[ { "batch_id": 0, "scenes": [ { "scene_id": "batch_audit_batch00_scene00", "scene_type": "park", "scene_description": "A peaceful park setting with trees, benches, and people walking their pets. The scene contains 6 objects: a person at center (bbox: x=0.50, y=0.49, w=0.07, h=0.15); a cat at middle-left (bbox: x=0.22, y=0.53, w=0.05, h=0.06); a bicycle at bottom-right (bbox: x=0.75, y=0.71, w=0.08, h=0.10); a bicycle at bottom-right (bbox: x=0.81, y=0.79, w=0.08, h=0.08); a tree at middle-left (bbox: x=0.18, y=0.27, w=0.13, h=0.35); a dog at top-right (bbox: x=0.76, y=0.15, w=0.09, h=0.07).", "objects": [ { "id": 0, "class_label": "person", "position": "center", "bbox": [ 0.4966, 0.4948, 0.0687, 0.1461 ] }, { "id": 1, "class_label": "cat", "position": "middle-left", "bbox": [ 0.2177, 0.5304, 0.0474, 0.0604 ] }, { "id": 2, "class_label": "bicycle", "position": "bottom-right", "bbox": [ 0.7455, 0.7082, 0.0788, 0.0972 ] }, { "id": 3, "class_label": "bicycle", "position": "bottom-right", "bbox": [ 0.8098, 0.7903, 0.0819, 0.083 ] }, { "id": 4, "class_label": "tree", "position": "middle-left", "bbox": [ 0.1769, 0.2729, 0.1284, 0.3485 ] }, { "id": 5, "class_label": "dog", "position": "top-right", "bbox": [ 0.7647, 0.15, 0.0935, 0.075 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.4966, 0.4948, 0.0687, 0.1461 ], "class_label": "person" }, { "id": 1, "bbox": [ 0.2177, 0.5304, 0.0474, 0.0604 ], "class_label": "cat" }, { "id": 2, "bbox": [ 0.7455, 0.7082, 0.0788, 0.0972 ], "class_label": "bicycle" }, { "id": 3, "bbox": [ 0.8098, 0.7903, 0.0819, 0.083 ], "class_label": "bicycle" }, { "id": 4, "bbox": [ 0.1769, 0.2729, 0.1284, 0.3485 ], "class_label": "tree" }, { "id": 5, "bbox": [ 0.7647, 0.15, 0.0935, 0.075 ], "class_label": "dog" } ], "batch_id": 0, "task_id": "batch_audit", "difficulty": "hard", "seed": 3000 }, { "scene_id": "batch_audit_batch00_scene01", "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 bottom-right (bbox: x=0.68, y=0.76, w=0.22, h=0.14); a bicycle at center (bbox: x=0.32, y=0.35, w=0.12, h=0.12); a tree at bottom-right (bbox: x=0.77, y=0.67, w=0.09, h=0.33); a person at middle-right (bbox: x=0.79, y=0.41, w=0.05, h=0.22); a tree at top-center (bbox: x=0.35, y=0.02, w=0.13, h=0.27); a car at top-right (bbox: x=0.78, y=0.08, w=0.14, h=0.11); a car at top-left (bbox: x=0.18, y=0.14, w=0.18, h=0.09); a building at middle-right (bbox: x=0.67, y=0.37, w=0.29, h=0.23); a tree at bottom-center (bbox: x=0.32, y=0.69, w=0.13, h=0.27); a truck at bottom-left (bbox: x=0.14, y=0.73, w=0.25, h=0.11).", "objects": [ { "id": 0, "class_label": "truck", "position": "bottom-right", "bbox": [ 0.6831, 0.7606, 0.2227, 0.1374 ] }, { "id": 1, "class_label": "bicycle", "position": "center", "bbox": [ 0.3223, 0.3507, 0.1188, 0.1191 ] }, { "id": 2, "class_label": "tree", "position": "bottom-right", "bbox": [ 0.7673, 0.6741, 0.0924, 0.3259 ] }, { "id": 3, "class_label": "person", "position": "middle-right", "bbox": [ 0.7861, 0.4101, 0.048, 0.2193 ] }, { "id": 4, "class_label": "tree", "position": "top-center", "bbox": [ 0.3476, 0.0234, 0.1274, 0.2692 ] }, { "id": 5, "class_label": "car", "position": "top-right", "bbox": [ 