, 36.12], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.9, \"box_2d\": [294.15, 196.3, 435.53, 282.07], \"box_3d\": [1.49, 1.61, 3.94, -5.07, 2.04, 15.34], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [378.86, 191.99, 477.99, 253.02], \"box_3d\": [1.48, 1.66, 3.9, -5.04, 2.06, 20.38], \"ry\": 1.61}, {\"category\": \"Car\", \"angle\": 2.22, \"box_2d\": [0.0, 220.97, 261.61, 374.0], \"box_3d\": [1.31, 1.58, 4.23, -5.46, 1.93, 7.35], \"ry\": 1.6}]\n```", - "options": null, - "id": 228 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006062", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.65, \"box_2d\": [619.72, 178.54, 714.02, 263.92], \"box_3d\": [1.46, 1.53, 3.17, 0.96, 1.58, 14.07], \"ry\": -1.58}]\n```", - "options": null, - "id": 229 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006072", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.27, \"box_2d\": [0.0, 183.86, 78.18, 239.95], \"box_3d\": [1.46, 1.4, 3.27, -16.8, 1.79, 20.12], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 2.2, \"box_2d\": [62.17, 171.1, 197.67, 228.53], \"box_3d\": [1.78, 1.6, 4.29, -16.24, 1.75, 24.64], \"ry\": 1.63}, {\"category\": \"Car\", \"angle\": 2.14, \"box_2d\": [182.71, 180.48, 270.47, 219.65], \"box_3d\": [1.5, 1.47, 3.62, -15.75, 1.83, 29.77], \"ry\": 1.65}]\n```", - "options": null, - "id": 230 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006142", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.59, \"box_2d\": [525.53, 180.74, 574.38, 224.33], \"box_3d\": [1.41, 1.59, 4.47, -2.15, 1.71, 25.96], \"ry\": 1.51}]\n```", - "options": null, - "id": 231 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006156", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.54, \"box_2d\": [0.0, 161.31, 165.84, 315.47], \"box_3d\": [1.61, 1.46, 2.33, -6.76, 1.51, 8.27], \"ry\": 3.08}, {\"category\": \"Car\", \"angle\": 0.44, \"box_2d\": [22.46, 175.55, 384.96, 289.95], \"box_3d\": [1.5, 1.66, 4.4, -5.83, 1.57, 10.68], \"ry\": -0.05}, {\"category\": \"Car\", \"angle\": -1.85, \"box_2d\": [731.6, 186.28, 965.43, 331.52], \"box_3d\": [1.41, 1.62, 4.19, 2.71, 1.61, 9.36], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": -0.34, \"box_2d\": [842.74, 180.88, 1043.6, 258.72], \"box_3d\": [1.42, 1.59, 3.21, 6.39, 1.6, 14.2], \"ry\": 0.07}, {\"category\": \"Car\", \"angle\": 2.86, \"box_2d\": [759.16, 178.74, 955.42, 240.36], \"box_3d\": [1.47, 1.77, 4.42, 6.18, 1.64, 18.52], \"ry\": -3.11}, {\"category\": \"Car\", \"angle\": -2.88, \"box_2d\": [303.89, 171.19, 465.79, 228.51], \"box_3d\": [1.54, 1.66, 4.16, -6.45, 1.53, 20.84], \"ry\": 3.11}, {\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [633.81, 175.41, 659.61, 194.47], \"box_3d\": [1.32, 1.63, 4.1, 2.62, 1.53, 52.61], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 0.14, \"box_2d\": [479.56, 170.21, 549.39, 195.94], \"box_3d\": [1.58, 1.56, 4.25, -6.11, 1.44, 46.26], \"ry\": 0.01}, {\"category\": \"Car\", \"angle\": -0.18, \"box_2d\": [684.41, 175.65, 763.49, 205.5], \"box_3d\": [1.5, 1.62, 3.88, 5.95, 1.67, 38.09], \"ry\": -0.02}, {\"category\": \"Car\", \"angle\": 0.29, \"box_2d\": [409.61, 173.98, 518.35, 214.44], \"box_3d\": [1.5, 1.49, 3.86, -5.66, 1.57, 