[ { "scene_id": "fix_bboxes_sample_000", "scene_type": "residential_area", "scene_description": "A quiet residential neighborhood with houses, trees, and occasional pedestrians. The scene contains 9 objects: a bench at top-center (bbox: x=0.35, y=0.27, w=0.11, h=0.05); a car at top-right (bbox: x=0.77, y=0.25, w=0.13, h=0.10); a cat at middle-right (bbox: x=0.68, y=0.62, w=0.05, h=0.05); a car at top-right (bbox: x=0.70, y=0.21, w=0.17, h=0.13); a building at bottom-right (bbox: x=0.76, y=0.60, w=0.24, h=0.40); a tree at top-center (bbox: x=0.30, y=0.16, w=0.10, h=0.21); a cat at center (bbox: x=0.49, y=0.38, w=0.05, h=0.06); a person at bottom-left (bbox: x=0.12, y=0.72, w=0.05, h=0.20); a bench at top-center (bbox: x=0.45, y=0.12, w=0.10, h=0.06).", "objects": [ { "id": 0, "class_label": "bench", "position": "top-center", "bbox": [ 0.346, 0.2698, 0.1076, 0.0519 ] }, { "id": 1, "class_label": "car", "position": "top-right", "bbox": [ 0.7692, 0.2477, 0.1333, 0.0968 ] }, { "id": 2, "class_label": "cat", "position": "middle-right", "bbox": [ 0.6836, 0.6154, 0.0491, 0.0537 ] }, { "id": 3, "class_label": "car", "position": "top-right", "bbox": [ 0.6985, 0.2126, 0.1692, 0.13 ] }, { "id": 4, "class_label": "building", "position": "bottom-right", "bbox": [ 0.7563, 0.5971, 0.2437, 0.4029 ] }, { "id": 5, "class_label": "tree", "position": "top-center", "bbox": [ 0.3015, 0.1592, 0.0977, 0.2072 ] }, { "id": 6, "class_label": "cat", "position": "center", "bbox": [ 0.4886, 0.3766, 0.0516, 0.0634 ] }, { "id": 7, "class_label": "person", "position": "bottom-left", "bbox": [ 0.1199, 0.7245, 0.0528, 0.1988 ] }, { "id": 8, "class_label": "bench", "position": "top-center", "bbox": [ 0.4452, 0.1242, 0.1006, 0.0637 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.346, 0.2698, 0.1076, 0.0519 ], "class_label": "bench" }, { "id": 1, "bbox": [ 0.7692, 0.2477, 0.1333, 0.0968 ], "class_label": "car" }, { "id": 2, "bbox": [ 0.6836, 0.6154, 0.0491, 0.0537 ], "class_label": "cat" }, { "id": 3, "bbox": [ 0.6985, 0.2126, 0.1692, 0.13 ], "class_label": "car" }, { "id": 4, "bbox": [ 0.7563, 0.5971, 0.2437, 0.4029 ], "class_label": "building" }, { "id": 5, "bbox": [ 0.3015, 0.1592, 0.0977, 0.2072 ], "class_label": "tree" }, { "id": 6, "bbox": [ 0.4886, 0.3766, 0.0516, 0.0634 ], "class_label": "cat" }, { "id": 7, "bbox": [ 0.1199, 0.7245, 0.0528, 0.1988 ], "class_label": "person" }, { "id": 8, "bbox": [ 0.4452, 0.1242, 0.1006, 0.0637 ], "class_label": "bench" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1000 }, { "scene_id": "fix_bboxes_sample_001", "scene_type": "urban_street", "scene_description": "A busy urban street scene with vehicles, pedestrians, and city infrastructure. The scene contains 6 objects: a truck at bottom-right (bbox: x=0.64, y=0.69, w=0.29, h=0.16); a tree at middle-left (bbox: x=0.19, y=0.48, w=0.12, h=0.16); a bench at middle-left (bbox: x=0.15, y=0.36, w=0.14, h=0.07); a truck at bottom-left (bbox: x=0.15, y=0.73, w=0.25, h=0.11); a truck at bottom-right (bbox: x=0.64, y=0.67, w=0.23, h=0.13); a car at top-right (bbox: x=0.59, y=0.15, w=0.23, h=0.12).", "objects": [ { "id": 0, "class_label": "truck", "position": "bottom-right", "bbox": [ 0.6386, 0.6935, 0.2854, 0.1629 ] }, { "id": 1, "class_label": "tree", "position": "middle-left", "bbox": [ 0.1861, 0.4825, 0.1225, 0.1601 ] }, { "id": 2, "class_label": "bench", "position": "middle-left", "bbox": [ 0.1505, 0.3586, 0.1436, 0.072 ] }, { "id": 3, "class_label": "truck", "position": "bottom-left", "bbox": [ 0.1499, 0.7329, 0.253, 0.108 ] }, { "id": 4, "class_label": "truck", "position": "bottom-right", "bbox": [ 0.6446, 0.6675, 0.2335, 0.1314 ] }, { "id": 5, "class_label": "car", "position": "top-right", "bbox": [ 0.5856, 0.1546, 0.231, 0.1225 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.6386, 0.6935, 0.2854, 0.1629 ], "class_label": "truck" }, { "id": 1, "bbox": [ 0.1861, 0.4825, 0.1225, 0.1601 ], "class_label": "tree" }, { "id": 2, "bbox": [ 0.1505, 0.3586, 0.1436, 0.072 ], "class_label": "bench" }, { "id": 3, "bbox": [ 0.1499, 0.7329, 0.253, 0.108 ], "class_label": "truck" }, { "id": 4, "bbox": [ 0.6446, 0.6675, 0.2335, 0.1314 ], "class_label": "truck" }, { "id": 5, "bbox": [ 0.5856, 0.1546, 0.231, 0.1225 ], "class_label": "car" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1001 }, { "scene_id": "fix_bboxes_sample_002", "scene_type": "intersection", "scene_description": "A road intersection with traffic lights, vehicles, and crossing pedestrians. The scene contains 10 objects: a traffic_light at middle-left (bbox: x=0.11, y=0.49, w=0.02, h=0.07); a car at bottom-left (bbox: x=0.21, y=0.66, w=0.18, h=0.10); a car at bottom-right (bbox: x=0.63, y=0.73, w=0.20, h=0.13); a building at center (bbox: x=0.51, y=0.42, w=0.17, h=0.41); a building at bottom-center (bbox: x=0.23, y=0.60, w=0.27, h=0.38); a car at top-center (bbox: x=0.45, y=0.06, w=0.16, h=0.15); a traffic_light at top-center (bbox: x=0.57, y=0.18, w=0.04, h=0.06); a car at top-center (bbox: x=0.44, y=0.14, w=0.11, h=0.11); a car at top-left (bbox: x=0.06, y=0.15, w=0.13, h=0.13); a car at top-center (bbox: x=0.48, y=0.11, w=0.22, h=0.08).", "objects": [ { "id": 0, "class_label": "traffic_light", "position": "middle-left", "bbox": [ 0.114, 0.4895, 0.0229, 0.0711 ] }, { "id": 1, "class_label": "car", "position": "bottom-left", "bbox": [ 0.211, 0.6596, 0.1765, 0.0969 ] }, { "id": 2, "class_label": "car", "position": "bottom-right", "bbox": [ 0.6318, 0.7268, 0.2015, 0.133 ] }, { "id": 3, "class_label": "building", "position": "center", "bbox": [ 0.5137, 0.4243, 0.166, 0.4143 ] }, { "id": 4, "class_label": "building", "position": "bottom-center", "bbox": [ 0.2343, 0.5981, 0.2745, 0.375 ] }, { "id": 5, "class_label": "car", "position": "top-center", "bbox": [ 0.4534, 0.0581, 0.1587, 0.1458 ] }, { "id": 6, "class_label": "traffic_light", "position": "top-center", "bbox": [ 0.5735, 0.1781, 0.0391, 0.0634 ] }, { "id": 7, "class_label": "car", "position": "top-center", "bbox": [ 0.4385, 0.1436, 0.1126, 0.1099 ] }, { "id": 8, "class_label": "car", "position": "top-left", "bbox": [ 0.0565, 0.151, 0.1311, 0.1298 ] }, { "id": 9, "class_label": "car", "position": "top-center", "bbox": [ 0.4754, 0.1058, 0.2241, 0.0807 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.114, 0.4895, 0.0229, 0.0711 ], "class_label": "traffic_light" }, { "id": 1, "bbox": [ 0.211, 0.6596, 0.1765, 0.0969 ], "class_label": "car" }, { "id": 2, "bbox": [ 0.6318, 0.7268, 0.2015, 0.133 ], "class_label": "car" }, { "id": 3, "bbox": [ 0.5137, 0.4243, 0.166, 0.4143 ], "class_label": "building" }, { "id": 4, "bbox": [ 0.2343, 0.5981, 0.2745, 0.375 ], "class_label": "building" }, { "id": 5, "bbox": [ 0.4534, 0.0581, 0.1587, 0.1458 ], "class_label": "car" }, { "id": 6, "bbox": [ 0.5735, 0.1781, 0.0391, 0.0634 ], "class_label": "traffic_light" }, { "id": 7, "bbox": [ 0.4385, 0.1436, 0.1126, 0.1099 ], "class_label": "car" }, { "id": 8, "bbox": [ 0.0565, 0.151, 0.1311, 0.1298 ], "class_label": "car" }, { "id": 9, "bbox": [ 0.4754, 0.1058, 0.2241, 0.0807 ], "class_label": "car" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1002 }, { "scene_id": "fix_bboxes_sample_003", "scene_type": "residential_area", "scene_description": "A quiet residential neighborhood with houses, trees, and occasional pedestrians. The scene contains 8 objects: a person at middle-left (bbox: x=0.16, y=0.27, w=0.07, h=0.17); a person at top-center (bbox: x=0.35, y=0.08, w=0.06, h=0.17); a person at bottom-right (bbox: x=0.67, y=0.63, w=0.07, h=0.23); a dog at bottom-right (bbox: x=0.79, y=0.68, w=0.05, h=0.05); a car at middle-right (bbox: x=0.83, y=0.42, w=0.10, h=0.10); a person at bottom-left (bbox: x=0.11, y=0.67, w=0.08, h=0.23); a tree at middle-left (bbox: x=0.19, y=0.46, w=0.13, h=0.24); a tree at top-right (bbox: x=0.69, y=0.00, w=0.12, h=0.27).", "objects": [ { "id": 0, "class_label": "person", "position": "middle-left", "bbox": [ 0.1562, 0.2743, 0.0689, 0.167 ] }, { "id": 1, "class_label": "person", "position": "top-center", "bbox": [ 0.3469, 0.0752, 0.059, 0.1677 ] }, { "id": 2, "class_label": "person", "position": "bottom-right", "bbox": [ 0.6723, 0.6283, 0.0687, 0.2348 ] }, { "id": 3, "class_label": "dog", "position": "bottom-right", "bbox": [ 0.7854, 0.6756, 0.0531, 0.0492 ] }, { "id": 4, "class_label": "car", "position": "middle-right", "bbox": [ 0.8267, 0.4176, 0.1047, 0.0968 ] }, { "id": 5, "class_label": "person", "position": "bottom-left", "bbox": [ 0.1127, 0.6712, 0.0765, 0.2284 ] }, { "id": 6, "class_label": "tree", "position": "middle-left", "bbox": [ 0.1873, 0.4559, 0.1335, 0.2408 ] }, { "id": 7, "class_label": "tree", "position": "top-right", "bbox": [ 0.6898, 0.0, 0.1208, 0.2742 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.1562, 0.2743, 0.0689, 0.167 ], "class_label": "person" }, { "id": 1, "bbox": [ 0.3469, 0.0752, 0.059, 0.1677 ], "class_label": "person" }, { "id": 2, "bbox": [ 0.6723, 0.6283, 0.0687, 0.2348 ], "class_label": "person" }, { "id": 3, "bbox": [ 0.7854, 0.6756, 0.0531, 0.0492 ], "class_label": "dog" }, { "id": 4, "bbox": [ 0.8267, 0.4176, 0.1047, 0.0968 ], "class_label": "car" }, { "id": 5, "bbox": [ 0.1127, 0.6712, 0.0765, 0.2284 ], "class_label": "person" }, { "id": 6, "bbox": [ 0.1873, 0.4559, 0.1335, 0.2408 ], "class_label": "tree" }, { "id": 7, "bbox": [ 0.6898, 0.0, 0.1208, 0.2742 ], "class_label": "tree" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1003 }, { "scene_id": "fix_bboxes_sample_004", "scene_type": "residential_area", "scene_description": "A quiet residential neighborhood with houses, trees, and occasional pedestrians. The scene contains 4 objects: a cat at bottom-right (bbox: x=0.70, y=0.84, w=0.07, h=0.06); a bench at middle-left (bbox: x=0.20, y=0.58, w=0.10, h=0.05); a tree at middle-right (bbox: x=0.66, y=0.47, w=0.08, h=0.20); a tree at top-right (bbox: x=0.74, y=0.00, w=0.09, h=0.25).", "objects": [ { "id": 0, "class_label": "cat", "position": "bottom-right", "bbox": [ 0.7037, 0.8384, 0.0677, 0.0553 ] }, { "id": 1, "class_label": "bench", "position": "middle-left", "bbox": [ 0.2018, 0.5795, 0.1039, 0.0518 ] }, { "id": 2, "class_label": "tree", "position": "middle-right", "bbox": [ 0.6588, 0.4658, 0.0848, 0.1972 ] }, { "id": 3, "class_label": "tree", "position": "top-right", "bbox": [ 0.743, 0.0, 0.0865, 0.2525 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.7037, 0.8384, 0.0677, 0.0553 ], "class_label": "cat" }, { "id": 1, "bbox": [ 0.2018, 0.5795, 0.1039, 0.0518 ], "class_label": "bench" }, { "id": 2, "bbox": [ 0.6588, 0.4658, 0.0848, 0.1972 ], "class_label": "tree" }, { "id": 3, "bbox": [ 0.743, 0.0, 0.0865, 0.2525 ], "class_label": "tree" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1004 }, { "scene_id": "fix_bboxes_sample_005", "scene_type": "residential_area", "scene_description": "A quiet residential neighborhood with houses, trees, and occasional pedestrians. The scene contains 7 objects: a dog at bottom-left (bbox: x=0.25, y=0.74, w=0.09, h=0.06); a cat at top-right (bbox: x=0.73, y=0.23, w=0.06, h=0.06); a bench at top-left (bbox: x=0.05, y=0.22, w=0.13, h=0.07); a person at bottom-center (bbox: x=0.42, y=0.66, w=0.04, h=0.13); a car at middle-left (bbox: x=0.06, y=0.40, w=0.23, h=0.12); a bench at bottom-right (bbox: x=0.79, y=0.72, w=0.08, h=0.07); a person at bottom-left (bbox: x=0.13, y=0.78, w=0.06, h=0.17).", "objects": [ { "id": 0, "class_label": "dog", "position": "bottom-left", "bbox": [ 0.252, 0.74, 0.0938, 0.0629 ] }, { "id": 1, "class_label": "cat", "position": "top-right", "bbox": [ 0.735, 0.2329, 0.0595, 0.0572 ] }, { "id": 2, "class_label": "bench", "position": "top-left", "bbox": [ 0.0457, 0.2208, 0.1263, 0.0674 ] }, { "id": 3, "class_label": "person", "position": "bottom-center", "bbox": [ 0.4166, 0.6571, 0.0438, 0.1348 ] }, { "id": 4, "class_label": "car", "position": "middle-left", "bbox": [ 0.0631, 0.3961, 0.2311, 0.1216 ] }, { "id": 5, "class_label": "bench", "position": "bottom-right", "bbox": [ 0.7936, 0.7242, 0.0833, 0.0738 ] }, { "id": 6, "class_label": "person", "position": "bottom-left", "bbox": [ 0.1345, 0.7783, 0.064, 0.1685 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.252, 0.74, 0.0938, 0.0629 ], "class_label": "dog" }, { "id": 1, "bbox": [ 0.735, 0.2329, 0.0595, 0.0572 ], "class_label": "cat" }, { "id": 2, "bbox": [ 0.0457, 0.2208, 0.1263, 0.0674 ], "class_label": "bench" }, { "id": 3, "bbox": [ 0.4166, 0.6571, 0.0438, 0.1348 ], "class_label": "person" }, { "id": 4, "bbox": [ 0.0631, 0.3961, 0.2311, 0.1216 ], "class_label": "car" }, { "id": 5, "bbox": [ 0.7936, 0.7242, 0.0833, 0.0738 ], "class_label": "bench" }, { "id": 6, "bbox": [ 0.1345, 0.7783, 0.064, 0.1685 ], "class_label": "person" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1005 }, { "scene_id": "fix_bboxes_sample_006", "scene_type": "parking_lot", "scene_description": "A parking lot with various vehicles and some pedestrians. The scene contains 5 objects: a truck at bottom-center (bbox: x=0.37, y=0.61, w=0.15, h=0.17); a truck at bottom-right (bbox: x=0.68, y=0.73, w=0.28, h=0.18); a bicycle at top-left (bbox: x=0.26, y=0.21, w=0.08, h=0.08); a car at top-left (bbox: x=0.22, y=0.22, w=0.14, h=0.10); a car at bottom-left (bbox: x=0.06, y=0.70, w=0.24, h=0.14).", "objects": [ { "id": 0, "class_label": "truck", "position": "bottom-center", "bbox": [ 0.3708, 0.6129, 0.1519, 0.1745 ] }, { "id": 1, "class_label": "truck", "position": "bottom-right", "bbox": [ 0.6849, 