0.7778, 0.0751, 0.1435, 0.109 ] }, { "id": 6, "class_label": "car", "position": "top-left", "bbox": [ 0.1845, 0.1401, 0.1778, 0.0907 ] }, { "id": 7, "class_label": "building", "position": "middle-right", "bbox": [ 0.6665, 0.3714, 0.2868, 0.2291 ] }, { "id": 8, "class_label": "tree", "position": "bottom-center", "bbox": [ 0.321, 0.685, 0.1313, 0.2679 ] }, { "id": 9, "class_label": "truck", "position": "bottom-left", "bbox": [ 0.1443, 0.73, 0.2532, 0.1126 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.6831, 0.7606, 0.2227, 0.1374 ], "class_label": "truck" }, { "id": 1, "bbox": [ 0.3223, 0.3507, 0.1188, 0.1191 ], "class_label": "bicycle" }, { "id": 2, "bbox": [ 0.7673, 0.6741, 0.0924, 0.3259 ], "class_label": "tree" }, { "id": 3, "bbox": [ 0.7861, 0.4101, 0.048, 0.2193 ], "class_label": "person" }, { "id": 4, "bbox": [ 0.3476, 0.0234, 0.1274, 0.2692 ], "class_label": "tree" }, { "id": 5, "bbox": [ 0.7778, 0.0751, 0.1435, 0.109 ], "class_label": "car" }, { "id": 6, "bbox": [ 0.1845, 0.1401, 0.1778, 0.0907 ], "class_label": "car" }, { "id": 7, "bbox": [ 0.6665, 0.3714, 0.2868, 0.2291 ], "class_label": "building" }, { "id": 8, "bbox": [ 0.321, 0.685, 0.1313, 0.2679 ], "class_label": "tree" }, { "id": 9, "bbox": [ 0.1443, 0.73, 0.2532, 0.1126 ], "class_label": "truck" } ], "batch_id": 0, "task_id": "batch_audit", "difficulty": "hard", "seed": 3001 }, { "scene_id": "batch_audit_batch00_scene02", "scene_type": "urban_street", "scene_description": "A busy urban street scene with vehicles, pedestrians, and city infrastructure. The scene contains 9 objects: a tree at top-right (bbox: x=0.82, y=0.13, w=0.14, h=0.17); a bicycle at top-center (bbox: x=0.39, y=0.06, w=0.08, h=0.09); a person at center (bbox: x=0.43, y=0.31, w=0.05, h=0.19); a person at middle-left (bbox: x=0.12, y=0.51, w=0.07, h=0.19); a car at middle-right (bbox: x=0.68, y=0.50, w=0.24, h=0.11); a person at top-left (bbox: x=0.11, y=0.10, w=0.06, h=0.15); a building at bottom-right (bbox: x=0.78, y=0.65, w=0.21, h=0.35); a bicycle at top-right (bbox: x=0.79, y=0.13, w=0.10, h=0.08); a bicycle at top-left (bbox: x=0.18, y=0.09, w=0.11, h=0.08).", "objects": [ { "id": 0, "class_label": "tree", "position": "top-right", "bbox": [ 0.8151, 0.1337, 0.1414, 0.1688 ] }, { "id": 1, "class_label": "bicycle", "position": "top-center", "bbox": [ 0.388, 0.064, 0.0805, 0.0883 ] }, { "id": 2, "class_label": "person", "position": "center", "bbox": [ 0.4322, 0.3097, 0.0549, 0.1949 ] }, { "id": 3, "class_label": "person", "position": "middle-left", "bbox": [ 0.119, 0.5149, 0.0679, 0.1945 ] }, { "id": 4, "class_label": "car", "position": "middle-right", "bbox": [ 0.6787, 0.499, 0.2392, 0.11 ] }, { "id": 5, "class_label": "person", "position": "top-left", "bbox": [ 0.1109, 0.0988, 0.0625, 0.1495 ] }, { "id": 6, "class_label": "building", "position": "bottom-right", "bbox": [ 0.7812, 0.6511, 0.2113, 0.3489 ] }, { "id": 7, "class_label": "bicycle", "position": "top-right", "bbox": [ 0.787, 0.1268, 0.0994, 0.0801 ] }, { "id": 8, "class_label": "bicycle", "position": "top-left", "bbox": [ 0.1752, 0.0851, 0.1114, 0.0756 