28.01], \"ry\": 0.09}, {\"category\": \"Car\", \"angle\": -0.3, \"box_2d\": [727.04, 173.0, 826.69, 214.47], \"box_3d\": [1.55, 1.58, 3.51, 6.52, 1.58, 28.51], \"ry\": -0.07}]\n```", - "options": null, - "id": 232 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006206", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.62, \"box_2d\": [527.47, 175.27, 558.1, 203.49], \"box_3d\": [1.7, 1.72, 4.11, -4.26, 1.88, 45.99], \"ry\": 1.53}]\n```", - "options": null, - "id": 233 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006209", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [531.01, 182.24, 565.63, 209.27], \"box_3d\": [1.52, 1.73, 4.28, -3.69, 2.1, 43.52], \"ry\": 1.57}]\n```", - "options": null, - "id": 234 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006234", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.8, \"box_2d\": [387.16, 191.36, 506.47, 274.84], \"box_3d\": [1.45, 1.53, 3.98, -3.32, 1.88, 15.21], \"ry\": 1.59}]\n```", - "options": null, - "id": 235 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006314", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.37, \"box_2d\": [0.0, 225.52, 134.43, 374.0], \"box_3d\": [1.42, 1.55, 3.85, -5.93, 1.97, 5.75], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 2.06, \"box_2d\": [63.97, 205.21, 311.58, 335.21], \"box_3d\": [1.43, 1.48, 3.71, -5.92, 1.97, 10.58], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": 1.91, \"box_2d\": [271.24, 196.16, 409.06, 271.43], \"box_3d\": [1.4, 1.65, 4.02, -6.01, 1.99, 16.45], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 2.52, \"box_2d\": [1000.82, 135.93, 1241.0, 241.17], \"box_3d\": [1.64, 1.59, 3.99, 9.03, 1.07, 12.23], \"ry\": -3.14}, {\"category\": \"Car\", \"angle\": 2.58, \"box_2d\": [961.34, 137.05, 1148.13, 227.09], \"box_3d\": [1.72, 1.68, 2.78, 9.01, 1.05, 14.85], \"ry\": 3.12}, {\"category\": \"Car\", \"angle\": 2.66, \"box_2d\": [889.38, 145.73, 1070.36, 221.29], \"box_3d\": [1.67, 1.67, 3.43, 8.66, 1.09, 17.16], \"ry\": 3.12}, {\"category\": \"Car\", \"angle\": 2.75, \"box_2d\": [829.45, 154.51, 987.62, 214.99], \"box_3d\": [1.57, 1.62, 3.69, 8.15, 1.11, 19.98], \"ry\": 3.13}, {\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [389.49, 188.9, 474.76, 237.88], \"box_3d\": [1.43, 1.61, 4.13, -5.84, 1.99, 24.01], \"ry\": 1.62}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [692.86, 175.82, 789.9, 241.68], \"box_3d\": [1.47, 1.6, 3.87, 3.16, 1.57, 18.26], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.7, \"box_2d\": [676.1, 173.6, 741.07, 226.18], \"box_3d\": [1.57, 1.57, 4.04, 3.13, 1.62, 23.81], \"ry\": -1.57}]\n```", - "options": null, - "id": 236 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006327", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.85, \"box_2d\": [321.2, 187.28, 501.22, 296.68], \"box_3d\": [1.44, 1.68, 4.39, -3.1, 1.71, 12.04], \"ry\": 1.6}]\n```", - "options": null, - "id": 237 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006352", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.95, \"box_2d\": [225.42, 187.24, 405.49, 281.2], \"box_3d\": [1.51, 1.81, 4.07, -5.55, 1.82, 14.09], \"ry\": 1.58}]\n```", - "options": null, - "id": 238 