0.7281, 0.2841, 0.1782 ] }, { "id": 2, "class_label": "bicycle", "position": "top-left", "bbox": [ 0.2575, 0.2095, 0.0784, 0.0764 ] }, { "id": 3, "class_label": "car", "position": "top-left", "bbox": [ 0.2176, 0.2228, 0.1393, 0.1002 ] }, { "id": 4, "class_label": "car", "position": "bottom-left", "bbox": [ 0.0578, 0.703, 0.2354, 0.1359 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.3708, 0.6129, 0.1519, 0.1745 ], "class_label": "truck" }, { "id": 1, "bbox": [ 0.6849, 0.7281, 0.2841, 0.1782 ], "class_label": "truck" }, { "id": 2, "bbox": [ 0.2575, 0.2095, 0.0784, 0.0764 ], "class_label": "bicycle" }, { "id": 3, "bbox": [ 0.2176, 0.2228, 0.1393, 0.1002 ], "class_label": "car" }, { "id": 4, "bbox": [ 0.0578, 0.703, 0.2354, 0.1359 ], "class_label": "car" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1006 }, { "scene_id": "fix_bboxes_sample_007", "scene_type": "urban_street", "scene_description": "A busy urban street scene with vehicles, pedestrians, and city infrastructure. The scene contains 9 objects: a person at center (bbox: x=0.38, y=0.53, w=0.07, h=0.22); a bicycle at bottom-right (bbox: x=0.81, y=0.78, w=0.06, h=0.09); a car at bottom-left (bbox: x=0.13, y=0.67, w=0.22, h=0.10); a car at center (bbox: x=0.32, y=0.36, w=0.24, h=0.13); a car at middle-right (bbox: x=0.80, y=0.32, w=0.19, h=0.09); a bench at top-center (bbox: x=0.42, y=0.24, w=0.11, h=0.06); a tree at center (bbox: x=0.40, y=0.47, w=0.15, h=0.33); a car at bottom-right (bbox: x=0.74, y=0.68, w=0.24, h=0.11); a bench at middle-left (bbox: x=0.12, y=0.33, w=0.09, h=0.05).", "objects": [ { "id": 0, "class_label": "person", "position": "center", "bbox": [ 0.3826, 0.5321, 0.0697, 0.2215 ] }, { "id": 1, "class_label": "bicycle", "position": "bottom-right", "bbox": [ 0.8148, 0.777, 0.0632, 0.0947 ] }, { "id": 2, "class_label": "car", "position": "bottom-left", "bbox": [ 0.1291, 0.6657, 0.2247, 0.0969 ] }, { "id": 3, "class_label": "car", "position": "center", "bbox": [ 0.3176, 0.3557, 0.2414, 0.1253 ] }, { "id": 4, "class_label": "car", "position": "middle-right", "bbox": [ 0.8018, 0.3215, 0.1856, 0.0864 ] }, { "id": 5, "class_label": "bench", "position": "top-center", "bbox": [ 0.4209, 0.2413, 0.1119, 0.0585 ] }, { "id": 6, "class_label": "tree", "position": "center", "bbox": [ 0.4, 0.4714, 0.1475, 0.3329 ] }, { "id": 7, "class_label": "car", "position": "bottom-right", "bbox": [ 0.7388, 0.6811, 0.2362, 0.1119 ] }, { "id": 8, "class_label": "bench", "position": "middle-left", "bbox": [ 0.1197, 0.3261, 0.088, 0.0521 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.3826, 0.5321, 0.0697, 0.2215 ], "class_label": "person" }, { "id": 1, "bbox": [ 0.8148, 0.777, 0.0632, 0.0947 ], "class_label": "bicycle" }, { "id": 2, "bbox": [ 0.1291, 0.6657, 0.2247, 0.0969 ], "class_label": "car" }, { "id": 3, "bbox": [ 0.3176, 0.3557, 0.2414, 0.1253 ], "class_label": "car" }, { "id": 4, "bbox": [ 0.8018, 0.3215, 0.1856, 0.0864 ], "class_label": "car" }, { "id": 5, "bbox": [ 0.4209, 0.2413, 0.1119, 0.0585 ], "class_label": "bench" }, { "id": 6, "bbox": [ 0.4, 0.4714, 0.1475, 0.3329 ], "class_label": "tree" }, { "id": 7, "bbox": [ 0.7388, 0.6811, 0.2362, 0.1119 ], "class_label": "car" }, { "id": 8, "bbox": [ 0.1197, 0.3261, 0.088, 0.0521 ], "class_label": "bench" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1007 }, { "scene_id": "fix_bboxes_sample_008", "scene_type": "intersection", "scene_description": "A road intersection with traffic lights, vehicles, and crossing pedestrians. The scene contains 9 objects: a person at top-center (bbox: x=0.33, y=0.11, w=0.07, h=0.24); a truck at top-center (bbox: x=0.39, y=0.20, w=0.24, h=0.15); a person at middle-left (bbox: x=0.11, y=0.48, w=0.06, h=0.18); a truck at bottom-left (bbox: x=0.03, y=0.66, w=0.20, h=0.17); a truck at middle-left (bbox: x=0.10, y=0.39, w=0.29, h=0.11); a truck at bottom-center (bbox: x=0.34, y=0.70, w=0.20, h=0.16); a person at middle-right (bbox: x=0.78, y=0.35, w=0.06, h=0.15); a truck at center (bbox: x=0.37, y=0.51, w=0.16, h=0.14); a traffic_light at bottom-center (bbox: x=0.61, y=0.84, w=0.03, h=0.08).", "objects": [ { "id": 0, "class_label": "person", "position": "top-center", "bbox": [ 0.3309, 0.1129, 0.0683, 0.2388 ] }, { "id": 1, "class_label": "truck", "position": "top-center", "bbox": [ 0.3924, 0.1969, 0.2388, 0.1535 ] }, { "id": 2, "class_label": "person", "position": "middle-left", "bbox": [ 0.1139, 0.4801, 0.0594, 0.1805 ] }, { "id": 3, "class_label": "truck", "position": "bottom-left", "bbox": [ 0.0283, 0.6601, 0.2025, 0.1693 ] }, { "id": 4, "class_label": "truck", "position": "middle-left", "bbox": [ 0.0997, 0.3858, 0.2894, 0.1083 ] }, { "id": 5, "class_label": "truck", "position": "bottom-center", "bbox": [ 0.3449, 0.7014, 0.2041, 0.1616 ] }, { "id": 6, "class_label": "person", "position": "middle-right", "bbox": [ 0.78, 0.3456, 0.0566, 0.1536 ] }, { "id": 7, "class_label": "truck", "position": "center", "bbox": [ 0.3661, 0.511, 0.1589, 0.14 ] }, { "id": 8, "class_label": "traffic_light", "position": "bottom-center", "bbox": [ 0.6079, 0.8371, 0.0293, 0.0815 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.3309, 0.1129, 0.0683, 0.2388 ], "class_label": "person" }, { "id": 1, "bbox": [ 0.3924, 0.1969, 0.2388, 0.1535 ], "class_label": "truck" }, { "id": 2, "bbox": [ 0.1139, 0.4801, 0.0594, 0.1805 ], "class_label": "person" }, { "id": 3, "bbox": [ 0.0283, 0.6601, 0.2025, 0.1693 ], "class_label": "truck" }, { "id": 4, "bbox": [ 0.0997, 0.3858, 0.2894, 0.1083 ], "class_label": "truck" }, { "id": 5, "bbox": [ 0.3449, 0.7014, 0.2041, 0.1616 ], "class_label": "truck" }, { "id": 6, "bbox": [ 0.78, 0.3456, 0.0566, 0.1536 ], "class_label": "person" }, { "id": 7, "bbox": [ 0.3661, 0.511, 0.1589, 0.14 ], "class_label": "truck" }, { "id": 8, "bbox": [ 0.6079, 0.8371, 0.0293, 0.0815 ], "class_label": "traffic_light" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1008 }, { "scene_id": "fix_bboxes_sample_009", "scene_type": "park", "scene_description": "A peaceful park setting with trees, benches, and people walking their pets. The scene contains 4 objects: a tree at top-center (bbox: x=0.54, y=0.01, w=0.13, h=0.25); a dog at middle-left (bbox: x=0.11, y=0.35, w=0.06, h=0.04); a tree at top-right (bbox: x=0.71, y=0.02, w=0.12, h=0.17); a dog at middle-right (bbox: x=0.79, y=0.46, w=0.10, h=0.05).", "objects": [ { "id": 0, "class_label": "tree", "position": "top-center", "bbox": [ 0.5431, 0.0106, 0.1317, 0.2457 ] }, { "id": 1, "class_label": "dog", "position": "middle-left", "bbox": [ 0.1074, 0.3528, 0.0637, 0.0413 ] }, { "id": 2, "class_label": "tree", "position": "top-right", "bbox": [ 0.7097, 0.0182, 0.1157, 0.1695 ] }, { "id": 3, "class_label": "dog", "position": "middle-right", "bbox": [ 0.789, 0.4644, 0.0985, 0.0455 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.5431, 0.0106, 0.1317, 0.2457 ], "class_label": "tree" }, { "id": 1, "bbox": [ 0.1074, 0.3528, 0.0637, 0.0413 ], "class_label": "dog" }, { "id": 2, "bbox": [ 0.7097, 0.0182, 0.1157, 0.1695 ], "class_label": "tree" }, { "id": 3, "bbox": [ 0.789, 0.4644, 0.0985, 0.0455 ], "class_label": "dog" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1009 }, { "scene_id": "fix_bboxes_sample_010", "scene_type": "intersection", "scene_description": "A road intersection with traffic lights, vehicles, and crossing pedestrians. The scene contains 7 objects: a bicycle at top-center (bbox: x=0.35, y=0.23, w=0.08, h=0.08); a truck at top-center (bbox: x=0.36, y=0.23, w=0.28, h=0.14); a bicycle at top-right (bbox: x=0.80, y=0.16, w=0.08, h=0.10); a car at bottom-right (bbox: x=0.76, y=0.83, w=0.18, h=0.13); a truck at bottom-right (bbox: x=0.70, y=0.81, w=0.30, h=0.11); a building at top-left (bbox: x=0.15, y=0.00, w=0.19, h=0.31); a building at top-left (bbox: x=0.17, y=0.11, w=0.16, h=0.21).", "objects": [ { "id": 0, "class_label": "bicycle", "position": "top-center", "bbox": [ 0.3504, 0.2325, 0.0846, 0.0847 ] }, { "id": 1, "class_label": "truck", "position": "top-center", "bbox": [ 0.3641, 0.2289, 0.2751, 0.1408 ] }, { "id": 2, "class_label": "bicycle", "position": "top-right", "bbox": [ 0.8038, 0.1557, 0.0835, 0.0961 ] }, { "id": 3, "class_label": "car", "position": "bottom-right", "bbox": [ 0.7604, 0.8326, 0.1778, 0.1321 ] }, { "id": 4, "class_label": "truck", "position": "bottom-right", "bbox": [ 0.7007, 0.81, 0.2993, 0.1138 ] }, { "id": 5, "class_label": "building", "position": "top-left", "bbox": [ 0.146, 0.0, 0.1882, 0.3083 ] }, { "id": 6, "class_label": "building", "position": "top-left", "bbox": [ 0.1709, 0.1099, 0.1626, 0.2128 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.3504, 0.2325, 0.0846, 0.0847 ], "class_label": "bicycle" }, { "id": 1, "bbox": [ 0.3641, 0.2289, 0.2751, 0.1408 ], "class_label": "truck" }, { "id": 2, "bbox": [ 0.8038, 0.1557, 0.0835, 0.0961 ], "class_label": "bicycle" }, { "id": 3, "bbox": [ 0.7604, 0.8326, 0.1778, 0.1321 ], "class_label": "car" }, { "id": 4, "bbox": [ 0.7007, 0.81, 0.2993, 0.1138 ], "class_label": "truck" }, { "id": 5, "bbox": [ 0.146, 0.0, 0.1882, 0.3083 ], "class_label": "building" }, { "id": 6, "bbox": [ 0.1709, 0.1099, 0.1626, 0.2128 ], "class_label": "building" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1010 }, { "scene_id": "fix_bboxes_sample_011", "scene_type": "parking_lot", "scene_description": "A parking lot with various vehicles and some pedestrians. The scene contains 9 objects: a car at bottom-left (bbox: x=0.19, y=0.73, w=0.13, h=0.11); a building at center (bbox: x=0.35, y=0.43, w=0.19, h=0.32); a bicycle at top-center (bbox: x=0.56, y=0.14, w=0.08, h=0.06); a car at bottom-center (bbox: x=0.40, y=0.77, w=0.23, h=0.14); a building at bottom-right (bbox: x=0.63, y=0.62, w=0.18, h=0.38); a truck at center (bbox: x=0.25, y=0.47, w=0.28, h=0.18); a car at top-right (bbox: x=0.66, y=0.15, w=0.16, h=0.10); a bicycle at bottom-left (bbox: x=0.10, y=0.69, w=0.06, h=0.07); a truck at top-left (bbox: x=0.06, y=0.07, w=0.18, h=0.14).", "objects": [ { "id": 0, "class_label": "car", "position": "bottom-left", "bbox": [ 0.1935, 0.7331, 0.1266, 0.1109 ] }, { "id": 1, "class_label": "building", "position": "center", "bbox": [ 0.3461, 0.4307, 0.1857, 0.3163 ] }, { "id": 2, "class_label": "bicycle", "position": "top-center", "bbox": [ 0.5562, 0.1376, 0.081, 0.0648 ] }, { "id": 3, "class_label": "car", "position": "bottom-center", "bbox": [ 0.3976, 0.7678, 0.2259, 0.138 ] }, { "id": 4, "class_label": "building", "position": "bottom-right", "bbox": [ 0.6278, 0.6209, 0.1831, 0.3791 ] }, { "id": 5, "class_label": "truck", "position": "center", "bbox": [ 0.2492, 0.475, 0.2801, 0.177 ] }, { "id": 6, "class_label": "car", "position": "top-right", "bbox": [ 0.6566, 0.1536, 0.1616, 0.0978 ] }, { "id": 7, "class_label": "bicycle", "position": "bottom-left", "bbox": [ 0.095, 0.694, 0.0633, 0.0699 ] }, { "id": 8, "class_label": "truck", "position": "top-left", "bbox": [ 0.0625, 0.0667, 0.1784, 0.1424 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.1935, 0.7331, 0.1266, 0.1109 ], "class_label": "car" }, { "id": 1, "bbox": [ 0.3461, 0.4307, 0.1857, 0.3163 ], "class_label": "building" }, { "id": 2, "bbox": [ 0.5562, 0.1376, 0.081, 0.0648 ], "class_label": "bicycle" }, { "id": 3, "bbox": [ 0.3976, 0.7678, 0.2259, 0.138 ], "class_label": "car" }, { "id": 4, "bbox": [ 0.6278, 0.6209, 0.1831, 0.3791 ], "class_label": "building" }, { "id": 5, "bbox": [ 0.2492, 0.475, 0.2801, 0.177 ], "class_label": "truck" }, { "id": 6, "bbox": [ 0.6566, 0.1536, 0.1616, 0.0978 ], "class_label": "car" }, { "id": 7, "bbox": [ 0.095, 0.694, 0.0633, 0.0699 ], "class_label": "bicycle" }, { "id": 8, "bbox": [ 0.0625, 0.0667, 0.1784, 0.1424 ], "class_label": "truck" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1011 }, { "scene_id": "fix_bboxes_sample_012", "scene_type": "parking_lot", "scene_description": "A parking lot with various vehicles and some pedestrians. The scene contains 7 objects: a truck at bottom-left (bbox: x=0.06, y=0.82, w=0.18, h=0.10); a building at middle-left (bbox: x=0.00, y=0.33, w=0.28, h=0.31); a car at bottom-center (bbox: x=0.37, y=0.80, w=0.19, h=0.10); a truck at middle-left (bbox: x=0.09, y=0.30, w=0.24, h=0.17); a building at center (bbox: x=0.55, y=0.41, w=0.20, h=0.38); a person at top-center (bbox: x=0.47, y=0.11, w=0.06, h=0.13); a truck at middle-left (bbox: x=0.11, y=0.50, w=0.29, h=0.17).", "objects": [ { "id": 0, "class_label": "truck", "position": "bottom-left", "bbox": [ 0.0607, 0.8189, 0.1812, 0.1024 ] }, { "id": 1, "class_label": "building", "position": "middle-left", "bbox": [ 0.0, 0.331, 0.279, 0.3101 ] }, { "id": 2, "class_label": "car", "position": "bottom-center", "bbox": [ 0.3712, 0.7963, 0.1916, 0.0957 ] }, { "id": 3, "class_label": "truck", "position": "middle-left", "bbox": [ 0.0911, 0.3025, 0.242, 0.1707 ] }, { "id": 4, "class_label": "building", "position": "center", "bbox": [ 0.5518, 0.4124, 0.1951, 0.3822 ] }, { "id": 5, "class_label": "person", "position": "top-center", "bbox": [ 0.4708, 0.1062, 0.0604, 0.1338 ] }, { "id": 6, "class_label": "truck", "position": "middle-left", "bbox": [ 0.1082, 0.4996, 0.2897, 0.172 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.0607, 0.8189, 0.1812, 0.1024 ], "class_label": "truck" }, { "id": 1, "bbox": [ 0.0, 0.331, 0.279, 0.3101 ], "class_label": "building" }, { "id": 2, "bbox": [ 0.3712, 0.7963, 0.1916, 0.0957 ], "class_label": "car" }, { "id": 3, "bbox": [ 0.0911, 0.3025, 0.242, 0.1707 ], "class_label": "truck" }, { "id": 4, "bbox": [ 0.5518, 0.4124, 0.1951, 0.3822 ], "class_label": "building" }, { "id": 5, "bbox": [ 0.4708, 0.1062, 0.0604, 0.1338 ], "class_label": "person" }, { "id": 6, "bbox": [ 0.1082, 0.4996, 0.2897, 0.172 ], "class_label": "truck" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1012 }, { "scene_id": "fix_bboxes_sample_013", "scene_type": "residential_area", "scene_description": "A quiet residential neighborhood with houses, trees, and occasional pedestrians. The scene contains 6 objects: a cat at top-right (bbox: x=0.84, y=0.09, w=0.06, h=0.04); a tree at center (bbox: x=0.49, y=0.44, w=0.08, h=0.22); a building at middle-left (bbox: x=0.14, y=0.28, w=0.20, h=0.38); a person at middle-right (bbox: x=0.82, y=0.29, w=0.07, h=0.20); a tree at bottom-left (bbox: x=0.13, y=0.71, w=0.09, h=0.17); a tree at top-center (bbox: x=0.37, y=0.02, w=0.14, h=0.28).", "objects": [ { "id": 0, "class_label": "cat", "position": "top-right", "bbox": [ 0.8417, 0.0907, 0.0627, 0.0408 ] }, { "id": 1, "class_label": "tree", "position": "center", "bbox": [ 0.4918, 0.4365, 0.0801, 0.2161 ] }, { "id": 2, "class_label": "building", "position": "middle-left", "bbox": [ 0.1356, 0.275, 0.2006, 0.3846 ] }, { "id": 3, "class_label": "person", "position": "middle-right", "bbox": [ 0.8238, 0.2859, 0.0651, 0.2015 ] }, { "id": 4, "class_label": "tree", "position": "bottom-left", "bbox": [ 0.1319, 0.713, 0.0873, 0.174 ] }, { "id": 5, "class_label": "tree", "position": "top-center", "bbox": [ 0.3683, 0.0189, 0.1402, 0.2775 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.8417, 0.0907, 0.0627, 0.0408 ], "class_label": "cat" }, { "id": 1, "bbox": [ 0.4918, 0.4365, 0.0801, 0.2161 ], "class_label": "tree" }, { "id": 2, "bbox": [ 0.1356, 0.275, 0.2006, 0.3846 ], "class_label": "building" }, { "id": 3, "bbox": [ 0.8238, 0.2859, 0.0651, 0.2015 ], "class_label": "person" }, { "id": 4, "bbox": [ 0.1319, 0.713, 0.0873, 0.174 ], "class_label": "tree" }, { "id": 5, "bbox": [ 0.3683, 0.0189, 0.1402, 0.2775 ], "class_label": "tree" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1013 }, { "scene_id": "fix_bboxes_sample_014", "scene_type": "intersection", "scene_description": "A road intersection with traffic lights, vehicles, and crossing pedestrians. The scene contains 6 objects: a truck at center (bbox: x=0.56, y=0.47, w=0.16, h=0.13); a traffic_light at bottom-right (bbox: x=0.77, y=0.75, w=0.03, h=0.09); a car at middle-left (bbox: x=0.04, y=0.55, w=0.21, h=0.13); a person at bottom-right (bbox: x=0.72, y=0.78, w=0.05, h=0.22); a car at center (bbox: x=0.42, y=0.32, w=0.17, h=0.11); a person at bottom-right (bbox: x=0.78, y=0.73, w=0.07, h=0.13).", "objects": [ { "id": 0, "class_label": "truck", "position": "center", "bbox": [ 0.5644, 0.4694, 0.1552, 0.1344 ] }, { "id": 1, "class_label": "traffic_light", "position": "bottom-right", "bbox": [ 0.7685, 0.7477, 0.0298, 0.0924 ] }, { "id": 2, "class_label": "car", "position": "middle-left", "bbox": [ 0.0432, 0.5505, 0.2128, 0.1307 ] }, { "id": 3, "class_label": "person", "position": "bottom-right", "bbox": [ 0.7225, 0.7759, 0.0532, 0.2165 ] }, { "id": 4, "class_label": "car", "position": "center", "bbox": [ 0.4181, 0.3191, 0.1693, 0.1142 ] }, { "id": 5, "class_label": "person", "position": "bottom-right", "bbox": [ 0.7797, 0.7283, 0.0747, 0.1264 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.5644, 0.4694, 0.1552, 0.1344 ], "class_label": "truck" }, { "id": 1, "bbox": [ 0.7685, 0.7477, 0.0298, 0.0924 ], "class_label": "traffic_light" }, { "id": 2, "bbox": [ 0.0432, 0.5505, 0.2128, 0.1307 ], "class_label": "car" }, { "id": 3, "bbox": [ 0.7225, 0.7759, 0.0532, 0.2165 ], "class_label": "person" }, { "id": 4, "bbox": [ 0.4181, 0.3191, 0.1693, 0.1142 ], "class_label": "car" }, { "id": 5, "bbox": [ 0.7797, 0.7283, 0.0747, 0.1264 ], "class_label": "person" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1014 }, { "scene_id": "fix_bboxes_sample_015", "scene_type": "urban_street", "scene_description": "A busy urban street scene with vehicles, pedestrians, and city infrastructure. The scene contains 7 objects: a bench at bottom-right (bbox: x=0.72, y=0.75, w=0.13, h=0.07); a traffic_light at bottom-center (bbox: x=0.34, y=0.84, w=0.03, h=0.08); a truck at bottom-left (bbox: x=0.06, y=0.74, w=0.27, h=0.17); a bicycle at middle-left (bbox: x=0.24, y=0.43, w=0.08, h=0.07); a truck at top-center (bbox: x=0.39, y=0.15, w=0.19, h=0.12); a traffic_light at middle-right (bbox: x=0.71, y=0.31, w=0.03, h=0.08); a car at bottom-left (bbox: x=0.06, y=0.66, w=0.16, h=0.09).", "objects": [ { "id": 0, "class_label": "bench", "position": "bottom-right", "bbox": [ 0.7151, 0.7509, 0.1303, 0.0728 ] }, { "id": 1, "class_label": "traffic_light", "position": "bottom-center", "bbox": [ 0.3355, 0.8389, 0.0347, 0.0849 ] }, { "id": 2, "class_label": "truck", "position": "bottom-left", "bbox": [ 0.0647, 0.7403, 0.2707, 0.1707 ] }, { "id": 3, "class_label": "bicycle", "position": "middle-left", "bbox": [ 0.2406, 0.4268, 0.0782, 0.0738 ] }, { "id": 4, "class_label": "truck", "position": "top-center", "bbox": [ 0.3888, 0.1494, 0.1883, 0.1225 ] }, { "id": 5, "class_label": "traffic_light", "position": "middle-right", "bbox": [ 0.7103, 0.3119, 0.0314, 0.0766 ] }, { "id": 6, "class_label": "car", "position": "bottom-left", "bbox": [ 0.0646, 0.6637, 0.1603, 0.0856 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.7151, 0.7509, 0.1303, 0.0728 ], "class_label": "bench" }, { "id": 1, "bbox": [ 0.3355, 0.8389, 0.0347, 0.0849 ], "class_label": "traffic_light" }, { "id": 2, "bbox": [ 0.0647, 0.7403, 0.2707, 0.1707 ], "class_label": "truck" }, { "id": 3, "bbox": [ 0.2406, 0.4268, 0.0782, 0.0738 ], "class_label": "bicycle" }, { "id": 4, "bbox": [ 0.3888, 0.1494, 0.1883, 0.1225 ], "class_label": "truck" }, { "id": 5, "bbox": [ 0.7103, 0.3119, 0.0314, 0.0766 ], "class_label": "traffic_light" }, { "id": 6, "bbox": [ 0.0646, 0.6637, 0.1603, 0.0856 ], "class_label": "car" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1015 }, { "scene_id": "fix_bboxes_sample_016", "scene_type": "parking_lot", "scene_description": "A parking lot with various vehicles and some pedestrians. The scene contains 11 objects: a building at middle-right (bbox: x=0.68, y=0.25, w=0.30, h=0.24); a bicycle at top-left (bbox: x=0.15, y=0.14, w=0.09, h=0.11); a building at top-right (bbox: x=0.68, y=0.04, w=0.27, h=0.25); a person at middle-right (bbox: x=0.72, y=0.46, w=0.08, h=0.15); a person at top-left (bbox: x=0.11, y=0.15, w=0.07, h=0.13); a person at middle-right (bbox: x=0.73, y=0.53, w=0.07, h=0.12); a bicycle at top-center (bbox: x=0.55, y=0.17, w=0.11, h=0.09); a truck at middle-left (bbox: x=0.18, y=0.38, w=0.20, h=0.12); a bicycle at middle-right (bbox: x=0.69, y=0.36, w=0.09, h=0.09); a truck at bottom-left (bbox: x=0.10, y=0.81, w=0.15, h=0.17); a bicycle at top-left (bbox: x=0.23, y=0.24, w=0.07, h=0.09).", "objects": [ { "id": 0, "class_label": "building", "position": "middle-right", "bbox": [ 0.6835, 0.2478, 0.2968, 0.2427 ] }, { "id": 1, "class_label": "bicycle", "position": "top-left", "bbox": [ 0.1514, 0.1444, 0.0881, 0.1145 ] }, { "id": 2, "class_label": "building", "position": "top-right", "bbox": [ 0.6847, 0.0425, 0.2652, 0.2451 ] }, { "id": 3, "class_label": "person", "position": "middle-right", "bbox": [ 0.7209, 0.4596, 0.0752, 0.1498 ] }, { "id": 4, "class_label": "person", "position": "top-left", "bbox": [ 0.1139, 0.1524, 0.0658, 0.1275 ] }, { "id": 5, "class_label": "person", "position": "middle-right", "bbox": [ 0.7319, 0.5332, 0.0706, 0.1206 ] }, { "id": 6, "class_label": "bicycle", "position": "top-center", "bbox": [ 0.5473, 0.1706, 0.1118, 0.0924 ] }, { "id": 7, "class_label": "truck", "position": "middle-left", "bbox": [ 0.1789, 0.3776, 0.1965, 0.1238 ] }, { "id": 8, "class_label": "bicycle", "position": "middle-right", "bbox": [ 0.687, 0.3584, 0.0911, 0.0862 ] }, { "id": 9, "class_label": "truck", "position": "bottom-left", "bbox": [ 0.1022, 0.8109, 0.1542, 0.1721 ] }, { "id": 10, "class_label": "bicycle", "position": "top-left", "bbox": [ 0.2267, 0.2414, 0.0651, 0.0881 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.6835, 0.2478, 0.2968, 0.2427 ], "class_label": "building" }, { "id": 1, "bbox": [ 0.1514, 0.1444, 0.0881, 0.1145 ], "class_label": "bicycle" }, { "id": 2, "bbox": [ 0.6847, 0.0425, 0.2652, 0.2451 ], "class_label": "building" }, { "id": 3, "bbox": [ 0.7209, 0.4596, 0.0752, 0.1498 ], "class_label": "person" }, { "id": 4, "bbox": [ 0.1139, 0.1524, 0.0658, 0.1275 ], "class_label": "person" }, { "id": 5, "bbox": [ 0.7319, 0.5332, 0.0706, 0.1206 ], "class_label": "person" }, { "id": 6, "bbox": [ 0.5473, 0.1706, 0.1118, 0.0924 ], "class_label": "bicycle" }, { "id": 7, "bbox": [ 0.1789, 0.3776, 0.1965, 0.1238 ], "class_label": "truck" }, { "id": 8, "bbox": [ 0.687, 0.3584, 0.0911, 0.0862 ], "class_label": "bicycle" }, { "id": 9, "bbox": [ 0.1022, 0.8109, 0.1542, 0.1721 ], "class_label": "truck" }, { "id": 10, "bbox": [ 0.2267, 0.2414, 0.0651, 0.0881 ], "class_label": "bicycle" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1016 }, { "scene_id": "fix_bboxes_sample_017", "scene_type": "residential_area", "scene_description": "A quiet residential neighborhood with houses, trees, and occasional pedestrians. The scene contains 9 objects: a car at middle-left (bbox: x=0.01, y=0.38, w=0.22, h=0.10); a bench at top-left (bbox: x=0.21, y=0.17, w=0.13, h=0.08); a car at bottom-center (bbox: x=0.48, y=0.74, w=0.19, h=0.10); a building at bottom-center (bbox: x=0.25, y=0.55, w=0.35, h=0.36); a bench at top-center (bbox: x=0.35, y=0.10, w=0.09, h=0.06); a car at bottom-left (bbox: x=0.16, y=0.75, w=0.16, h=0.09); a bench at top-left (bbox: x=0.15, y=0.26, w=0.13, h=0.07); a building at bottom-left (bbox: x=0.07, y=0.58, w=0.21, h=0.27); a building at top-right (bbox: x=0.68, y=0.04, w=0.24, h=0.33).", "objects": [ { "id": 0, "class_label": "car", "position": "middle-left", "bbox": [ 0.0103, 0.3792, 0.2201, 0.1047 ] }, { "id": 1, "class_label": "bench", "position": "top-left", "bbox": [ 0.2069, 0.167, 0.1265, 0.0754 ] }, { "id": 2, "class_label": "car", "position": "bottom-center", "bbox": [ 0.4759, 0.7439, 0.1868, 0.0983 ] }, { "id": 3, "class_label": "building", "position": "bottom-center", "bbox": [ 0.2524, 0.5471, 0.3467, 0.3594 ] }, { "id": 4, "class_label": "bench", "position": "top-center", "bbox": [ 0.3476, 0.1013, 0.0909, 0.0561 ] }, { "id": 5, "class_label": "car", "position": "bottom-left", "bbox": [ 0.163, 0.7538, 0.159, 0.0875 ] }, { "id": 6, "class_label": "bench", "position": "top-left", "bbox": [ 0.1537, 0.2634, 0.1251, 0.0694 ] }, { "id": 7, "class_label": "building", "position": "bottom-left", "bbox": [ 0.072, 0.5765, 0.2148, 0.2665 ] }, { "id": 8, "class_label": "building", "position": "top-right", "bbox": [ 0.6767, 0.0397, 0.2447, 0.3283 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.0103, 0.3792, 0.2201, 0.1047 ], "class_label": "car" }, { "id": 1, "bbox": [ 0.2069, 0.167, 0.1265, 0.0754 ], "class_label": "bench" }, { "id": 2, "bbox": [ 0.4759, 0.7439, 0.1868, 0.0983 ], "class_label": "car" }, { "id": 3, "bbox": [ 0.2524, 0.5471, 0.3467, 0.3594 ], "class_label": "building" }, { "id": 4, "bbox": [ 0.3476, 0.1013, 0.0909, 0.0561 ], "class_label": "bench" }, { "id": 5, "bbox": [ 0.163, 0.7538, 0.159, 0.0875 ], "class_label": "car" }, { "id": 6, "bbox": [ 0.1537, 0.2634, 0.1251, 0.0694 ], "class_label": "bench" }, { "id": 7, "bbox": [ 0.072, 0.5765, 0.2148, 0.2665 ], "class_label": "building" }, { "id": 8, "bbox": [ 0.6767, 0.0397, 0.2447, 0.3283 ], "class_label": "building" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1017 }, { "scene_id": "fix_bboxes_sample_018", "scene_type": "residential_area", "scene_description": "A quiet residential neighborhood with houses, trees, and occasional pedestrians. The scene contains 9 objects: a bench at bottom-left (bbox: x=0.19, y=0.73, w=0.13, h=0.07); a cat at top-right (bbox: x=0.74, y=0.23, w=0.08, h=0.06); a bench at top-center (bbox: x=0.39, y=0.17, w=0.14, h=0.07); a tree at top-right (bbox: x=0.82, y=0.11, w=0.12, h=0.16); a tree at bottom-left (bbox: x=0.09, y=0.60, w=0.11, h=0.34); a person at bottom-center (bbox: x=0.43, y=0.83, w=0.06, h=0.10); a building at middle-right (bbox: x=0.73, y=0.36, w=0.17, h=0.39); a building at top-center (bbox: x=0.31, y=0.00, w=0.33, h=0.25); a person at middle-left (bbox: x=0.19, y=0.53, w=0.05, h=0.13).", "objects": [ { "id": 0, "class_label": "bench", "position": "bottom-left", "bbox": [ 0.1925, 0.7282, 0.1296, 0.0703 ] }, { "id": 1, "class_label": "cat", "position": "top-right", "bbox": [ 0.7442, 0.2291, 0.0761, 0.0617 ] }, { "id": 2, "class_label": "bench", "position": "top-center", "bbox": [ 0.3946, 0.1722, 0.1436, 0.0712 ] }, { "id": 3, "class_label": "tree", "position": "top-right", "bbox": [ 0.819, 0.1056, 0.121, 0.1604 ] }, { "id": 4, "class_label": "tree", "position": "bottom-left", "bbox": [ 0.0874, 0.5951, 0.1055, 0.3371 ] }, { "id": 5, "class_label": "person", "position": "bottom-center", "bbox": [ 0.4286, 0.8321, 0.0595, 0.1022 ] }, { "id": 6, "class_label": "building", "position": "middle-right", "bbox": [ 0.7296, 0.3554, 0.1745, 0.393 ] }, { "id": 7, "class_label": "building", "position": "top-center", "bbox": [ 0.3066, 0.0, 0.3277, 0.2518 ] }, { "id": 8, "class_label": "person", "position": "middle-left", "bbox": [ 0.1897, 0.5302, 0.0467, 0.1304 