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.8151, 0.1337, 0.1414, 0.1688 ], "class_label": "tree" }, { "id": 1, "bbox": [ 0.388, 0.064, 0.0805, 0.0883 ], "class_label": "bicycle" }, { "id": 2, "bbox": [ 0.4322, 0.3097, 0.0549, 0.1949 ], "class_label": "person" }, { "id": 3, "bbox": [ 0.119, 0.5149, 0.0679, 0.1945 ], "class_label": "person" }, { "id": 4, "bbox": [ 0.6787, 0.499, 0.2392, 0.11 ], "class_label": "car" }, { "id": 5, "bbox": [ 0.1109, 0.0988, 0.0625, 0.1495 ], "class_label": "person" }, { "id": 6, "bbox": [ 0.7812, 0.6511, 0.2113, 0.3489 ], "class_label": "building" }, { "id": 7, "bbox": [ 0.787, 0.1268, 0.0994, 0.0801 ], "class_label": "bicycle" }, { "id": 8, "bbox": [ 0.1752, 0.0851, 0.1114, 0.0756 ], "class_label": "bicycle" } ], "batch_id": 0, "task_id": "batch_audit", "difficulty": "hard", "seed": 3002 }, { "scene_id": "batch_audit_batch00_scene03", "scene_type": "parking_lot", "scene_description": "A parking lot with various vehicles and some pedestrians. The scene contains 10 objects: a person at center (bbox: x=0.53, y=0.34, w=0.06, h=0.17); a truck at top-left (bbox: x=0.06, y=0.17, w=0.24, h=0.14); a bicycle at bottom-center (bbox: x=0.50, y=0.82, w=0.10, h=0.11); a car at middle-right (bbox: x=0.71, y=0.49, w=0.16, h=0.10); a building at center (bbox: x=0.45, y=0.31, w=0.21, h=0.27); a building at top-right (bbox: x=0.71, y=0.00, w=0.19, h=0.38); a truck at middle-right (bbox: x=0.62, y=0.49, w=0.18, h=0.12); a truck at middle-right (bbox: x=0.61, y=0.28, w=0.29, h=0.16); a bicycle at bottom-right (bbox: x=0.84, y=0.83, w=0.09, h=0.09); a truck at top-right (bbox: x=0.57, y=0.11, w=0.28, h=0.17).", "objects": [ { "id": 0, "class_label": "person", "position": "center", "bbox": [ 0.53, 0.3387, 0.0557, 0.173 ] }, { "id": 1, "class_label": "truck", "position": "top-left", "bbox": [ 0.0589, 0.1705, 0.2447, 0.1431 ] }, { "id": 2, "class_label": "bicycle", "position": "bottom-center", "bbox": [ 0.4994, 0.8235, 0.1043, 0.1059 ] }, { "id": 3, "class_label": "car", "position": "middle-right", "bbox": [ 0.707, 0.493, 0.1615, 0.102 ] }, { "id": 4, "class_label": "building", "position": "center", "bbox": [ 0.4467, 0.3086, 0.2111, 0.2652 ] }, { "id": 5, "class_label": "building", "position": "top-right", "bbox": [ 0.7105, 0.0, 0.1915, 0.3785 ] }, { "id": 6, "class_label": "truck", "position": "middle-right", "bbox": [ 0.6171, 0.4916, 0.1774, 0.1203 ] }, { "id": 7, "class_label": "truck", "position": "middle-right", "bbox": [ 0.6085, 0.2805, 0.2896, 0.1613 ] }, { "id": 8, "class_label": "bicycle", "position": "bottom-right", "bbox": [ 0.8437, 0.8275, 0.0866, 0.0922 ] }, { "id": 9, "class_label": "truck", "position": "top-right", "bbox": [ 0.5672, 0.1103, 0.2767, 0.1744 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.53, 0.3387, 0.0557, 0.173 ], "class_label": "person" }, { "id": 1, "bbox": [ 0.0589, 0.1705, 0.2447, 0.1431 ], "class_label": "truck" }, { "id": 2, "bbox": [ 0.4994, 0.8235, 0.1043, 0.1059 ], "class_label": "bicycle" }, { "id": 3, "bbox": [ 0.707, 0.493, 0.1615, 0.102 ], "class_label": "car" }, { "id": 4, "bbox": [ 0.4467, 0.3086, 0.2111, 0.2652 ], "class_label": "building" }, { "id": 