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006359", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.57, \"box_2d\": [586.32, 178.28, 608.96, 198.48], \"box_3d\": [1.41, 1.59, 4.47, -0.95, 1.83, 53.34], \"ry\": 1.55}]\n```", - "options": null, - "id": 239 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006391", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -2.21, \"box_2d\": [64.72, 209.46, 382.83, 364.74], \"box_3d\": [1.47, 1.58, 3.55, -4.73, 1.95, 8.71], \"ry\": -2.7}, {\"category\": \"Car\", \"angle\": -1.46, \"box_2d\": [770.17, 169.28, 876.08, 255.98], \"box_3d\": [1.6, 1.75, 3.89, 4.62, 1.55, 15.42], \"ry\": -1.18}, {\"category\": \"Car\", \"angle\": -1.9, \"box_2d\": [1058.85, 154.32, 1241.0, 266.96], \"box_3d\": [1.59, 1.64, 4.03, 9.11, 1.35, 12.21], \"ry\": -1.27}, {\"category\": \"Car\", \"angle\": -2.02, \"box_2d\": [0.0, 226.58, 166.93, 349.11], \"box_3d\": [1.37, 1.7, 4.38, -8.97, 2.28, 10.89], \"ry\": -2.7}, {\"category\": \"Car\", \"angle\": -1.72, \"box_2d\": [1017.51, 156.72, 1137.93, 225.21], \"box_3d\": [1.5, 1.8, 4.5, 11.57, 1.17, 18.22], \"ry\": -1.16}, {\"category\": \"Car\", \"angle\": -1.6, \"box_2d\": [985.96, 153.04, 1044.02, 204.59], \"box_3d\": [1.83, 1.75, 4.52, 15.56, 1.15, 27.94], \"ry\": -1.1}]\n```", - "options": null, - "id": 240 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006429", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.92, \"box_2d\": [166.86, 181.05, 476.48, 374.0], \"box_3d\": [1.63, 1.6, 3.73, -2.6, 1.72, 7.36], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -2.03, \"box_2d\": [794.85, 188.67, 1241.0, 374.0], \"box_3d\": [1.59, 1.49, 3.34, 2.58, 1.74, 5.43], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.8, \"box_2d\": [692.66, 188.36, 861.73, 315.27], \"box_3d\": [1.54, 1.6, 3.52, 2.26, 1.8, 10.83], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": -1.42, \"box_2d\": [386.58, 181.38, 506.39, 284.9], \"box_3d\": [1.54, 1.5, 3.48, -2.76, 1.7, 12.64], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [440.21, 184.16, 529.65, 250.39], \"box_3d\": [1.44, 1.63, 3.32, -3.02, 1.74, 17.77], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.78, \"box_2d\": [666.69, 180.26, 760.67, 251.56], \"box_3d\": [1.55, 1.55, 3.27, 2.38, 1.74, 17.53], \"ry\": -1.65}, {\"category\": \"Car\", \"angle\": -1.69, \"box_2d\": [643.57, 178.94, 710.66, 231.45], \"box_3d\": [1.48, 1.6, 3.79, 1.97, 1.69, 22.52], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [515.92, 178.33, 560.04, 214.75], \"box_3d\": [1.56, 1.55, 3.64, -3.27, 1.82, 32.97], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [538.31, 169.65, 576.49, 207.65], \"box_3d\": [1.91, 1.66, 4.22, -2.79, 1.77, 38.61], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [556.27, 172.53, 591.03, 203.88], \"box_3d\": [1.83, 1.69, 4.28, -2.22, 1.84, 44.53], \"ry\": 1.62}]\n```", - "options": null, - "id": 241 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006554", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.63, \"box_2d\": [707.23, 168.35, 729.04, 185.77], \"box_3d\": [1.4, 1.55, 4.06, 9.08, 1.07, 60.87], \"ry\": -1.48}]\n```", - "options": null, - "id": 242 