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.1925, 0.7282, 0.1296, 0.0703 ], "class_label": "bench" }, { "id": 1, "bbox": [ 0.7442, 0.2291, 0.0761, 0.0617 ], "class_label": "cat" }, { "id": 2, "bbox": [ 0.3946, 0.1722, 0.1436, 0.0712 ], "class_label": "bench" }, { "id": 3, "bbox": [ 0.819, 0.1056, 0.121, 0.1604 ], "class_label": "tree" }, { "id": 4, "bbox": [ 0.0874, 0.5951, 0.1055, 0.3371 ], "class_label": "tree" }, { "id": 5, "bbox": [ 0.4286, 0.8321, 0.0595, 0.1022 ], "class_label": "person" }, { "id": 6, "bbox": [ 0.7296, 0.3554, 0.1745, 0.393 ], "class_label": "building" }, { "id": 7, "bbox": [ 0.3066, 0.0, 0.3277, 0.2518 ], "class_label": "building" }, { "id": 8, "bbox": [ 0.1897, 0.5302, 0.0467, 0.1304 ], "class_label": "person" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1018 }, { "scene_id": "fix_bboxes_sample_019", "scene_type": "urban_street", "scene_description": "A busy urban street scene with vehicles, pedestrians, and city infrastructure. The scene contains 7 objects: a bicycle at top-center (bbox: x=0.59, y=0.26, w=0.11, h=0.09); a bicycle at center (bbox: x=0.37, y=0.39, w=0.11, h=0.08); a building at bottom-left (bbox: x=0.08, y=0.59, w=0.16, h=0.41); a bench at top-left (bbox: x=0.15, y=0.11, w=0.14, h=0.06); a tree at bottom-left (bbox: x=0.22, y=0.60, w=0.11, h=0.24); a traffic_light at top-center (bbox: x=0.58, y=0.21, w=0.03, h=0.08); a truck at center (bbox: x=0.35, y=0.42, w=0.18, h=0.13).", "objects": [ { "id": 0, "class_label": "bicycle", "position": "top-center", "bbox": [ 0.5925, 0.2551, 0.1066, 0.0868 ] }, { "id": 1, "class_label": "bicycle", "position": "center", "bbox": [ 0.374, 0.3867, 0.1079, 0.0832 ] }, { "id": 2, "class_label": "building", "position": "bottom-left", "bbox": [ 0.0778, 0.5896, 0.1597, 0.4099 ] }, { "id": 3, "class_label": "bench", "position": "top-left", "bbox": [ 0.1513, 0.1056, 0.1385, 0.0648 ] }, { "id": 4, "class_label": "tree", "position": "bottom-left", "bbox": [ 0.2196, 0.6017, 0.1148, 0.2375 ] }, { "id": 5, "class_label": "traffic_light", "position": "top-center", "bbox": [ 0.5764, 0.2124, 0.029, 0.0831 ] }, { "id": 6, "class_label": "truck", "position": "center", "bbox": [ 0.3484, 0.4227, 0.1764, 0.1335 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.5925, 0.2551, 0.1066, 0.0868 ], "class_label": "bicycle" }, { "id": 1, "bbox": [ 0.374, 0.3867, 0.1079, 0.0832 ], "class_label": "bicycle" }, { "id": 2, "bbox": [ 0.0778, 0.5896, 0.1597, 0.4099 ], "class_label": "building" }, { "id": 3, "bbox": [ 0.1513, 0.1056, 0.1385, 0.0648 ], "class_label": "bench" }, { "id": 4, "bbox": [ 0.2196, 0.6017, 0.1148, 0.2375 ], "class_label": "tree" }, { "id": 5, "bbox": [ 0.5764, 0.2124, 0.029, 0.0831 ], "class_label": "traffic_light" }, { "id": 6, "bbox": [ 0.3484, 0.4227, 0.1764, 0.1335 ], "class_label": "truck" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1019 }, { "scene_id": "fix_bboxes_sample_020", "scene_type": "urban_street", "scene_description": "A busy urban street scene with vehicles, pedestrians, and city infrastructure. The scene contains 5 objects: a bench at bottom-left (bbox: x=0.08, y=0.72, w=0.12, h=0.06); a tree at bottom-right (bbox: x=0.84, y=0.54, w=0.12, h=0.33); a person at top-left (bbox: x=0.15, y=0.14, w=0.06, h=0.17); a truck at top-right (bbox: x=0.61, y=0.13, w=0.21, h=0.13); a tree at bottom-left (bbox: x=0.21, y=0.62, w=0.10, h=0.21).", "objects": [ { "id": 0, "class_label": "bench", "position": "bottom-left", "bbox": [ 0.0755, 0.7245, 0.1208, 0.0589 ] }, { "id": 1, "class_label": "tree", "position": "bottom-right", "bbox": [ 0.8413, 0.5443, 0.1167, 0.3309 ] }, { "id": 2, "class_label": "person", "position": "top-left", "bbox": [ 0.1521, 0.1438, 0.0642, 0.1744 ] }, { "id": 3, "class_label": "truck", "position": "top-right", "bbox": [ 0.614, 0.1308, 0.2084, 0.1323 ] }, { "id": 4, "class_label": "tree", "position": "bottom-left", "bbox": [ 0.2108, 0.6214, 0.0957, 0.2075 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.0755, 0.7245, 0.1208, 0.0589 ], "class_label": "bench" }, { "id": 1, "bbox": [ 0.8413, 0.5443, 0.1167, 0.3309 ], "class_label": "tree" }, { "id": 2, "bbox": [ 0.1521, 0.1438, 0.0642, 0.1744 ], "class_label": "person" }, { "id": 3, "bbox": [ 0.614, 0.1308, 0.2084, 0.1323 ], "class_label": "truck" }, { "id": 4, "bbox": [ 0.2108, 0.6214, 0.0957, 0.2075 ], "class_label": "tree" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1020 }, { "scene_id": "fix_bboxes_sample_021", "scene_type": "intersection", "scene_description": "A road intersection with traffic lights, vehicles, and crossing pedestrians. The scene contains 10 objects: a truck at middle-left (bbox: x=0.08, y=0.48, w=0.29, h=0.14); a person at top-center (bbox: x=0.38, y=0.19, w=0.04, h=0.17); a building at top-left (bbox: x=0.19, y=0.09, w=0.21, h=0.20); a traffic_light at top-center (bbox: x=0.47, y=0.15, w=0.02, h=0.07); a car at top-left (bbox: x=0.09, y=0.19, w=0.23, h=0.15); a bicycle at top-center (bbox: x=0.57, y=0.13, w=0.06, h=0.08); a truck at top-right (bbox: x=0.71, y=0.03, w=0.29, h=0.16); a traffic_light at bottom-right (bbox: x=0.83, y=0.73, w=0.02, h=0.08); a car at top-center (bbox: x=0.36, y=0.18, w=0.11, h=0.11); a traffic_light at center (bbox: x=0.38, y=0.52, w=0.02, h=0.10).", "objects": [ { "id": 0, "class_label": "truck", "position": "middle-left", "bbox": [ 0.0761, 0.4831, 0.2878, 0.1428 ] }, { "id": 1, "class_label": "person", "position": "top-center", "bbox": [ 0.3819, 0.1896, 0.0442, 0.1685 ] }, { "id": 2, "class_label": "building", "position": "top-left", "bbox": [ 0.1888, 0.0936, 0.2143, 0.2014 ] }, { "id": 3, "class_label": "traffic_light", "position": "top-center", "bbox": [ 0.4742, 0.1519, 0.0248, 0.0749 ] }, { "id": 4, "class_label": "car", "position": "top-left", "bbox": [ 0.0897, 0.1915, 0.2349, 0.1488 ] }, { "id": 5, "class_label": "bicycle", "position": "top-center", "bbox": [ 0.5714, 0.127, 0.0629, 0.0773 ] }, { "id": 6, "class_label": "truck", "position": "top-right", "bbox": [ 0.7055, 0.0257, 0.2945, 0.1563 ] }, { "id": 7, "class_label": "traffic_light", "position": "bottom-right", "bbox": [ 0.8327, 0.7276, 0.0218, 0.0759 ] }, { "id": 8, "class_label": "car", "position": "top-center", "bbox": [ 0.3557, 0.1788, 0.1101, 0.108 ] }, { "id": 9, "class_label": "traffic_light", "position": "center", "bbox": [ 0.3759, 0.5249, 0.0204, 0.0964 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.0761, 0.4831, 0.2878, 0.1428 ], "class_label": "truck" }, { "id": 1, "bbox": [ 0.3819, 0.1896, 0.0442, 0.1685 ], "class_label": "person" }, { "id": 2, "bbox": [ 0.1888, 0.0936, 0.2143, 0.2014 ], "class_label": "building" }, { "id": 3, "bbox": [ 0.4742, 0.1519, 0.0248, 0.0749 ], "class_label": "traffic_light" }, { "id": 4, "bbox": [ 0.0897, 0.1915, 0.2349, 0.1488 ], "class_label": "car" }, { "id": 5, "bbox": [ 0.5714, 0.127, 0.0629, 0.0773 ], "class_label": "bicycle" }, { "id": 6, "bbox": [ 0.7055, 0.0257, 0.2945, 0.1563 ], "class_label": "truck" }, { "id": 7, "bbox": [ 0.8327, 0.7276, 0.0218, 0.0759 ], "class_label": "traffic_light" }, { "id": 8, "bbox": [ 0.3557, 0.1788, 0.1101, 0.108 ], "class_label": "car" }, { "id": 9, "bbox": [ 0.3759, 0.5249, 0.0204, 0.0964 ], "class_label": "traffic_light" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1021 }, { "scene_id": "fix_bboxes_sample_022", "scene_type": "park", "scene_description": "A peaceful park setting with trees, benches, and people walking their pets. The scene contains 7 objects: a bench at middle-right (bbox: x=0.79, y=0.34, w=0.09, h=0.06); a cat at middle-left (bbox: x=0.17, y=0.40, w=0.06, h=0.06); a bicycle at bottom-left (bbox: x=0.20, y=0.80, w=0.08, h=0.07); a person at top-right (bbox: x=0.73, y=0.23, w=0.07, h=0.13); a bicycle at top-right (bbox: x=0.72, y=0.16, w=0.10, h=0.09); a bench at top-center (bbox: x=0.57, y=0.21, w=0.09, h=0.05); a bench at bottom-right (bbox: x=0.65, y=0.82, w=0.15, h=0.06).", "objects": [ { "id": 0, "class_label": "bench", "position": "middle-right", "bbox": [ 0.7906, 0.3396, 0.086, 0.0617 ] }, { "id": 1, "class_label": "cat", "position": "middle-left", "bbox": [ 0.168, 0.4021, 0.0573, 0.0628 ] }, { "id": 2, "class_label": "bicycle", "position": "bottom-left", "bbox": [ 0.1968, 0.7971, 0.0791, 0.0691 ] }, { "id": 3, "class_label": "person", "position": "top-right", "bbox": [ 0.726, 0.226, 0.0741, 0.1251 ] }, { "id": 4, "class_label": "bicycle", "position": "top-right", "bbox": [ 0.7193, 0.1624, 0.1014, 0.0882 ] }, { "id": 5, "class_label": "bench", "position": "top-center", "bbox": [ 0.5658, 0.2097, 0.0859, 0.053 ] }, { "id": 6, "class_label": "bench", "position": "bottom-right", "bbox": [ 0.6502, 0.8219, 0.1454, 0.0561 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.7906, 0.3396, 0.086, 0.0617 ], "class_label": "bench" }, { "id": 1, "bbox": [ 0.168, 0.4021, 0.0573, 0.0628 ], "class_label": "cat" }, { "id": 2, "bbox": [ 0.1968, 0.7971, 0.0791, 0.0691 ], "class_label": "bicycle" }, { "id": 3, "bbox": [ 0.726, 0.226, 0.0741, 0.1251 ], "class_label": "person" }, { "id": 4, "bbox": [ 0.7193, 0.1624, 0.1014, 0.0882 ], "class_label": "bicycle" }, { "id": 5, "bbox": [ 0.5658, 0.2097, 0.0859, 0.053 ], "class_label": "bench" }, { "id": 6, "bbox": [ 0.6502, 0.8219, 0.1454, 0.0561 ], "class_label": "bench" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1022 }, { "scene_id": "fix_bboxes_sample_023", "scene_type": "residential_area", "scene_description": "A quiet residential neighborhood with houses, trees, and occasional pedestrians. The scene contains 9 objects: a car at bottom-center (bbox: x=0.23, y=0.83, w=0.25, h=0.11); a car at middle-right (bbox: x=0.67, y=0.32, w=0.22, h=0.11); a dog at center (bbox: x=0.47, y=0.38, w=0.07, h=0.04); a dog at middle-left (bbox: x=0.25, y=0.56, w=0.06, h=0.05); a cat at middle-right (bbox: x=0.76, y=0.36, w=0.06, h=0.06); a bench at bottom-right (bbox: x=0.73, y=0.74, w=0.15, h=0.07); a dog at bottom-left (bbox: x=0.23, y=0.82, w=0.06, h=0.07); a tree at top-left (bbox: x=0.03, y=0.15, w=0.15, h=0.29); a dog at bottom-center (bbox: x=0.48, y=0.75, w=0.06, h=0.06).", "objects": [ { "id": 0, "class_label": "car", "position": "bottom-center", "bbox": [ 0.226, 0.8343, 0.2494, 0.1095 ] }, { "id": 1, "class_label": "car", "position": "middle-right", "bbox": [ 0.6739, 0.317, 0.2182, 0.1057 ] }, { "id": 2, "class_label": "dog", "position": "center", "bbox": [ 0.4658, 0.3784, 0.0732, 0.0442 ] }, { "id": 3, "class_label": "dog", "position": "middle-left", "bbox": [ 0.2471, 0.5592, 0.0551, 0.0505 ] }, { "id": 4, "class_label": "cat", "position": "middle-right", "bbox": [ 0.7589, 0.3571, 0.0588, 0.0555 ] }, { "id": 5, "class_label": "bench", "position": "bottom-right", "bbox": [ 0.7251, 0.7353, 0.1463, 0.0674 ] }, { "id": 6, "class_label": "dog", "position": "bottom-left", "bbox": [ 0.2316, 0.8176, 0.0595, 0.0653 ] }, { "id": 7, "class_label": "tree", "position": "top-left", "bbox": [ 0.0289, 0.1459, 0.1498, 0.2879 ] }, { "id": 8, "class_label": "dog", "position": "bottom-center", "bbox": [ 0.4767, 0.746, 0.0611, 0.0558 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.226, 0.8343, 0.2494, 0.1095 ], "class_label": "car" }, { "id": 1, "bbox": [ 0.6739, 0.317, 0.2182, 0.1057 ], "class_label": "car" }, { "id": 2, "bbox": [ 0.4658, 0.3784, 0.0732, 0.0442 ], "class_label": "dog" }, { "id": 3, "bbox": [ 0.2471, 0.5592, 0.0551, 0.0505 ], "class_label": "dog" }, { "id": 4, "bbox": [ 0.7589, 0.3571, 0.0588, 0.0555 ], "class_label": "cat" }, { "id": 5, "bbox": [ 0.7251, 0.7353, 0.1463, 0.0674 ], "class_label": "bench" }, { "id": 6, "bbox": [ 0.2316, 0.8176, 0.0595, 0.0653 ], "class_label": "dog" }, { "id": 7, "bbox": [ 0.0289, 0.1459, 0.1498, 0.2879 ], "class_label": "tree" }, { "id": 8, "bbox": [ 0.4767, 0.746, 0.0611, 0.0558 ], "class_label": "dog" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1023 }, { "scene_id": "fix_bboxes_sample_024", "scene_type": "urban_street", "scene_description": "A busy urban street scene with vehicles, pedestrians, and city infrastructure. The scene contains 8 objects: a tree at middle-right (bbox: x=0.65, y=0.34, w=0.14, h=0.32); a building at bottom-left (bbox: x=0.09, y=0.62, w=0.31, h=0.22); a tree at top-center (bbox: x=0.43, y=0.09, w=0.11, h=0.24); a building at top-center (bbox: x=0.50, y=0.01, w=0.16, h=0.22); a building at middle-right (bbox: x=0.66, y=0.30, w=0.34, h=0.20); a person at bottom-left (bbox: x=0.09, y=0.75, w=0.05, h=0.23); a truck at bottom-center (bbox: x=0.43, y=0.71, w=0.25, h=0.14); a car at top-left (bbox: x=0.16, y=0.08, w=0.24, h=0.10).", "objects": [ { "id": 0, "class_label": "tree", "position": "middle-right", "bbox": [ 0.6511, 0.3418, 0.1381, 0.3241 ] }, { "id": 1, "class_label": "building", "position": "bottom-left", "bbox": [ 0.0946, 0.616, 0.3088, 0.224 ] }, { "id": 2, "class_label": "tree", "position": "top-center", "bbox": [ 0.4295, 0.0864, 0.1087, 0.2372 ] }, { "id": 3, "class_label": "building", "position": "top-center", "bbox": [ 0.5025, 0.006, 0.157, 0.2209 ] }, { "id": 4, "class_label": "building", "position": "middle-right", "bbox": [ 0.6567, 0.3043, 0.3433, 0.2008 ] }, { "id": 5, "class_label": "person", "position": "bottom-left", "bbox": [ 0.0925, 0.7492, 0.0514, 0.2298 ] }, { "id": 6, "class_label": "truck", "position": "bottom-center", "bbox": [ 0.4332, 0.7096, 0.2523, 0.1449 ] }, { "id": 7, "class_label": "car", "position": "top-left", "bbox": [ 0.1615, 0.0823, 0.2365, 0.0973 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.6511, 0.3418, 