5, "bbox": [ 0.7105, 0.0, 0.1915, 0.3785 ], "class_label": "building" }, { "id": 6, "bbox": [ 0.6171, 0.4916, 0.1774, 0.1203 ], "class_label": "truck" }, { "id": 7, "bbox": [ 0.6085, 0.2805, 0.2896, 0.1613 ], "class_label": "truck" }, { "id": 8, "bbox": [ 0.8437, 0.8275, 0.0866, 0.0922 ], "class_label": "bicycle" }, { "id": 9, "bbox": [ 0.5672, 0.1103, 0.2767, 0.1744 ], "class_label": "truck" } ], "batch_id": 0, "task_id": "batch_audit", "difficulty": "hard", "seed": 3003 }, { "scene_id": "batch_audit_batch00_scene04", "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-left (bbox: x=0.18, y=0.21, w=0.08, h=0.11); a person at center (bbox: x=0.49, y=0.40, w=0.04, h=0.23); a tree at bottom-right (bbox: x=0.73, y=0.60, w=0.15, h=0.22); a dog at top-right (bbox: x=0.67, y=0.08, w=0.09, h=0.08); a bench at top-left (bbox: x=0.07, y=0.26, w=0.15, h=0.07).", "objects": [ { "id": 0, "class_label": "bicycle", "position": "top-left", "bbox": [ 0.1759, 0.2072, 0.0753, 0.1081 ] }, { "id": 1, "class_label": "person", "position": "center", "bbox": [ 0.4904, 0.3971, 0.0419, 0.2332 ] }, { "id": 2, "class_label": "tree", "position": "bottom-right", "bbox": [ 0.7303, 0.5983, 0.1499, 0.2246 ] }, { "id": 3, "class_label": "dog", "position": "top-right", "bbox": [ 0.6715, 0.076, 0.0917, 0.0789 ] }, { "id": 4, "class_label": "bench", "position": "top-left", "bbox": [ 0.0736, 0.2599, 0.1459, 0.0719 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.1759, 0.2072, 0.0753, 0.1081 ], "class_label": "bicycle" }, { "id": 1, "bbox": [ 0.4904, 0.3971, 0.0419, 0.2332 ], "class_label": "person" }, { "id": 2, "bbox": [ 0.7303, 0.5983, 0.1499, 0.2246 ], "class_label": "tree" }, { "id": 3, "bbox": [ 0.6715, 0.076, 0.0917, 0.0789 ], "class_label": "dog" }, { "id": 4, "bbox": [ 0.0736, 0.2599, 0.1459, 0.0719 ], "class_label": "bench" } ], "batch_id": 0, "task_id": "batch_audit", "difficulty": "hard", "seed": 3004 } ] }, { "batch_id": 1, "scenes": [ { "scene_id": "batch_audit_batch01_scene00", "scene_type": "residential_area", "scene_description": "A quiet residential neighborhood with houses, trees, and occasional pedestrians. The scene contains 9 objects: a bench at middle-left (bbox: x=0.24, y=0.51, w=0.09, h=0.07); a person at top-left (bbox: x=0.20, y=0.09, w=0.07, h=0.18); a tree at bottom-center (bbox: x=0.42, y=0.78, w=0.14, h=0.21); a cat at top-right (bbox: x=0.87, y=0.13, w=0.04, h=0.07); a building at top-left (bbox: x=0.02, y=0.11, w=0.19, h=0.36); a tree at bottom-center (bbox: x=0.46, y=0.55, w=0.12, h=0.30); a building at center (bbox: x=0.24, y=0.45, w=0.30, h=0.23); a dog at top-right (bbox: x=0.74, y=0.08, w=0.07, h=0.06); a tree at middle-right (bbox: x=0.78, y=0.47, w=0.12, h=0.30).", "objects": [ { "id": 0, "class_label": "bench", "position": "middle-left", "bbox": [ 0.2374, 0.5094, 0.0917, 0.065 ] }, { "id": 1, "class_label": "person", "position": "top-left", "bbox": [ 0.1992, 0.0885, 0.0737, 0.1819 ] }, { "id": 2, "class_label": "tree", "position": "bottom-center", "bbox": [ 0.4183, 0.7777, 0.1351, 0.2143 ] }, { "id": 3, "class_label": "cat", "position": "top-right", "bbox": [ 0.8749, 