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006584", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [363.31, 180.35, 475.8, 261.1], \"box_3d\": [1.7, 1.72, 4.11, -4.48, 1.89, 17.43], \"ry\": 1.54}]\n```", - "options": null, - "id": 243 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006607", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.6, \"box_2d\": [460.12, 187.4, 519.59, 241.59], \"box_3d\": [1.41, 1.59, 4.47, -3.61, 1.88, 21.89], \"ry\": 1.44}]\n```", - "options": null, - "id": 244 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006617", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (707.05, 707.05), (cx, cy) = (604.08, 180.51). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.58, \"box_2d\": [526.65, 162.16, 567.82, 200.96], \"box_3d\": [1.5, 1.59, 3.6, -2.41, 0.78, 29.51], \"ry\": 1.5}, {\"category\": \"Car\", \"angle\": 1.61, \"box_2d\": [547.1, 166.77, 579.72, 193.84], \"box_3d\": [1.36, 1.57, 4.06, -2.19, 0.66, 37.97], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.82, \"box_2d\": [581.35, 167.14, 609.55, 183.07], \"box_3d\": [1.5, 1.7, 4.45, -0.94, 0.23, 69.62], \"ry\": -1.83}]\n```", - "options": null, - "id": 245 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006659", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.98, \"box_2d\": [799.24, 203.87, 1220.71, 374.0], \"box_3d\": [1.37, 1.51, 3.39, 2.67, 1.68, 5.79], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [507.64, 179.42, 566.98, 228.03], \"box_3d\": [1.55, 1.68, 4.42, -2.51, 1.79, 25.48], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.68, \"box_2d\": [657.48, 178.46, 702.04, 216.04], \"box_3d\": [1.6, 1.53, 4.18, 3.1, 1.87, 33.06], \"ry\": -1.59}]\n```", - "options": null, - "id": 246 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006691", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [448.97, 184.6, 539.26, 243.12], \"box_3d\": [1.48, 1.87, 4.42, -3.28, 1.84, 21.0], \"ry\": 1.58}]\n```", - "options": null, - "id": 247 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006733", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.12, \"box_2d\": [0.0, 212.8, 339.5, 374.0], \"box_3d\": [1.41, 1.76, 4.4, -3.66, 1.81, 5.25], \"ry\": 1.54}, {\"category\": \"Car\", \"angle\": 2.38, \"box_2d\": [0.0, 220.88, 70.46, 374.0], \"box_3d\": [1.43, 1.7, 3.74, -6.67, 1.94, 5.87], \"ry\": 1.56}, {\"category\": \"Car\", \"angle\": -1.67, \"box_2d\": [654.1, 179.32, 709.89, 223.85], \"box_3d\": [1.55, 1.66, 4.15, 2.62, 1.8, 27.41], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.87, \"box_2d\": [292.86, 194.36, 454.13, 295.55], \"box_3d\": [1.41, 1.66, 3.35, -3.91, 1.82, 12.35], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 2.77, \"box_2d\": [805.7, 165.78, 996.06, 222.26], \"box_3d\": [1.47, 1.7, 4.65, 8.03, 1.32, 20.29], \"ry\": -3.14}, {\"category\": \"Car\", \"angle\": 2.82, \"box_2d\": [786.12, 153.89, 940.81, 216.89], \"box_3d\": [1.91, 1.81, 4.33, 8.05, 1.35, 23.28], \"ry\": -3.14}, {\"category\": \"Car\", \"angle\": -1.37, \"box_2d\": [431.2, 183.4, 486.79, 224.15], \"box_3d\": [1.63, 1.61, 3.81, -6.53, 2.11, 31.39], \"ry\": -1.57}, {\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [336.89, 192.01, 431.93, 249.74], \"box_3d\": [1.43, 1.57, 