0.1381, 0.3241 ], "class_label": "tree" }, { "id": 1, "bbox": [ 0.0946, 0.616, 0.3088, 0.224 ], "class_label": "building" }, { "id": 2, "bbox": [ 0.4295, 0.0864, 0.1087, 0.2372 ], "class_label": "tree" }, { "id": 3, "bbox": [ 0.5025, 0.006, 0.157, 0.2209 ], "class_label": "building" }, { "id": 4, "bbox": [ 0.6567, 0.3043, 0.3433, 0.2008 ], "class_label": "building" }, { "id": 5, "bbox": [ 0.0925, 0.7492, 0.0514, 0.2298 ], "class_label": "person" }, { "id": 6, "bbox": [ 0.4332, 0.7096, 0.2523, 0.1449 ], "class_label": "truck" }, { "id": 7, "bbox": [ 0.1615, 0.0823, 0.2365, 0.0973 ], "class_label": "car" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1024 }, { "scene_id": "fix_bboxes_sample_025", "scene_type": "park", "scene_description": "A peaceful park setting with trees, benches, and people walking their pets. The scene contains 4 objects: a person at middle-left (bbox: x=0.21, y=0.40, w=0.05, h=0.13); a cat at bottom-center (bbox: x=0.40, y=0.69, w=0.08, h=0.04); a person at middle-right (bbox: x=0.78, y=0.37, w=0.06, h=0.13); a bicycle at top-center (bbox: x=0.32, y=0.11, w=0.07, h=0.09).", "objects": [ { "id": 0, "class_label": "person", "position": "middle-left", "bbox": [ 0.2085, 0.398, 0.0538, 0.1344 ] }, { "id": 1, "class_label": "cat", "position": "bottom-center", "bbox": [ 0.4002, 0.6919, 0.076, 0.0441 ] }, { "id": 2, "class_label": "person", "position": "middle-right", "bbox": [ 0.7813, 0.3744, 0.0625, 0.131 ] }, { "id": 3, "class_label": "bicycle", "position": "top-center", "bbox": [ 0.3187, 0.1085, 0.0668, 0.091 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.2085, 0.398, 0.0538, 0.1344 ], "class_label": "person" }, { "id": 1, "bbox": [ 0.4002, 0.6919, 0.076, 0.0441 ], "class_label": "cat" }, { "id": 2, "bbox": [ 0.7813, 0.3744, 0.0625, 0.131 ], "class_label": "person" }, { "id": 3, "bbox": [ 0.3187, 0.1085, 0.0668, 0.091 ], "class_label": "bicycle" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1025 }, { "scene_id": "fix_bboxes_sample_026", "scene_type": "urban_street", "scene_description": "A busy urban street scene with vehicles, pedestrians, and city infrastructure. The scene contains 8 objects: a traffic_light at bottom-left (bbox: x=0.25, y=0.76, w=0.03, h=0.07); a building at bottom-right (bbox: x=0.64, y=0.66, w=0.30, h=0.23); a bicycle at center (bbox: x=0.30, y=0.30, w=0.10, h=0.11); a bench at middle-right (bbox: x=0.72, y=0.37, w=0.14, h=0.06); a tree at middle-right (bbox: x=0.80, y=0.31, w=0.09, h=0.33); a traffic_light at middle-left (bbox: x=0.10, y=0.45, w=0.03, h=0.08); a traffic_light at bottom-right (bbox: x=0.70, y=0.78, w=0.03, h=0.07); a tree at top-left (bbox: x=0.09, y=0.11, w=0.08, h=0.17).", "objects": [ { "id": 0, "class_label": "traffic_light", "position": "bottom-left", "bbox": [ 0.2471, 0.756, 0.0339, 0.0688 ] }, { "id": 1, "class_label": "building", "position": "bottom-right", "bbox": [ 0.6409, 0.6564, 0.2984, 0.2278 ] }, { "id": 2, "class_label": "bicycle", "position": "center", "bbox": [ 0.2987, 0.2987, 0.1032, 0.1091 ] }, { "id": 3, "class_label": "bench", "position": "middle-right", "bbox": [ 0.722, 0.3712, 0.1415, 0.0552 ] }, { "id": 4, "class_label": "tree", "position": "middle-right", "bbox": [ 0.8004, 0.3118, 0.0934, 0.3345 ] }, { "id": 5, "class_label": "traffic_light", "position": "middle-left", "bbox": [ 0.1024, 0.4536, 0.0277, 0.0794 ] }, { "id": 6, "class_label": "traffic_light", "position": "bottom-right", "bbox": [ 0.7002, 0.7791, 0.0348, 0.071 ] }, { "id": 7, "class_label": "tree", "position": "top-left", "bbox": [ 0.0887, 0.1055, 0.0817, 0.1699 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.2471, 0.756, 0.0339, 0.0688 ], "class_label": "traffic_light" }, { "id": 1, "bbox": [ 0.6409, 0.6564, 0.2984, 0.2278 ], "class_label": "building" }, { "id": 2, "bbox": [ 0.2987, 0.2987, 0.1032, 0.1091 ], "class_label": "bicycle" }, { "id": 3, "bbox": [ 0.722, 0.3712, 0.1415, 0.0552 ], "class_label": "bench" }, { "id": 4, "bbox": [ 0.8004, 0.3118, 0.0934, 0.3345 ], "class_label": "tree" }, { "id": 5, "bbox": [ 0.1024, 0.4536, 0.0277, 0.0794 ], "class_label": "traffic_light" }, { "id": 6, "bbox": [ 0.7002, 0.7791, 0.0348, 0.071 ], "class_label": "traffic_light" }, { "id": 7, "bbox": [ 0.0887, 0.1055, 0.0817, 0.1699 ], "class_label": "tree" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1026 }, { "scene_id": "fix_bboxes_sample_027", "scene_type": "urban_street", "scene_description": "A busy urban street scene with vehicles, pedestrians, and city infrastructure. The scene contains 5 objects: a truck at bottom-right (bbox: x=0.71, y=0.65, w=0.29, h=0.18); a tree at bottom-right (bbox: x=0.73, y=0.71, w=0.13, h=0.25); a traffic_light at middle-right (bbox: x=0.82, y=0.42, w=0.03, h=0.10); a truck at top-center (bbox: x=0.37, y=0.16, w=0.27, h=0.11); a traffic_light at top-left (bbox: x=0.18, y=0.14, w=0.03, h=0.07).", "objects": [ { "id": 0, "class_label": "truck", "position": "bottom-right", "bbox": [ 0.7115, 0.6546, 0.2885, 0.1785 ] }, { "id": 1, "class_label": "tree", "position": "bottom-right", "bbox": [ 0.7349, 0.7124, 0.1341, 0.2523 ] }, { "id": 2, "class_label": "traffic_light", "position": "middle-right", "bbox": [ 0.82, 0.4151, 0.0348, 0.0961 ] }, { "id": 3, "class_label": "truck", "position": "top-center", "bbox": [ 0.3718, 0.1645, 0.2717, 0.1062 ] }, { "id": 4, "class_label": "traffic_light", "position": "top-left", "bbox": [ 0.1825, 0.1439, 0.0288, 0.0749 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.7115, 0.6546, 0.2885, 0.1785 ], "class_label": "truck" }, { "id": 1, "bbox": [ 0.7349, 0.7124, 0.1341, 0.2523 ], "class_label": "tree" }, { "id": 2, "bbox": [ 0.82, 0.4151, 0.0348, 0.0961 ], "class_label": "traffic_light" }, { "id": 3, "bbox": [ 0.3718, 0.1645, 0.2717, 0.1062 ], "class_label": "truck" }, { "id": 4, "bbox": [ 0.1825, 0.1439, 0.0288, 0.0749 ], "class_label": "traffic_light" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1027 }, { "scene_id": "fix_bboxes_sample_028", "scene_type": "residential_area", "scene_description": "A quiet residential neighborhood with houses, trees, and occasional pedestrians. The scene contains 8 objects: a car at center (bbox: x=0.35, y=0.42, w=0.22, h=0.09); a tree at top-left (bbox: x=0.22, y=0.08, w=0.08, h=0.29); a building at middle-left (bbox: x=0.00, y=0.33, w=0.26, h=0.22); a building at top-right (bbox: x=0.78, y=0.00, w=0.22, h=0.31); a cat at center (bbox: x=0.49, y=0.36, w=0.04, h=0.05); a tree at bottom-center (bbox: x=0.47, y=0.57, w=0.08, h=0.29); a car at top-right (bbox: x=0.77, y=0.18, w=0.19, h=0.09); a car at top-center (bbox: x=0.54, y=0.14, w=0.15, h=0.09).", "objects": [ { "id": 0, "class_label": "car", "position": "center", "bbox": [ 0.3453, 0.4187, 0.2164, 0.0938 ] }, { "id": 1, "class_label": "tree", "position": "top-left", "bbox": [ 0.221, 0.0848, 0.0818, 0.2851 ] }, { "id": 2, "class_label": "building", "position": "middle-left", "bbox": [ 0.0, 0.3346, 0.2646, 0.2193 ] }, { "id": 3, "class_label": "building", "position": "top-right", "bbox": [ 0.7796, 0.0, 0.2204, 0.307 ] }, { "id": 4, "class_label": "cat", "position": "center", "bbox": [ 0.4851, 0.3594, 0.045, 0.0515 ] }, { "id": 5, "class_label": "tree", "position": "bottom-center", "bbox": [ 0.4723, 0.5729, 0.0825, 0.2899 ] }, { "id": 6, "class_label": "car", "position": "top-right", "bbox": [ 0.7663, 0.1775, 0.1874, 0.0895 ] }, { "id": 7, "class_label": "car", "position": "top-center", "bbox": [ 0.5394, 0.1364, 0.1549, 0.0877 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.3453, 0.4187, 0.2164, 0.0938 ], "class_label": "car" }, { "id": 1, "bbox": [ 0.221, 0.0848, 0.0818, 0.2851 ], "class_label": "tree" }, { "id": 2, "bbox": [ 0.0, 0.3346, 0.2646, 0.2193 ], "class_label": "building" }, { "id": 3, "bbox": [ 0.7796, 0.0, 0.2204, 0.307 ], "class_label": "building" }, { "id": 4, "bbox": [ 0.4851, 0.3594, 0.045, 0.0515 ], "class_label": "cat" }, { "id": 5, "bbox": [ 0.4723, 0.5729, 0.0825, 0.2899 ], "class_label": "tree" }, { "id": 6, "bbox": [ 0.7663, 0.1775, 0.1874, 0.0895 ], "class_label": "car" }, { "id": 7, "bbox": [ 0.5394, 0.1364, 0.1549, 0.0877 ], "class_label": "car" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1028 }, { "scene_id": "fix_bboxes_sample_029", "scene_type": "intersection", "scene_description": "A road intersection with traffic lights, vehicles, and crossing pedestrians. The scene contains 8 objects: a building at bottom-right (bbox: x=0.80, y=0.63, w=0.18, h=0.33); a traffic_light at center (bbox: x=0.62, y=0.34, w=0.04, h=0.07); a person at center (bbox: x=0.42, y=0.39, w=0.06, h=0.18); a building at top-center (bbox: x=0.40, y=0.17, w=0.33, h=0.25); a building at top-left (bbox: x=0.02, y=0.18, w=0.22, h=0.21); a bicycle at middle-right (bbox: x=0.70, y=0.46, w=0.09, h=0.11); a car at center (bbox: x=0.57, y=0.55, w=0.15, h=0.14); a bicycle at top-left (bbox: x=0.21, y=0.12, w=0.11, h=0.11).", "objects": [ { "id": 0, "class_label": "building", "position": "bottom-right", "bbox": [ 0.7968, 0.6344, 0.1824, 0.3336 ] }, { "id": 1, "class_label": "traffic_light", "position": "center", "bbox": [ 0.6215, 0.3431, 0.0353, 0.0739 ] }, { "id": 2, "class_label": "person", "position": "center", "bbox": [ 0.4191, 0.3949, 0.0565, 0.1808 ] }, { "id": 3, "class_label": "building", "position": "top-center", "bbox": [ 0.3959, 0.1729, 0.3252, 0.2495 ] }, { "id": 4, "class_label": "building", "position": "top-left", "bbox": [ 0.021, 0.18, 0.2179, 0.207 ] }, { "id": 5, "class_label": "bicycle", "position": "middle-right", "bbox": [ 0.6958, 0.4637, 0.0918, 0.1119 ] }, { "id": 6, "class_label": "car", "position": "center", "bbox": [ 0.5678, 0.5531, 0.1537, 0.1404 ] }, { "id": 7, "class_label": "bicycle", "position": "top-left", "bbox": [ 0.2134, 0.115, 0.113, 0.114 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.7968, 0.6344, 0.1824, 0.3336 ], "class_label": "building" }, { "id": 1, "bbox": [ 0.6215, 0.3431, 0.0353, 0.0739 ], "class_label": "traffic_light" }, { "id": 2, "bbox": [ 0.4191, 0.3949, 0.0565, 0.1808 ], "class_label": "person" }, { "id": 3, "bbox": [ 0.3959, 0.1729, 0.3252, 0.2495 ], "class_label": "building" }, { "id": 4, "bbox": [ 0.021, 0.18, 0.2179, 0.207 ], "class_label": "building" }, { "id": 5, "bbox": [ 0.6958, 0.4637, 0.0918, 0.1119 ], "class_label": "bicycle" }, { "id": 6, "bbox": [ 0.5678, 0.5531, 0.1537, 0.1404 ], "class_label": "car" }, { "id": 7, "bbox": [ 0.2134, 0.115, 0.113, 0.114 ], "class_label": "bicycle" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1029 }, { "scene_id": "fix_bboxes_sample_030", "scene_type": "urban_street", "scene_description": "A busy urban street scene with vehicles, pedestrians, and city infrastructure. The scene contains 8 objects: a person at middle-left (bbox: x=0.23, y=0.55, w=0.08, h=0.17); a tree at top-center (bbox: x=0.35, y=0.19, w=0.11, h=0.18); a person at top-right (bbox: x=0.74, y=0.14, w=0.04, h=0.23); a building at bottom-left (bbox: x=0.08, y=0.63, w=0.29, h=0.31); a bicycle at top-left (bbox: x=0.11, y=0.15, w=0.11, h=0.07); a building at bottom-left (bbox: x=0.21, y=0.60, w=0.18, h=0.40); a person at bottom-left (bbox: x=0.15, y=0.70, w=0.05, h=0.24); a bench at top-right (bbox: x=0.66, y=0.12, w=0.08, h=0.06).", "objects": [ { "id": 0, "class_label": "person", "position": "middle-left", "bbox": [ 0.2348, 0.549, 0.0776, 0.1678 ] }, { "id": 1, "class_label": "tree", "position": "top-center", "bbox": [ 0.3508, 0.1897, 0.1114, 0.1785 ] }, { "id": 2, "class_label": "person", "position": "top-right", "bbox": [ 0.7399, 0.1387, 0.0405, 0.2253 ] }, { "id": 3, "class_label": "building", "position": "bottom-left", "bbox": [ 0.0809, 0.6257, 0.2899, 0.3076 ] }, { "id": 4, "class_label": "bicycle", "position": "top-left", "bbox": [ 0.11, 0.1537, 0.1064, 0.0687 ] }, { "id": 5, "class_label": "building", "position": "bottom-left", "bbox": [ 0.2061, 0.6045, 0.1813, 0.3955 ] }, { "id": 6, "class_label": "person", "position": "bottom-left", "bbox": [ 0.149, 0.702, 0.0497, 0.2428 ] }, { "id": 7, "class_label": "bench", "position": "top-right", "bbox": [ 0.6585, 0.1248, 0.0831, 0.0594 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.2348, 0.549, 0.0776, 0.1678 ], "class_label": "person" }, { "id": 1, "bbox": [ 0.3508, 0.1897, 0.1114, 0.1785 ], "class_label": "tree" }, { "id": 2, "bbox": [ 0.7399, 0.1387, 0.0405, 0.2253 ], "class_label": "person" }, { "id": 3, "bbox": [ 0.0809, 0.6257, 0.2899, 0.3076 ], "class_label": "building" }, { "id": 4, "bbox": [ 0.11, 0.1537, 0.1064, 0.0687 ], "class_label": "bicycle" }, { "id": 5, "bbox": [ 0.2061, 0.6045, 0.1813, 0.3955 ], "class_label": "building" }, { "id": 6, "bbox": [ 0.149, 0.702, 0.0497, 0.2428 ], "class_label": "person" }, { "id": 7, "bbox": [ 0.6585, 0.1248, 0.0831, 0.0594 ], "class_label": "bench" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1030 }, { "scene_id": "fix_bboxes_sample_031", "scene_type": "intersection", "scene_description": "A road intersection with traffic lights, vehicles, and crossing pedestrians. The scene contains 10 objects: a car at top-left (bbox: x=0.00, y=0.21, w=0.23, h=0.10); a person at center (bbox: x=0.42, y=0.54, w=0.06, h=0.16); a traffic_light at top-right (bbox: x=0.73, y=0.21, w=0.02, h=0.09); a bicycle at top-left (bbox: x=0.20, y=0.09, w=0.09, h=0.11); a car at middle-right (bbox: x=0.67, y=0.47, w=0.10, h=0.09); a traffic_light at bottom-left (bbox: x=0.16, y=0.84, w=0.02, h=0.09); a person at bottom-center (bbox: x=0.46, y=0.61, w=0.06, h=0.22); a