3.48, -6.35, 2.01, 20.52], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 2.9, \"box_2d\": [745.32, 164.58, 862.58, 206.86], \"box_3d\": [1.63, 1.67, 4.32, 7.81, 1.33, 29.44], \"ry\": -3.13}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [478.07, 184.49, 516.12, 212.15], \"box_3d\": [1.53, 1.68, 3.86, -6.69, 2.24, 42.9], \"ry\": 1.56}]\n```", - "options": null, - "id": 248 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006735", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.24, \"box_2d\": [235.06, 186.69, 490.27, 358.89], \"box_3d\": [1.77, 1.75, 4.67, -3.06, 1.99, 9.96], \"ry\": -1.53}, {\"category\": \"Car\", \"angle\": 1.73, \"box_2d\": [476.29, 189.64, 556.59, 253.18], \"box_3d\": [1.6, 1.68, 3.94, -2.64, 2.12, 20.84], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.63, \"box_2d\": [569.48, 189.33, 597.28, 211.86], \"box_3d\": [1.42, 1.66, 4.11, -1.82, 2.59, 49.75], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": -0.32, \"box_2d\": [1145.02, 114.32, 1241.0, 276.27], \"box_3d\": [1.66, 1.15, 3.39, 7.66, 1.05, 7.39], \"ry\": 0.47}]\n```", - "options": null, - "id": 249 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006743", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.55, \"box_2d\": [561.94, 176.74, 613.3, 226.46], \"box_3d\": [1.61, 1.66, 3.2, -0.79, 1.76, 25.22], \"ry\": -1.58}, {\"category\": \"Car\", \"angle\": 1.67, \"box_2d\": [533.36, 179.46, 551.8, 195.58], \"box_3d\": [1.66, 1.61, 3.46, -7.15, 2.37, 76.74], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.94, \"box_2d\": [841.41, 169.72, 884.2, 189.42], \"box_3d\": [1.36, 1.5, 4.21, 18.31, 1.17, 52.49], \"ry\": -1.6}]\n```", - "options": null, - "id": 250 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006774", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [478.67, 172.98, 545.2, 232.89], \"box_3d\": [1.97, 1.87, 4.53, -3.5, 2.01, 26.25], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.66, \"box_2d\": [543.26, 180.21, 570.52, 205.79], \"box_3d\": [1.7, 1.53, 3.82, -3.7, 2.23, 50.4], \"ry\": 1.59}]\n```", - "options": null, - "id": 251 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006784", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.54, \"box_2d\": [542.09, 187.74, 572.59, 219.04], \"box_3d\": [1.68, 1.63, 3.9, -3.14, 2.58, 42.06], \"ry\": 1.46}]\n```", - "options": null, - "id": 252 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006857", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.13, \"box_2d\": [0.0, 197.59, 261.67, 324.11], \"box_3d\": [1.48, 1.61, 3.61, -6.95, 1.9, 10.82], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.58, \"box_2d\": [618.39, 175.36, 641.62, 197.44], \"box_3d\": [1.54, 1.63, 3.67, 1.42, 1.75, 52.85], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 1.88, \"box_2d\": [482.43, 180.31, 526.02, 207.0], \"box_3d\": [1.56, 1.59, 3.69, -6.49, 2.03, 44.42], \"ry\": 1.74}, {\"category\": \"Car\", \"angle\": 1.84, \"box_2d\": [505.26, 172.32, 544.18, 202.15], \"box_3d\": [1.93, 1.59, 4.01, -5.77, 1.93, 49.1], \"ry\": 1.73}, {\"category\": \"Car\", \"angle\": -1.59, \"box_2d\": [658.12, 174.2, 682.29, 196.88], \"box_3d\": [1.58, 1.67, 3.84, 4.35, 1.71, 52.8], \"ry\": -1.51}]\n```", - "options": null, - "id": 253 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006890", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.69, \"box_2d\": [499.64, 183.66, 551.68, 222.36], \"box_3d\": [1.52, 1.73, 4.28, -3.6, 2.01, 31.21], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [549.05, 180.93, 575.09, 200.74], \"box_3d\": [1.44, 1.7, 3.65, -3.65, 2.07, 55.11], \"ry\": 1.58}]\n```", - "options": null, - "id": 254 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "006905", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.64, \"box_2d\": [413.64, 186.81, 509.45, 275.85], \"box_3d\": [1.65, 1.75, 3.39, -3.12, 1.97, 15.51], \"ry\": 1.44}, {\"category\": \"Car\", \"angle\": 1.72, \"box_2d\": [465.77, 187.73, 533.46, 243.68], \"box_3d\": [1.57, 1.54, 4.01, -3.46, 2.07, 22.97], \"ry\": 1.57}]\n```", - "options": null, - "id": 255 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "007024", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.7, \"box_2d\": [461.89, 180.0, 540.68, 237.0], \"box_3d\": [1.42, 1.66, 4.11, -2.98, 1.63, 20.28], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.59, \"box_2d\": [566.59, 158.48, 588.47, 181.05], \"box_3d\": [1.97, 1.87, 4.53, -2.94, 0.74, 65.86], \"ry\": 1.55}]\n```", - "options": null, - "id": 256 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "007057", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.17, \"box_2d\": [217.07, 181.49, 285.97, 211.0], \"box_3d\": [1.42, 1.7, 4.14, -18.49, 1.88, 37.35], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": -1.11, \"box_2d\": [158.6, 178.65, 234.05, 210.36], \"box_3d\": [1.52, 1.71, 4.11, -21.23, 1.84, 37.16], \"ry\": -1.63}, {\"category\": \"Car\", \"angle\": 2.14, \"box_2d\": [35.65, 182.94, 150.33, 228.73], \"box_3d\": [1.42, 1.55, 3.45, -17.54, 1.78, 24.61], \"ry\": 1.53}]\n```", - "options": null, - "id": 257 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "007069", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.66, \"box_2d\": [663.45, 181.42, 711.93, 217.65], \"box_3d\": [1.41, 1.65, 4.28, 3.21, 1.79, 30.78], \"ry\": -1.56}, {\"category\": \"Car\", \"angle\": 0.21, \"box_2d\": [377.44, 177.36, 517.01, 221.73], \"box_3d\": [1.44, 1.8, 4.44, -5.61, 1.62, 25.08], \"ry\": -0.0}, {\"category\": \"Car\", \"angle\": 0.34, \"box_2d\": [303.62, 180.4, 480.53, 236.61], \"box_3d\": [1.5, 1.8, 4.5, -6.19, 1.74, 20.66], \"ry\": 0.05}, {\"category\": \"Car\", \"angle\": 0.39, \"box_2d\": [273.65, 181.58, 472.25, 248.95], \"box_3d\": [1.53, 1.8, 4.22, -5.72, 1.76, 17.58], \"ry\": 0.08}, {\"category\": \"Car\", \"angle\": 0.44, \"box_2d\": [236.71, 182.54, 448.75, 258.48], \"box_3d\": [1.5, 1.46, 3.95, -5.61, 1.73, 15.23], \"ry\": 0.1}, {\"category\": \"Car\", \"angle\": 0.42, \"box_2d\": [151.35, 180.49, 398.09, 275.34], \"box_3d\": [1.53, 1.64, 3.6, -5.85, 1.69, 12.78], \"ry\": 0.0}, {\"category\": \"Car\", \"angle\": -2.44, \"box_2d\": [0.0, 186.56, 191.35, 349.83], \"box_3d\": [1.53, 1.51, 3.38, -6.59, 1.7, 7.6], \"ry\": -3.14}]\n```", - "options": null, - "id": 258 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "007086", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [559.56, 179.69, 