person at bottom-center (bbox: x=0.42, y=0.77, w=0.05, h=0.18); a car at middle-left (bbox: x=0.15, y=0.51, w=0.13, h=0.12); a bicycle at bottom-center (bbox: x=0.56, y=0.84, w=0.09, h=0.11).", "objects": [ { "id": 0, "class_label": "car", "position": "top-left", "bbox": [ 0.0, 0.2107, 0.2272, 0.102 ] }, { "id": 1, "class_label": "person", "position": "center", "bbox": [ 0.4151, 0.5396, 0.0586, 0.1595 ] }, { "id": 2, "class_label": "traffic_light", "position": "top-right", "bbox": [ 0.7295, 0.2146, 0.0202, 0.0927 ] }, { "id": 3, "class_label": "bicycle", "position": "top-left", "bbox": [ 0.2031, 0.0868, 0.0947, 0.1104 ] }, { "id": 4, "class_label": "car", "position": "middle-right", "bbox": [ 0.6692, 0.4736, 0.105, 0.0916 ] }, { "id": 5, "class_label": "traffic_light", "position": "bottom-left", "bbox": [ 0.1598, 0.8413, 0.0233, 0.0907 ] }, { "id": 6, "class_label": "person", "position": "bottom-center", "bbox": [ 0.4627, 0.6145, 0.055, 0.2173 ] }, { "id": 7, "class_label": "person", "position": "bottom-center", "bbox": [ 0.4204, 0.7677, 0.0539, 0.183 ] }, { "id": 8, "class_label": "car", "position": "middle-left", "bbox": [ 0.1522, 0.5082, 0.1294, 0.115 ] }, { "id": 9, "class_label": "bicycle", "position": "bottom-center", "bbox": [ 0.5584, 0.8444, 0.0886, 0.1096 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.0, 0.2107, 0.2272, 0.102 ], "class_label": "car" }, { "id": 1, "bbox": [ 0.4151, 0.5396, 0.0586, 0.1595 ], "class_label": "person" }, { "id": 2, "bbox": [ 0.7295, 0.2146, 0.0202, 0.0927 ], "class_label": "traffic_light" }, { "id": 3, "bbox": [ 0.2031, 0.0868, 0.0947, 0.1104 ], "class_label": "bicycle" }, { "id": 4, "bbox": [ 0.6692, 0.4736, 0.105, 0.0916 ], "class_label": "car" }, { "id": 5, "bbox": [ 0.1598, 0.8413, 0.0233, 0.0907 ], "class_label": "traffic_light" }, { "id": 6, "bbox": [ 0.4627, 0.6145, 0.055, 0.2173 ], "class_label": "person" }, { "id": 7, "bbox": [ 0.4204, 0.7677, 0.0539, 0.183 ], "class_label": "person" }, { "id": 8, "bbox": [ 0.1522, 0.5082, 0.1294, 0.115 ], "class_label": "car" }, { "id": 9, "bbox": [ 0.5584, 0.8444, 0.0886, 0.1096 ], "class_label": "bicycle" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1031 }, { "scene_id": "fix_bboxes_sample_032", "scene_type": "residential_area", "scene_description": "A quiet residential neighborhood with houses, trees, and occasional pedestrians. The scene contains 9 objects: a dog at bottom-left (bbox: x=0.15, y=0.82, w=0.09, h=0.05); a building at middle-right (bbox: x=0.71, y=0.38, w=0.15, h=0.28); a person at bottom-right (bbox: x=0.79, y=0.75, w=0.04, h=0.23); a person at center (bbox: x=0.58, y=0.31, w=0.04, h=0.11); a car at bottom-right (bbox: x=0.59, y=0.69, w=0.23, h=0.15); a car at bottom-right (bbox: x=0.69, y=0.82, w=0.21, h=0.13); a cat at bottom-center (bbox: x=0.35, y=0.80, w=0.04, h=0.06); a car at bottom-left (bbox: x=0.15, y=0.69, w=0.18, h=0.09); a dog at bottom-center (bbox: x=0.54, y=0.71, w=0.07, h=0.08).", "objects": [ { "id": 0, "class_label": "dog", "position": "bottom-left", "bbox": [ 0.1472, 0.8167, 0.0891, 0.0535 ] }, { "id": 1, "class_label": "building", "position": "middle-right", "bbox": [ 0.7055, 0.3759, 0.1533, 0.2769 ] }, { "id": 2, "class_label": "person", "position": "bottom-right", "bbox": [ 0.7937, 0.7475, 0.041, 0.2309 ] }, { "id": 3, "class_label": "person", "position": "center", "bbox": [ 0.5798, 0.3081, 0.0429, 0.1066 ] }, { "id": 4, "class_label": "car", "position": "bottom-right", "bbox": [ 0.5904, 0.6895, 0.2349, 0.1475 ] }, { "id": 5, "class_label": "car", "position": "bottom-right", "bbox": [ 0.6867, 0.8245, 0.2102, 0.1252 ] }, { "id": 6, "class_label": "cat", "position": "bottom-center", "bbox": [ 0.3479, 0.7959, 0.042, 0.0597 ] }, { "id": 7, "class_label": "car", "position": "bottom-left", "bbox": [ 0.1513, 0.6902, 0.1826, 0.0943 ] }, { "id": 8, "class_label": "dog", "position": "bottom-center", "bbox": [ 0.5369, 0.709, 0.0672, 0.0785 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.1472, 0.8167, 0.0891, 0.0535 ], "class_label": "dog" }, { "id": 1, "bbox": [ 0.7055, 0.3759, 0.1533, 0.2769 ], "class_label": "building" }, { "id": 2, "bbox": [ 0.7937, 0.7475, 0.041, 0.2309 ], "class_label": "person" }, { "id": 3, "bbox": [ 0.5798, 0.3081, 0.0429, 0.1066 ], "class_label": "person" }, { "id": 4, "bbox": [ 0.5904, 0.6895, 0.2349, 0.1475 ], "class_label": "car" }, { "id": 5, "bbox": [ 0.6867, 0.8245, 0.2102, 0.1252 ], "class_label": "car" }, { "id": 6, "bbox": [ 0.3479, 0.7959, 0.042, 0.0597 ], "class_label": "cat" }, { "id": 7, "bbox": [ 0.1513, 0.6902, 0.1826, 0.0943 ], "class_label": "car" }, { "id": 8, "bbox": [ 0.5369, 0.709, 0.0672, 0.0785 ], "class_label": "dog" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1032 }, { "scene_id": "fix_bboxes_sample_033", "scene_type": "park", "scene_description": "A peaceful park setting with trees, benches, and people walking their pets. The scene contains 6 objects: a tree at top-center (bbox: x=0.48, y=0.00, w=0.12, h=0.28); a tree at top-right (bbox: x=0.80, y=0.14, w=0.14, h=0.18); a person at top-right (bbox: x=0.77, y=0.14, w=0.06, h=0.11); a person at bottom-left (bbox: x=0.08, y=0.76, w=0.07, h=0.16); a cat at bottom-right (bbox: x=0.86, y=0.69, w=0.05, h=0.06); a person at center (bbox: x=0.44, y=0.40, w=0.04, h=0.10).", "objects": [ { "id": 0, "class_label": "tree", "position": "top-center", "bbox": [ 0.4849, 0.0, 0.1245, 0.2817 ] }, { "id": 1, "class_label": "tree", "position": "top-right", "bbox": [ 0.798, 0.1428, 0.1358, 0.1828 ] }, { "id": 2, "class_label": "person", "position": "top-right", "bbox": [ 0.7704, 0.1359, 0.0592, 0.1139 ] }, { "id": 3, "class_label": "person", "position": "bottom-left", "bbox": [ 0.0835, 0.7585, 0.0655, 0.1561 ] }, { "id": 4, "class_label": "cat", "position": "bottom-right", "bbox": [ 0.8637, 0.686, 0.0463, 0.0624 ] }, { "id": 5, "class_label": "person", "position": "center", "bbox": [ 0.4352, 0.4014, 0.0406, 0.1034 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.4849, 0.0, 0.1245, 0.2817 ], "class_label": "tree" }, { "id": 1, "bbox": [ 0.798, 0.1428, 0.1358, 0.1828 ], "class_label": "tree" }, { "id": 2, "bbox": [ 0.7704, 0.1359, 0.0592, 0.1139 ], "class_label": "person" }, { "id": 3, "bbox": [ 0.0835, 0.7585, 0.0655, 0.1561 ], "class_label": "person" }, { "id": 4, "bbox": [ 0.8637, 0.686, 0.0463, 0.0624 ], "class_label": "cat" }, { "id": 5, "bbox": [ 0.4352, 0.4014, 0.0406, 0.1034 ], "class_label": "person" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1033 }, { "scene_id": "fix_bboxes_sample_034", "scene_type": "urban_street", "scene_description": "A busy urban street scene with vehicles, pedestrians, and city infrastructure. The scene contains 10 objects: a truck at top-center (bbox: x=0.50, y=0.20, w=0.16, h=0.11); a bicycle at top-right (bbox: x=0.67, y=0.18, w=0.08, h=0.06); a bicycle at bottom-center (bbox: x=0.34, y=0.71, w=0.09, h=0.07); a truck at bottom-right (bbox: x=0.69, y=0.64, w=0.28, h=0.14); a traffic_light at middle-left (bbox: x=0.25, y=0.50, w=0.03, h=0.06); a person at top-center (bbox: x=0.46, y=0.04, w=0.06, h=0.22); a tree at bottom-center (bbox: x=0.35, y=0.70, w=0.11, h=0.28); a tree at middle-left (bbox: x=0.24, y=0.55, w=0.10, h=0.18); a person at top-center (bbox: x=0.42, y=0.14, w=0.04, h=0.17); a building at top-left (bbox: x=0.09, y=0.05, w=0.28, h=0.43).", "objects": [ { "id": 0, "class_label": "truck", "position": "top-center", "bbox": [ 0.4954, 0.1986, 0.1643, 0.106 ] }, { "id": 1, "class_label": "bicycle", "position": "top-right", "bbox": [ 0.671, 0.1829, 0.0817, 0.061 ] }, { "id": 2, "class_label": "bicycle", "position": "bottom-center", "bbox": [ 0.3411, 0.7143, 0.0878, 0.0676 ] }, { "id": 3, "class_label": "truck", "position": "bottom-right", "bbox": [ 0.6887, 0.6388, 0.2841, 0.1367 ] }, { "id": 4, "class_label": "traffic_light", "position": "middle-left", "bbox": [ 0.2519, 0.5019, 0.0266, 0.0604 ] }, { "id": 5, "class_label": "person", "position": "top-center", "bbox": [ 0.4613, 0.0433, 0.0575, 0.2195 ] }, { "id": 6, "class_label": "tree", "position": "bottom-center", "bbox": [ 0.3545, 0.7046, 0.1075, 0.285 ] }, { "id": 7, "class_label": "tree", "position": "middle-left", "bbox": [ 0.2367, 0.5461, 0.0953, 0.1762 ] }, { "id": 8, "class_label": "person", "position": "top-center", "bbox": [ 0.4185, 0.1374, 0.0416, 0.1651 ] }, { "id": 9, "class_label": "building", "position": "top-left", "bbox": [ 0.0872, 0.0469, 0.276, 0.4264 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.4954, 0.1986, 0.1643, 0.106 ], "class_label": "truck" }, { "id": 1, "bbox": [ 0.671, 0.1829, 0.0817, 0.061 ], "class_label": "bicycle" }, { "id": 2, "bbox": [ 0.3411, 0.7143, 0.0878, 0.0676 ], "class_label": "bicycle" }, { "id": 3, "bbox": [ 0.6887, 0.6388, 0.2841, 0.1367 ], "class_label": "truck" }, { "id": 4, "bbox": [ 0.2519, 0.5019, 0.0266, 0.0604 ], "class_label": "traffic_light" }, { "id": 5, "bbox": [ 0.4613, 0.0433, 0.0575, 0.2195 ], "class_label": "person" }, { "id": 6, "bbox": [ 0.3545, 0.7046, 0.1075, 0.285 ], "class_label": "tree" }, { "id": 7, "bbox": [ 0.2367, 0.5461, 0.0953, 0.1762 ], "class_label": "tree" }, { "id": 8, "bbox": [ 0.4185, 0.1374, 0.0416, 0.1651 ], "class_label": "person" }, { "id": 9, "bbox": [ 0.0872, 0.0469, 0.276, 0.4264 ], "class_label": "building" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1034 }, { "scene_id": "fix_bboxes_sample_035", "scene_type": "residential_area", "scene_description": "A quiet residential neighborhood with houses, trees, and occasional pedestrians. The scene contains 5 objects: a person at top-right (bbox: x=0.78, y=0.07, w=0.04, h=0.16); a tree at top-left (bbox: x=0.17, y=0.04, w=0.10, h=0.21); a tree at bottom-left (bbox: x=0.19, y=0.62, w=0.15, h=0.23); a cat at center (bbox: x=0.37, y=0.35, w=0.04, h=0.05); a dog at middle-right (bbox: x=0.76, y=0.48, w=0.10, h=0.08).", "objects": [ { "id": 0, "class_label": "person", "position": "top-right", "bbox": [ 0.7817, 0.0675, 0.0429, 0.1569 ] }, { "id": 1, "class_label": "tree", "position": "top-left", "bbox": [ 0.1659, 0.0423, 0.1038, 0.2107 ] }, { "id": 2, "class_label": "tree", "position": "bottom-left", "bbox": [ 0.1877, 0.6171, 0.1469, 0.2276 ] }, { "id": 3, "class_label": "cat", "position": "center", "bbox": [ 0.3696, 0.354, 0.0412, 0.0459 ] }, { "id": 4, "class_label": "dog", "position": "middle-right", "bbox": [ 0.76, 0.4825, 0.0957, 0.0767 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.7817, 0.0675, 0.0429, 0.1569 ], "class_label": "person" }, { "id": 1, "bbox": [ 0.1659, 0.0423, 0.1038, 0.2107 ], "class_label": "tree" }, { "id": 2, "bbox": [ 0.1877, 0.6171, 0.1469, 0.2276 ], "class_label": "tree" }, { "id": 3, "bbox": [ 0.3696, 0.354, 0.0412, 0.0459 ], "class_label": "cat" }, { "id": 4, "bbox": [ 0.76, 0.4825, 0.0957, 0.0767 ], "class_label": "dog" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1035 }, { "scene_id": "fix_bboxes_sample_036", "scene_type": "residential_area", "scene_description": "A quiet residential neighborhood with houses, trees, and occasional pedestrians. The scene contains 9 objects: a cat at top-center (bbox: x=0.43, y=0.11, w=0.07, h=0.04); a person at middle-right (bbox: x=0.77, y=0.53, w=0.06, h=0.12); a person at top-right (bbox: x=0.78, y=0.15, w=0.07, h=0.24); a building at top-right (bbox: x=0.76, y=0.01, w=0.21, h=0.29); a tree at middle-right (bbox: x=0.71, y=0.27, w=0.08, h=0.33); a person at middle-right (bbox: x=0.85, y=0.46, w=0.06, h=0.24); a building at bottom-center (bbox: x=0.29, y=0.73, w=0.34, h=0.22); a dog at top-left (bbox: x=0.10, y=0.13, w=0.06, h=0.07); a building at middle-left (bbox: x=0.11, y=0.47, w=0.28, h=0.27).", "objects": [ { "id": 0, "class_label": "cat", "position": "top-center", "bbox": [ 0.4257, 0.1077, 0.0688, 0.0412 ] }, { "id": 1, "class_label": "person", "position": "middle-right", "bbox": [ 0.7702, 0.5322, 0.0648, 0.1158 ] }, { "id": 2, "class_label": "person", "position": "top-right", "bbox": [ 0.7805, 0.1484, 0.0713, 0.2435 ] }, { "id": 3, "class_label": "building", "position": "top-right", "bbox": [ 0.7631, 0.0057, 0.2099, 0.2926 ] }, { "id": 4, "class_label": "tree", "position": "middle-right", "bbox": [ 0.7127, 0.2677, 0.0807, 0.3346 ] }, { "id": 5, "class_label": "person", "position": "middle-right", "bbox": [ 0.8519, 0.4633, 0.0617, 0.243 ] }, { "id": 6, "class_label": "building", "position": "bottom-center", "bbox": [ 0.2918, 0.729, 0.3355, 0.22 ] }, { "id": 7, "class_label": "dog", "position": "top-left", "bbox": [ 0.0957, 0.1269, 0.0625, 0.0733 ] }, { "id": 8, "class_label": "building", "position": "middle-left", "bbox": [ 0.1101, 0.4735, 0.2836, 0.2651 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.4257, 0.1077, 0.0688, 0.0412 ], "class_label": "cat" }, { "id": 1, "bbox": [ 0.7702, 0.5322, 0.0648, 0.1158 ], "class_label": "person" }, { "id": 2, "bbox": [ 0.7805, 0.1484, 0.0713, 0.2435 ], "class_label": "person" }, { "id": 3, "bbox": [ 0.7631, 0.0057, 0.2099, 0.2926 ], "class_label": "building" }, { "id": 4, "bbox": [ 0.7127, 0.2677, 0.0807, 0.3346 ], "class_label": "tree" }, { "id": 5, "bbox": [ 0.8519, 0.4633, 0.0617, 0.243 ], "class_label": "person" }, { "id": 6, "bbox": [ 0.2918, 0.729, 0.3355, 0.22 ], "class_label": "building" }, { "id": 7, "bbox": [ 0.0957, 0.1269, 0.0625, 0.0733 ], "class_label": "dog" }, { "id": 8, "bbox": [ 0.1101, 0.4735, 0.2836, 0.2651 ], "class_label": "building" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1036 }, { "scene_id": "fix_bboxes_sample_037", "scene_type": "urban_street", "scene_description": "A busy urban street scene with vehicles, pedestrians, and city infrastructure. The scene contains 8 objects: a bench at top-center (bbox: x=0.39, y=0.12, w=0.08, h=0.06); a person at top-left (bbox: x=0.08, y=0.01, w=0.06, h=0.20); a tree at top-left (bbox: x=0.03, y=0.07, w=0.14, h=0.21); a bicycle at top-right (bbox: x=0.76, y=0.24, w=0.11, h=0.08); a bicycle at bottom-left (bbox: x=0.15, y=0.68, w=0.08, h=0.09); a traffic_light at middle-right (bbox: x=0.84, y=0.53, w=0.04, h=0.07); a bicycle at top-right (bbox: x=0.74, y=0.22, w=0.10, h=0.09); a building at middle-right (bbox: x=0.65, y=0.22, w=0.30, h=0.28).", "objects": [ { "id": 0, "class_label": "bench", "position": "top-center", "bbox": [ 0.3941, 0.1196, 0.0832, 0.0606 ] }, { "id": 1, "class_label": "person", "position": "top-left", "bbox": [ 0.0781, 0.0149, 0.0605, 0.1986 ] }, { "id": 2, "class_label": "tree", "position": "top-left", "bbox": [ 0.0315, 0.071, 0.139, 0.2109 ] }, { "id": 3, "class_label": "bicycle", "position": "top-right", "bbox": [ 0.7555, 0.243, 0.1105, 0.0825 ] }, { "id": 4, "class_label": "bicycle", "position": "bottom-left", "bbox": [ 0.1459, 0.6822, 0.0845, 0.0883 ] }, { "id": 5, "class_label": "traffic_light", "position": "middle-right", "bbox": [ 0.8411, 0.5273, 0.0355, 0.0747 ] }, { "id": 6, "class_label": "bicycle", "position": "top-right", "bbox": [ 0.7415, 0.2194, 0.096, 0.0944 ] }, { "id": 7, "class_label": "building", "position": "middle-right", "bbox": [ 0.6526, 0.2154, 0.2994, 0.2751 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.3941, 0.1196, 0.0832, 0.0606 ], "class_label": "bench" }, { "id": 1, "bbox": [ 0.0781, 0.0149, 0.0605, 0.1986 ], "class_label": "person" }, { "id": 2, "bbox": [ 0.0315, 0.071, 0.139, 0.2109 ], "class_label": "tree" }, { "id": 3, "bbox": [ 0.7555, 0.243, 0.1105, 0.0825 ], "class_label": "bicycle" }, { "id": 4, "bbox": [ 0.1459, 0.6822, 0.0845, 0.0883 ], "class_label": "bicycle" }, { "id": 5, "bbox": [ 0.8411, 0.5273, 0.0355, 0.0747 ], "class_label": "traffic_light" }, { "id": 6, "bbox": [ 0.7415, 0.2194, 0.096, 0.0944 ], "class_label": "bicycle" }, { "id": 7, "bbox": [ 0.6526, 0.2154, 0.2994, 0.2751 ], "class_label": "building" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1037 }, { "scene_id": "fix_bboxes_sample_038", "scene_type": "urban_street", "scene_description": "A busy urban street scene with vehicles, pedestrians, and city infrastructure. The scene contains 5 objects: a traffic_light at top-right (bbox: x=0.78, y=0.15, w=0.02, h=0.08); a person at top-left (bbox: x=0.13, y=0.00, w=0.04, h=0.24); a truck at bottom-left (bbox: x=0.16, y=0.81, w=0.24, h=0.15); a building at bottom-left (bbox: x=0.00, y=0.58, w=0.30, h=0.42); a bicycle at top-right (bbox: x=0.70, y=0.18, w=0.10, h=0.11).", "objects": [ { "id": 0, "class_label": "traffic_light", "position": "top-right", "bbox": [ 0.7804, 0.1539, 0.0225, 0.0796 ] }, { "id": 1, "class_label": "person", "position": "top-left", "bbox": [ 0.1323, 0.0, 0.0416, 0.2396 ] }, { "id": 2, "class_label": "truck", "position": "bottom-left", "bbox": [ 0.1603, 0.8082, 0.2447, 0.1516 ] }, { "id": 3, "class_label": "building", "position": "bottom-left", "bbox": [ 0.0, 0.5787, 0.2987, 0.4213 ] }, { "id": 4, "class_label": "bicycle", "position": "top-right", "bbox": [ 0.6981, 0.1837, 0.0988, 0.1115 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.7804, 0.1539, 0.0225, 0.0796 ], "class_label": "traffic_light" }, { "id": 1, "bbox": [ 0.1323, 0.0, 0.0416, 0.2396 ], "class_label": "person" }, { "id": 2, "bbox": [ 0.1603, 0.8082, 0.2447, 0.1516 ], "class_label": "truck" }, { "id": 3, "bbox": [ 0.0, 0.5787, 0.2987, 0.4213 ], "class_label": "building" }, { "id": 4, "bbox": [ 0.6981, 0.1837, 0.0988, 0.1115 ], "class_label": "bicycle" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1038 }, { "scene_id": "fix_bboxes_sample_039", "scene_type": "park", "scene_description": "A peaceful park setting with trees, benches, and people walking their pets. The scene contains 5 objects: a bicycle at top-center (bbox: x=0.56, y=0.23, w=0.06, h=0.11); a dog at top-right (bbox: x=0.87, y=0.21, w=0.05, h=0.04); a bench at bottom-right (bbox: x=0.78, y=0.81, w=0.10, h=0.07); a person at bottom-left (bbox: x=0.14, y=0.82, w=0.07, h=0.15); a cat at top-left (bbox: x=0.12, y=0.10, w=0.07, h=0.04).", "objects": [ { "id": 0, "class_label": "bicycle", "position": "top-center", "bbox": [ 0.5574, 0.2275, 0.0633, 0.114 ] }, { "id": 1, "class_label": "dog", "position": "top-right", "bbox": [ 0.8738, 0.2147, 0.0512, 0.042 ] }, { "id": 2, "class_label": "bench", "position": "bottom-right", "bbox": [ 0.7845, 0.81, 0.1016, 0.0664 ] }, { "id": 3, "class_label": "person", "position": "bottom-left", "bbox": [ 0.1429, 0.8182, 0.0722, 0.1486 ] }, { "id": 4, "class_label": "cat", "position": "top-left", "bbox": [ 0.1181, 0.095, 0.0747, 0.0432 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.5574, 0.2275, 0.0633, 0.114 ], "class_label": "bicycle" }, { "id": 1, "bbox": [ 0.8738, 0.2147, 0.0512, 0.042 ], "class_label": "dog" }, { "id": 2, "bbox": [ 0.7845, 0.81, 0.1016, 0.0664 ], "class_label": "bench" }, { "id": 3, "bbox": [ 0.1429, 0.8182, 0.0722, 0.1486 ], "class_label": "person" }, { "id": 4, "bbox": [ 0.1181, 0.095, 0.0747, 0.0432 ], "class_label": "cat" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1039 }, { "scene_id": "fix_bboxes_sample_040", "scene_type": "parking_lot", "scene_description": "A parking lot with various vehicles and some pedestrians. The scene contains 7 objects: a building at middle-left (bbox: x=0.12, y=0.13, w=0.29, h=0.45); a person at middle-right (bbox: x=0.83, y=0.41, w=0.05, h=0.24); a bicycle at middle-left (bbox: x=0.17, y=0.42, w=0.07, h=0.06); a car at bottom-left (bbox: x=0.15, y=0.72, w=0.16, h=0.12); a building at top-left (bbox: x=0.13, y=0.07, w=0.20, h=0.35); a car at bottom-right (bbox: x=0.61, y=0.74, w=0.20, h=0.12); a bicycle at bottom-left (bbox: x=0.12, y=0.72, w=0.10, h=0.10).", "objects": [ { "id": 0, "class_label": "building", "position": "middle-left", "bbox": [ 0.1238, 0.1291, 0.2856, 0.445 ] }, { "id": 1, "class_label": "person", "position": "middle-right", "bbox": [ 0.8302, 0.4131, 0.0522, 0.237 ] }, { "id": 2, "class_label": "bicycle", "position": "middle-left", "bbox": [ 0.1672, 0.418, 0.0745, 0.0617 ] }, { "id": 3, "class_label": "car", "position": "bottom-left", "bbox": [ 0.1473, 0.7221, 0.1551, 0.1206 ] }, { "id": 4, "class_label": "building", "position": "top-left", "bbox": [ 0.1272, 0.0663, 0.203, 0.3508 ] }, { "id": 5, "class_label": "car", "position": "bottom-right", "bbox": [ 0.6101, 0.742, 0.1976, 0.1215 ] }, { "id": 6, "class_label": "bicycle", "position": "bottom-left", "bbox": [ 0.1206, 0.7237, 0.1027, 0.0957 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.1238, 0.1291, 0.2856, 0.445 ], "class_label": "building" }, { "id": 1, "bbox": [ 0.8302, 0.4131, 0.0522, 0.237 ], "class_label": "person" }, { "id": 2, "bbox": [ 0.1672, 0.418, 0.0745, 0.0617 ], "class_label": "bicycle" }, { "id": 3, "bbox": [ 0.1473, 0.7221, 0.1551, 0.1206 ], "class_label": "car" }, { "id": 4, "bbox": [ 0.1272, 0.0663, 0.203, 0.3508 ], "class_label": "building" }, { "id": 5, "bbox": [ 0.6101, 0.742, 0.1976, 0.1215 ], "class_label": "car" }, { "id": 6, "bbox": [ 0.1206, 0.7237, 0.1027, 0.0957 ], "class_label": "bicycle" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1040 }, { "scene_id": "fix_bboxes_sample_041", "scene_type": "residential_area", "scene_description": "A quiet residential neighborhood with houses, trees, and occasional pedestrians. The scene contains 5 objects: a car at bottom-left (bbox: x=0.09, y=0.65, w=0.10, h=0.13); a car at bottom-right (bbox: x=0.62, y=0.74, w=0.23, h=0.09); a building at bottom-center (bbox: x=0.34, y=0.64, w=0.27, h=0.33); a cat at top-right (bbox: x=0.79, y=0.11, w=0.08, h=0.06); a building at center (bbox: x=0.35, y=0.47, w=0.29, h=0.25).", "objects": [ { "id": 0, "class_label": "car", "position": "bottom-left", "bbox": [ 0.0907, 0.6485, 0.1014, 0.134 ] }, { "id": 1, "class_label": "car", "position": "bottom-right", "bbox": [ 0.6151, 0.7441, 0.231, 0.0949 ] }, { "id": 2, "class_label": "building", "position": "bottom-center", "bbox": [ 0.3383, 0.637, 0.2707, 0.3311 ] }, { "id": 3, "class_label": "cat", "position": "top-right", "bbox": [ 0.7914, 0.1104, 0.0782, 0.0571 ] }, { "id": 4, "class_label": "building", "position": "center", "bbox": [ 0.349, 0.4704, 0.2873, 0.2543 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.0907, 0.6485, 0.1014, 0.134 ], "class_label": "car" }, { "id": 1, "bbox": [ 0.6151, 0.7441, 0.231, 0.0949 ], "class_label": "car" }, { "id": 2, "bbox": [ 0.3383, 0.637, 0.2707, 0.3311 ], "class_label": "building" }, { "id": 3, "bbox": [ 0.7914, 0.1104, 0.0782, 0.0571 ], "class_label": "cat" }, { "id": 4, "bbox": [ 0.349, 0.4704, 0.2873, 0.2543 ], "class_label": "building" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1041 }, { "scene_id": "fix_bboxes_sample_042", "scene_type": "park", "scene_description": "A peaceful park setting with trees, benches, and people walking their pets. The scene contains 5 objects: a cat at top-left (bbox: x=0.16, y=0.27, w=0.08, h=0.07); a dog at bottom-center (bbox: x=0.36, y=0.77, w=0.07, h=0.05); a tree at center (bbox: x=0.39, y=0.40, w=0.08, h=0.16); a bicycle at bottom-center (bbox: x=0.37, y=0.67, w=0.09, h=0.11); a person at middle-right (bbox: x=0.77, y=0.31, w=0.08, h=0.18).", "objects": [ { "id": 0, "class_label": "cat", "position": "top-left", "bbox": [ 0.1592, 0.2661, 0.0774, 0.0665 ] }, { "id": 1, "class_label": "dog", "position": "bottom-center", "bbox": [ 0.3647, 0.7749, 0.0737, 0.0527 ] }, { "id": 2, "class_label": "tree", "position": "center", "bbox": [ 0.386, 0.3963, 0.0816, 0.163 ] }, { "id": 3, "class_label": "bicycle", "position": "bottom-center", "bbox": [ 0.3655, 0.668, 0.091, 0.1058 ] }, { "id": 4, "class_label": "person", "position": "middle-right", "bbox": [ 0.7668, 0.3115, 0.079, 0.1814 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.1592, 0.2661, 0.0774, 0.0665 ], "class_label": "cat" }, { "id": 1, "bbox": [ 0.3647, 0.7749, 0.0737, 0.0527 ], "class_label": "dog" }, { "id": 2, "bbox": [ 0.386, 0.3963, 0.0816, 0.163 ], "class_label": "tree" }, { "id": 3, "bbox": [ 0.3655, 0.668, 0.091, 0.1058 ], "class_label": "bicycle" }, { "id": 4, "bbox": [ 0.7668, 0.3115, 0.079, 0.1814 ], "class_label": "person" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1042 }, { "scene_id": "fix_bboxes_sample_043", "scene_type": "intersection", "scene_description": "A road intersection with traffic lights, vehicles, and crossing pedestrians. The scene contains 8 objects: a person at bottom-center (bbox: x=0.34, y=0.77, w=0.05, h=0.13); a person at top-right (bbox: x=0.81, y=0.09, w=0.07, h=0.13); a car at bottom-left (bbox: x=0.06, y=0.77, w=0.16, h=0.10); a person at center (bbox: x=0.50, y=0.29, w=0.08, h=0.22); a person at top-left (bbox: x=0.26, y=0.04, w=0.06, h=0.24); a car at middle-left (bbox: x=0.20, y=0.48, w=0.20, h=0.08); a person at bottom-right (bbox: x=0.86, y=0.79, w=0.05, h=0.19); a person at top-center (bbox: x=0.60, y=0.22, w=0.05, h=0.11).", "objects": [ { "id": 0, "class_label": "person", "position": "bottom-center", "bbox": [ 0.3383, 0.7662, 0.0496, 0.1334 ] }, { "id": 1, "class_label": "person", "position": "top-right", "bbox": [ 0.8124, 0.0914, 0.0705, 0.1347 ] }, { "id": 2, "class_label": "car", "position": "bottom-left", "bbox": [ 0.0644, 0.7725, 0.1641, 0.1023 ] }, { "id": 3, "class_label": "person", "position": "center", "bbox": [ 0.4992, 0.2873, 0.0758, 0.2173 ] }, { "id": 4, "class_label": "person", "position": "top-left", "bbox": [ 0.2622, 0.0447, 0.0632, 0.2358 ] }, { "id": 5, "class_label": "car", "position": "middle-left", "bbox": [ 0.2004, 0.4833, 0.1965, 0.0845 ] }, { "id": 6, "class_label": "person", "position": "bottom-right", "bbox": [ 0.8573, 0.7946, 0.0486, 0.192 ] }, { "id": 7, "class_label": "person", "position": "top-center", "bbox": [ 0.6029, 0.2246, 0.0548, 0.1134 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.3383, 0.7662, 0.0496, 0.1334 ], "class_label": "person" }, { "id": 1, "bbox": [ 0.8124, 0.0914, 0.0705, 0.1347 ], "class_label": "person" }, { "id": 2, "bbox": [ 0.0644, 0.7725, 0.1641, 0.1023 ], "class_label": "car" }, { "id": 3, "bbox": [ 0.4992, 0.2873, 0.0758, 0.2173 ], "class_label": "person" }, { "id": 4, "bbox": [ 0.2622, 0.0447, 0.0632, 0.2358 ], "class_label": "person" }, { "id": 5, "bbox": [ 0.2004, 0.4833, 0.1965, 0.0845 ], "class_label": "car" }, { "id": 6, "bbox": [ 0.8573, 0.7946, 0.0486, 0.192 ], "class_label": "person" }, { "id": 7, "bbox": [ 0.6029, 0.2246, 0.0548, 0.1134 ], "class_label": "person" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1043 }, { "scene_id": "fix_bboxes_sample_044", "scene_type": "park", "scene_description": "A peaceful park setting with trees, benches, and people walking their pets. The scene contains 6 objects: a tree at bottom-center (bbox: x=0.29, y=0.61, w=0.14, h=0.27); a dog at top-right (bbox: x=0.67, y=0.25, w=0.08, h=0.06); a cat at top-left (bbox: x=0.20, y=0.10, w=0.06, h=0.05); a bench at center (bbox: x=0.34, y=0.39, w=0.13, h=0.06); a bicycle at middle-left (bbox: x=0.05, y=0.36, w=0.12, h=0.11); a bicycle at middle-right (bbox: x=0.77, y=0.43, w=0.07, h=0.10).", "objects": [ { "id": 