582.35, 196.58], \"box_3d\": [1.44, 1.72, 4.14, -3.47, 2.07, 64.52], \"ry\": 1.59}]\n```", - "options": null, - "id": 259 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "007131", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.26, \"box_2d\": [765.46, 179.64, 1037.29, 365.87], \"box_3d\": [1.62, 1.58, 3.89, 2.92, 1.7, 8.25], \"ry\": 1.59}, {\"category\": \"Car\", \"angle\": 1.79, \"box_2d\": [398.72, 186.23, 528.92, 271.26], \"box_3d\": [1.45, 1.67, 4.43, -2.88, 1.76, 14.95], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": -1.79, \"box_2d\": [701.31, 187.08, 819.62, 269.69], \"box_3d\": [1.47, 1.59, 3.62, 2.96, 1.79, 15.08], \"ry\": -1.61}, {\"category\": \"Car\", \"angle\": 1.75, \"box_2d\": [461.07, 180.34, 542.23, 239.56], \"box_3d\": [1.56, 1.68, 3.69, -3.1, 1.79, 21.02], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": 1.45, \"box_2d\": [674.94, 183.2, 754.7, 241.44], \"box_3d\": [1.44, 1.64, 4.5, 2.82, 1.76, 20.55], \"ry\": 1.58}, {\"category\": \"Car\", \"angle\": 1.65, \"box_2d\": [520.57, 180.41, 563.7, 218.25], \"box_3d\": [1.46, 1.51, 3.24, -2.78, 1.78, 29.75], \"ry\": 1.55}, {\"category\": \"Car\", \"angle\": 1.5, \"box_2d\": [664.35, 180.78, 704.78, 211.68], \"box_3d\": [1.38, 1.59, 4.19, 3.49, 1.77, 34.68], \"ry\": 1.6}, {\"category\": \"Car\", \"angle\": -1.61, \"box_2d\": [653.43, 178.07, 676.62, 198.91], \"box_3d\": [1.51, 1.56, 3.81, 4.12, 1.92, 54.65], \"ry\": -1.54}]\n```", - "options": null, - "id": 260 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "007149", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.4, \"box_2d\": [437.29, 189.0, 551.09, 263.75], \"box_3d\": [1.33, 1.71, 4.09, -2.38, 1.72, 15.55], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -1.71, \"box_2d\": [672.18, 171.3, 726.0, 211.18], \"box_3d\": [1.49, 1.56, 4.34, 3.52, 1.46, 29.43], \"ry\": -1.59}, {\"category\": \"Car\", \"angle\": -1.45, \"box_2d\": [513.96, 180.97, 574.52, 227.36], \"box_3d\": [1.42, 1.53, 4.12, -2.19, 1.71, 24.51], \"ry\": -1.54}, {\"category\": \"Car\", \"angle\": -1.73, \"box_2d\": [706.71, 174.34, 803.49, 250.27], \"box_3d\": [1.59, 1.59, 3.89, 3.3, 1.66, 17.24], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": -1.56, \"box_2d\": [656.24, 174.02, 683.1, 196.56], \"box_3d\": [1.38, 1.64, 3.51, 3.82, 1.47, 46.3], \"ry\": -1.48}]\n```", - "options": null, - "id": 261 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "007282", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.45, \"box_2d\": [767.9, 181.48, 975.64, 349.99], \"box_3d\": [1.57, 1.73, 4.15, 2.9, 1.69, 8.98], \"ry\": 1.75}]\n```", - "options": null, - "id": 262 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "007298", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.86, \"box_2d\": [431.98, 183.79, 504.36, 228.35], \"box_3d\": [1.48, 1.61, 3.61, -5.11, 1.89, 26.24], \"ry\": 1.67}, {\"category\": \"Car\", \"angle\": 1.83, \"box_2d\": [569.7, 178.78, 601.21, 199.12], \"box_3d\": [1.56, 1.59, 3.69, -1.94, 2.04, 57.53], \"ry\": 1.8}]\n```", - "options": null, - "id": 263 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "007311", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 2.14, \"box_2d\": [0.0, 156.2, 370.35, 374.0], \"box_3d\": [1.7, 1.63, 4.08, -3.01, 1.67, 4.4], \"ry\": 1.57}, {\"category\": \"Car\", \"angle\": -1.31, \"box_2d\": [342.56, 190.41, 502.1, 291.62], \"box_3d\": [1.4, 1.67, 3.72, -3.05, 1.74, 12.33], \"ry\": -1.55}, {\"category\": \"Car\", \"angle\": 1.46, \"box_2d\": [633.03, 183.76, 683.26, 223.94], \"box_3d\": [1.49, 1.61, 3.74, 1.85, 1.94, 29.17], \"ry\": 1.53}]\n```", - "options": null, - "id": 264 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "007384", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (721.54, 721.54), (cx, cy) = (609.56, 172.85). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": -1.91, \"box_2d\": [861.04, 185.34, 1241.0, 374.0], \"box_3d\": [1.66, 1.56, 3.42, 3.05, 1.77, 5.53], \"ry\": -1.43}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [548.51, 179.62, 608.65, 225.09], \"box_3d\": [1.45, 1.64, 4.21, -1.06, 1.7, 25.33], \"ry\": 1.64}, {\"category\": \"Car\", \"angle\": 1.53, \"box_2d\": [725.1, 178.39, 762.25, 213.82], \"box_3d\": [1.62, 1.58, 3.89, 6.41, 1.9, 35.06], \"ry\": 1.71}, {\"category\": \"Car\", \"angle\": 1.68, \"box_2d\": [609.97, 177.96, 646.71, 204.22], \"box_3d\": [1.45, 1.67, 4.43, 1.11, 1.76, 42.43], \"ry\": 1.71}]\n```", - "options": null, - "id": 265 - }, - { - "dataset": "Mono3DRefer", - "scene_name": "007399", - "question_type": "mobj_3dDetect_json", - "question": "I hope you can understand all the Car objects in this scene to evalue KITTI 3D Object Detection Benchmark. Please ignore the objects that are too small or too far away to be recognized. Additionally, known the camera intrinsic parameters: (fx, fy) = (718.34, 718.34), (cx, cy) = (600.39, 181.51). You'll uderstand the meaning from the phrases I have provided, and answer the 3D attributes of all target objects with JSON format including:\n- category: the category of the object.\n- angle: observation angle of the object on the image plane, ranging from -pi to pi. When the 2D object faces the image center, the angle is 0, and facing left is -pi/2, facing right is pi/2.\n- box_2d: the 2D bounding box of the object in the image coordinate (list of [x1, y1, x2, y2] with unit in pixels).\n- box_3d: the size and location of the 3D object in the real-world coordinate (list of [height, width, length, x, y, z], where x and y are the coordinates of center point of the 3D object's upper plane, and z is the distance from the camera to the object center.\n- ry: Rotation around the Y-axis (up axis) in real-world coordinate, range from -pi to pi. When the head of the object facing away from the camera (-pi/2), ry is 0, and range from 0 to -pi (counter-clockwise). When the head of the object facing towards the camera (pi/2), ry is 0, and range from 0 to pi (counter-clockwise).\n\nNoting the answered JSON format should be like:\n```JSON\n[{\"category\": category, \"angle\": angle, \"box_2d\": [x1, y1, x2, y2], \"box_3d\": [height, width, length, x, y, z], \"ry\": ry}, ...]\n```", - "ground_truth": "```JSON\n[{\"category\": \"Car\", \"angle\": 1.89, \"box_2d\": [340.09, 181.95, 457.46, 245.91], \"box_3d\": [1.42, 1.78, 3.62, -4.99, 1.45, 18.11], \"ry\": 1.62}]\n```", - "options": null, - "id": 266 - } -] \ No newline at end of file