0, "class_label": "tree", "position": "bottom-center", "bbox": [ 0.2938, 0.6063, 0.1357, 0.2716 ] }, { "id": 1, "class_label": "dog", "position": "top-right", "bbox": [ 0.6661, 0.2486, 0.0776, 0.0597 ] }, { "id": 2, "class_label": "cat", "position": "top-left", "bbox": [ 0.2047, 0.101, 0.0612, 0.0486 ] }, { "id": 3, "class_label": "bench", "position": "center", "bbox": [ 0.3436, 0.3904, 0.1314, 0.0558 ] }, { "id": 4, "class_label": "bicycle", "position": "middle-left", "bbox": [ 0.0508, 0.3578, 0.1187, 0.1122 ] }, { "id": 5, "class_label": "bicycle", "position": "middle-right", "bbox": [ 0.7673, 0.4268, 0.0738, 0.0987 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.2938, 0.6063, 0.1357, 0.2716 ], "class_label": "tree" }, { "id": 1, "bbox": [ 0.6661, 0.2486, 0.0776, 0.0597 ], "class_label": "dog" }, { "id": 2, "bbox": [ 0.2047, 0.101, 0.0612, 0.0486 ], "class_label": "cat" }, { "id": 3, "bbox": [ 0.3436, 0.3904, 0.1314, 0.0558 ], "class_label": "bench" }, { "id": 4, "bbox": [ 0.0508, 0.3578, 0.1187, 0.1122 ], "class_label": "bicycle" }, { "id": 5, "bbox": [ 0.7673, 0.4268, 0.0738, 0.0987 ], "class_label": "bicycle" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1044 }, { "scene_id": "fix_bboxes_sample_045", "scene_type": "urban_street", "scene_description": "A busy urban street scene with vehicles, pedestrians, and city infrastructure. The scene contains 10 objects: a tree at bottom-center (bbox: x=0.43, y=0.61, w=0.11, h=0.30); a bench at bottom-center (bbox: x=0.46, y=0.77, w=0.13, h=0.07); a bicycle at top-left (bbox: x=0.07, y=0.11, w=0.09, h=0.08); a bicycle at top-center (bbox: x=0.54, y=0.10, w=0.09, h=0.08); a tree at middle-left (bbox: x=0.23, y=0.36, w=0.12, h=0.18); a bicycle at center (bbox: x=0.39, y=0.49, w=0.08, h=0.10); a bench at top-left (bbox: x=0.13, y=0.25, w=0.10, h=0.06); a traffic_light at middle-right (bbox: x=0.82, y=0.50, w=0.03, h=0.08); a bench at center (bbox: x=0.48, y=0.36, w=0.11, h=0.05); a truck at middle-left (bbox: x=0.00, y=0.42, w=0.26, h=0.13).", "objects": [ { "id": 0, "class_label": "tree", "position": "bottom-center", "bbox": [ 0.4328, 0.6145, 0.1133, 0.3014 ] }, { "id": 1, "class_label": "bench", "position": "bottom-center", "bbox": [ 0.4622, 0.7677, 0.1273, 0.0658 ] }, { "id": 2, "class_label": "bicycle", "position": "top-left", "bbox": [ 0.0724, 0.1102, 0.0944, 0.0795 ] }, { "id": 3, "class_label": "bicycle", "position": "top-center", "bbox": [ 0.5354, 0.0989, 0.0882, 0.0757 ] }, { "id": 4, "class_label": "tree", "position": "middle-left", "bbox": [ 0.2333, 0.3614, 0.1219, 0.1809 ] }, { "id": 5, "class_label": "bicycle", "position": "center", "bbox": [ 0.3946, 0.4908, 0.0831, 0.1013 ] }, { "id": 6, "class_label": "bench", "position": "top-left", "bbox": [ 0.1289, 0.2492, 0.103, 0.064 ] }, { "id": 7, "class_label": "traffic_light", "position": "middle-right", "bbox": [ 0.8234, 0.4954, 0.0252, 0.0806 ] }, { "id": 8, "class_label": "bench", "position": "center", "bbox": [ 0.4816, 0.3593, 0.1127, 0.0523 ] }, { "id": 9, "class_label": "truck", "position": "middle-left", "bbox": [ 0.0, 0.4244, 0.2573, 0.1277 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.4328, 0.6145, 0.1133, 0.3014 ], "class_label": "tree" }, { "id": 1, "bbox": [ 0.4622, 0.7677, 0.1273, 0.0658 ], "class_label": "bench" }, { "id": 2, "bbox": [ 0.0724, 0.1102, 0.0944, 0.0795 ], "class_label": "bicycle" }, { "id": 3, "bbox": [ 0.5354, 0.0989, 0.0882, 0.0757 ], "class_label": "bicycle" }, { "id": 4, "bbox": [ 0.2333, 0.3614, 0.1219, 0.1809 ], "class_label": "tree" }, { "id": 5, "bbox": [ 0.3946, 0.4908, 0.0831, 0.1013 ], "class_label": "bicycle" }, { "id": 6, "bbox": [ 0.1289, 0.2492, 0.103, 0.064 ], "class_label": "bench" }, { "id": 7, "bbox": [ 0.8234, 0.4954, 0.0252, 0.0806 ], "class_label": "traffic_light" }, { "id": 8, "bbox": [ 0.4816, 0.3593, 0.1127, 0.0523 ], "class_label": "bench" }, { "id": 9, "bbox": [ 0.0, 0.4244, 0.2573, 0.1277 ], "class_label": "truck" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1045 }, { "scene_id": "fix_bboxes_sample_046", "scene_type": "residential_area", "scene_description": "A quiet residential neighborhood with houses, trees, and occasional pedestrians. The scene contains 8 objects: a bench at bottom-center (bbox: x=0.36, y=0.84, w=0.09, h=0.06); a bench at top-left (bbox: x=0.07, y=0.08, w=0.14, h=0.07); a person at middle-left (bbox: x=0.18, y=0.38, w=0.07, h=0.18); a car at middle-left (bbox: x=0.14, y=0.31, w=0.13, h=0.13); a building at top-center (bbox: x=0.28, y=0.13, w=0.24, h=0.24); a dog at top-center (bbox: x=0.49, y=0.14, w=0.10, h=0.05); a car at bottom-right (bbox: x=0.61, y=0.70, w=0.19, h=0.15); a car at middle-left (bbox: x=0.07, y=0.52, w=0.15, h=0.12).", "objects": [ { "id": 0, "class_label": "bench", "position": "bottom-center", "bbox": [ 0.3633, 0.8432, 0.0862, 0.0609 ] }, { "id": 1, "class_label": "bench", "position": "top-left", "bbox": [ 0.0654, 0.0801, 0.1405, 0.0695 ] }, { "id": 2, "class_label": "person", "position": "middle-left", "bbox": [ 0.1812, 0.3838, 0.0697, 0.1799 ] }, { "id": 3, "class_label": "car", "position": "middle-left", "bbox": [ 0.1431, 0.3145, 0.1332, 0.1322 ] }, { "id": 4, "class_label": "building", "position": "top-center", "bbox": [ 0.2785, 0.1345, 0.242, 0.2357 ] }, { "id": 5, "class_label": "dog", "position": "top-center", "bbox": [ 0.4948, 0.1373, 0.0965, 0.0459 ] }, { "id": 6, "class_label": "car", "position": "bottom-right", "bbox": [ 0.6138, 0.7, 0.1902, 0.1462 ] }, { "id": 7, "class_label": "car", "position": "middle-left", "bbox": [ 0.0745, 0.5209, 0.1529, 0.1192 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.3633, 0.8432, 0.0862, 0.0609 ], "class_label": "bench" }, { "id": 1, "bbox": [ 0.0654, 0.0801, 0.1405, 0.0695 ], "class_label": "bench" }, { "id": 2, "bbox": [ 0.1812, 0.3838, 0.0697, 0.1799 ], "class_label": "person" }, { "id": 3, "bbox": [ 0.1431, 0.3145, 0.1332, 0.1322 ], "class_label": "car" }, { "id": 4, "bbox": [ 0.2785, 0.1345, 0.242, 0.2357 ], "class_label": "building" }, { "id": 5, "bbox": [ 0.4948, 0.1373, 0.0965, 0.0459 ], "class_label": "dog" }, { "id": 6, "bbox": [ 0.6138, 0.7, 0.1902, 0.1462 ], "class_label": "car" }, { "id": 7, "bbox": [ 0.0745, 0.5209, 0.1529, 0.1192 ], "class_label": "car" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1046 }, { "scene_id": "fix_bboxes_sample_047", "scene_type": "urban_street", "scene_description": "A busy urban street scene with vehicles, pedestrians, and city infrastructure. The scene contains 7 objects: a tree at center (bbox: x=0.40, y=0.30, w=0.11, h=0.24); a truck at top-center (bbox: x=0.47, y=0.09, w=0.17, h=0.17); a traffic_light at top-center (bbox: x=0.47, y=0.20, w=0.03, h=0.06); a traffic_light at bottom-left (bbox: x=0.12, y=0.77, w=0.03, h=0.08); a bench at top-right (bbox: x=0.75, y=0.18, w=0.12, h=0.05); a tree at top-left (bbox: x=0.10, y=0.01, w=0.13, h=0.35); a traffic_light at center (bbox: x=0.39, y=0.54, w=0.03, h=0.07).", "objects": [ { "id": 0, "class_label": "tree", "position": "center", "bbox": [ 0.4029, 0.3027, 0.1124, 0.2439 ] }, { "id": 1, "class_label": "truck", "position": "top-center", "bbox": [ 0.47, 0.088, 0.1684, 0.1714 ] }, { "id": 2, "class_label": "traffic_light", "position": "top-center", "bbox": [ 0.4748, 0.201, 0.0266, 0.0644 ] }, { "id": 3, "class_label": "traffic_light", "position": "bottom-left", "bbox": [ 0.1249, 0.7745, 0.0256, 0.0772 ] }, { "id": 4, "class_label": "bench", "position": "top-right", "bbox": [ 0.7488, 0.1808, 0.1229, 0.0503 ] }, { "id": 5, "class_label": "tree", "position": "top-left", "bbox": [ 0.0974, 0.0131, 0.132, 0.3477 ] }, { "id": 6, "class_label": "traffic_light", "position": "center", "bbox": [ 0.3876, 0.5448, 0.0331, 0.0703 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.4029, 0.3027, 0.1124, 0.2439 ], "class_label": "tree" }, { "id": 1, "bbox": [ 0.47, 0.088, 0.1684, 0.1714 ], "class_label": "truck" }, { "id": 2, "bbox": [ 0.4748, 0.201, 0.0266, 0.0644 ], "class_label": "traffic_light" }, { "id": 3, "bbox": [ 0.1249, 0.7745, 0.0256, 0.0772 ], "class_label": "traffic_light" }, { "id": 4, "bbox": [ 0.7488, 0.1808, 0.1229, 0.0503 ], "class_label": "bench" }, { "id": 5, "bbox": [ 0.0974, 0.0131, 0.132, 0.3477 ], "class_label": "tree" }, { "id": 6, "bbox": [ 0.3876, 0.5448, 0.0331, 0.0703 ], "class_label": "traffic_light" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1047 }, { "scene_id": "fix_bboxes_sample_048", "scene_type": "parking_lot", "scene_description": "A parking lot with various vehicles and some pedestrians. The scene contains 9 objects: a truck at middle-left (bbox: x=0.01, y=0.39, w=0.30, h=0.15); a car at bottom-center (bbox: x=0.27, y=0.68, w=0.16, h=0.09); a truck at bottom-center (bbox: x=0.43, y=0.66, w=0.29, h=0.12); a car at bottom-center (bbox: x=0.28, y=0.69, w=0.25, h=0.11); a car at middle-right (bbox: x=0.60, y=0.50, w=0.24, h=0.15); a person at top-center (bbox: x=0.57, y=0.19, w=0.04, h=0.14); a bicycle at bottom-center (bbox: x=0.37, y=0.75, w=0.09, h=0.08); a car at bottom-center (bbox: x=0.49, y=0.70, w=0.21, h=0.14); a bicycle at bottom-center (bbox: x=0.34, y=0.82, w=0.11, h=0.06).", "objects": [ { "id": 0, "class_label": "truck", "position": "middle-left", "bbox": [ 0.0114, 0.3913, 0.2964, 0.1527 ] }, { "id": 1, "class_label": "car", "position": "bottom-center", "bbox": [ 0.274, 0.6824, 0.1601, 0.0866 ] }, { "id": 2, "class_label": "truck", "position": "bottom-center", "bbox": [ 0.4317, 0.6607, 0.2875, 0.1169 ] }, { "id": 3, "class_label": "car", "position": "bottom-center", "bbox": [ 0.2815, 0.6944, 0.2488, 0.1068 ] }, { "id": 4, "class_label": "car", "position": "middle-right", "bbox": [ 0.5977, 0.4969, 0.245, 0.1472 ] }, { "id": 5, "class_label": "person", "position": "top-center", "bbox": [ 0.569, 0.1918, 0.0416, 0.1365 ] }, { "id": 6, "class_label": "bicycle", "position": "bottom-center", "bbox": [ 0.3671, 0.7471, 0.0929, 0.0825 ] }, { "id": 7, "class_label": "car", "position": "bottom-center", "bbox": [ 0.4911, 0.7023, 0.2149, 0.1417 ] }, { "id": 8, "class_label": "bicycle", "position": "bottom-center", "bbox": [ 0.3365, 0.8246, 0.1131, 0.0641 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.0114, 0.3913, 0.2964, 0.1527 ], "class_label": "truck" }, { "id": 1, "bbox": [ 0.274, 0.6824, 0.1601, 0.0866 ], "class_label": "car" }, { "id": 2, "bbox": [ 0.4317, 0.6607, 0.2875, 0.1169 ], "class_label": "truck" }, { "id": 3, "bbox": [ 0.2815, 0.6944, 0.2488, 0.1068 ], "class_label": "car" }, { "id": 4, "bbox": [ 0.5977, 0.4969, 0.245, 0.1472 ], "class_label": "car" }, { "id": 5, "bbox": [ 0.569, 0.1918, 0.0416, 0.1365 ], "class_label": "person" }, { "id": 6, "bbox": [ 0.3671, 0.7471, 0.0929, 0.0825 ], "class_label": "bicycle" }, { "id": 7, "bbox": [ 0.4911, 0.7023, 0.2149, 0.1417 ], "class_label": "car" }, { "id": 8, "bbox": [ 0.3365, 0.8246, 0.1131, 0.0641 ], "class_label": "bicycle" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1048 }, { "scene_id": "fix_bboxes_sample_049", "scene_type": "intersection", "scene_description": "A road intersection with traffic lights, vehicles, and crossing pedestrians. The scene contains 9 objects: a car at middle-left (bbox: x=0.07, y=0.33, w=0.17, h=0.14); a car at top-center (bbox: x=0.35, y=0.09, w=0.12, h=0.10); a person at bottom-left (bbox: x=0.18, y=0.77, w=0.07, h=0.19); a car at middle-right (bbox: x=0.71, y=0.34, w=0.18, h=0.09); a person at middle-right (bbox: x=0.87, y=0.49, w=0.05, h=0.16); a building at top-left (bbox: x=0.08, y=0.16, w=0.17, h=0.23); a building at center (bbox: x=0.47, y=0.21, w=0.34, h=0.29); a truck at bottom-right (bbox: x=0.64, y=0.65, w=0.20, h=0.12); a person at top-center (bbox: x=0.62, y=0.09, w=0.07, h=0.10).", "objects": [ { "id": 0, "class_label": "car", "position": "middle-left", "bbox": [ 0.0662, 0.3322, 0.1717, 0.143 ] }, { "id": 1, "class_label": "car", "position": "top-center", "bbox": [ 0.3499, 0.091, 0.1171, 0.1015 ] }, { "id": 2, "class_label": "person", "position": "bottom-left", "bbox": [ 0.1824, 0.7735, 0.065, 0.1875 ] }, { "id": 3, "class_label": "car", "position": "middle-right", "bbox": [ 0.7074, 0.3393, 0.18, 0.0875 ] }, { "id": 4, "class_label": "person", "position": "middle-right", "bbox": [ 0.8676, 0.4865, 0.054, 0.1612 ] }, { "id": 5, "class_label": "building", "position": "top-left", "bbox": [ 0.0793, 0.1644, 0.1737, 0.2331 ] }, { "id": 6, "class_label": "building", "position": "center", "bbox": [ 0.4713, 0.2133, 0.3397, 0.2863 ] }, { "id": 7, "class_label": "truck", "position": "bottom-right", "bbox": [ 0.6376, 0.6455, 0.1966, 0.1198 ] }, { "id": 8, "class_label": "person", "position": "top-center", "bbox": [ 0.6156, 0.0939, 0.0675, 0.1042 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.0662, 0.3322, 0.1717, 0.143 ], "class_label": "car" }, { "id": 1, "bbox": [ 0.3499, 0.091, 0.1171, 0.1015 ], "class_label": "car" }, { "id": 2, "bbox": [ 0.1824, 0.7735, 0.065, 0.1875 ], "class_label": "person" }, { "id": 3, "bbox": [ 0.7074, 0.3393, 0.18, 0.0875 ], "class_label": "car" }, { "id": 4, "bbox": [ 0.8676, 0.4865, 0.054, 0.1612 ], "class_label": "person" }, { "id": 5, "bbox": [ 0.0793, 0.1644, 0.1737, 0.2331 ], "class_label": "building" }, { "id": 6, "bbox": [ 0.4713, 0.2133, 0.3397, 0.2863 ], "class_label": "building" }, { "id": 7, "bbox": [ 0.6376, 0.6455, 0.1966, 0.1198 ], "class_label": "truck" }, { "id": 8, "bbox": [ 0.6156, 0.0939, 0.0675, 0.1042 ], "class_label": "person" } ], "task_id": "fix_bboxes", "difficulty": "easy", "seed": 1049 } ]