0.1299, 0.0405, 0.068 ] }, { "id": 4, "class_label": "building", "position": "top-left", "bbox": [ 0.0196, 0.1105, 0.1895, 0.3573 ] }, { "id": 5, "class_label": "tree", "position": "bottom-center", "bbox": [ 0.4584, 0.5521, 0.1234, 0.3045 ] }, { "id": 6, "class_label": "building", "position": "center", "bbox": [ 0.2426, 0.4451, 0.3042, 0.2271 ] }, { "id": 7, "class_label": "dog", "position": "top-right", "bbox": [ 0.743, 0.0797, 0.0704, 0.0649 ] }, { "id": 8, "class_label": "tree", "position": "middle-right", "bbox": [ 0.7755, 0.4661, 0.123, 0.299 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.2374, 0.5094, 0.0917, 0.065 ], "class_label": "bench" }, { "id": 1, "bbox": [ 0.1992, 0.0885, 0.0737, 0.1819 ], "class_label": "person" }, { "id": 2, "bbox": [ 0.4183, 0.7777, 0.1351, 0.2143 ], "class_label": "tree" }, { "id": 3, "bbox": [ 0.8749, 0.1299, 0.0405, 0.068 ], "class_label": "cat" }, { "id": 4, "bbox": [ 0.0196, 0.1105, 0.1895, 0.3573 ], "class_label": "building" }, { "id": 5, "bbox": [ 0.4584, 0.5521, 0.1234, 0.3045 ], "class_label": "tree" }, { "id": 6, "bbox": [ 0.2426, 0.4451, 0.3042, 0.2271 ], "class_label": "building" }, { "id": 7, "bbox": [ 0.743, 0.0797, 0.0704, 0.0649 ], "class_label": "dog" }, { "id": 8, "bbox": [ 0.7755, 0.4661, 0.123, 0.299 ], "class_label": "tree" } ], "batch_id": 1, "task_id": "batch_audit", "difficulty": "hard", "seed": 3100 }, { "scene_id": "batch_audit_batch01_scene01", "scene_type": "urban_street", "scene_description": "A busy urban street scene with vehicles, pedestrians, and city infrastructure. The scene contains 5 objects: a tree at top-center (bbox: x=0.31, y=0.06, w=0.12, h=0.19); a traffic_light at middle-left (bbox: x=0.14, y=0.38, w=0.04, h=0.09); a building at bottom-left (bbox: x=0.05, y=0.66, w=0.25, h=0.34); a person at middle-left (bbox: x=0.12, y=0.36, w=0.06, h=0.21); a traffic_light at top-right (bbox: x=0.73, y=0.14, w=0.04, h=0.07).", "objects": [ { "id": 0, "class_label": "tree", "position": "top-center", "bbox": [ 0.3148, 0.0612, 0.1243, 0.192 ] }, { "id": 1, "class_label": "traffic_light", "position": "middle-left", "bbox": [ 0.1442, 0.3794, 0.0393, 0.092 ] }, { "id": 2, "class_label": "building", "position": "bottom-left", "bbox": [ 0.0505, 0.6628, 0.2539, 0.3372 ] }, { "id": 3, "class_label": "person", "position": "middle-left", "bbox": [ 0.1185, 0.3613, 0.0567, 0.2071 ] }, { "id": 4, "class_label": "traffic_light", "position": "top-right", "bbox": [ 0.7313, 0.1402, 0.0362, 0.0747 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.3148, 0.0612, 0.1243, 0.192 ], "class_label": "tree" }, { "id": 1, "bbox": [ 0.1442, 0.3794, 0.0393, 0.092 ], "class_label": "traffic_light" }, { "id": 2, "bbox": [ 0.0505, 0.6628, 0.2539, 0.3372 ], "class_label": "building" }, { "id": 3, "bbox": [ 0.1185, 0.3613, 0.0567, 0.2071 ], "class_label": "person" }, { "id": 4, "bbox": [ 0.7313, 0.1402, 0.0362, 0.0747 ], "class_label": "traffic_light" } ], "batch_id": 1, "task_id": "batch_audit", "difficulty": "hard", "seed": 3101 }, { "scene_id": "batch_audit_batch01_scene02", "scene_type": "residential_area", "scene_description": "A quiet residential neighborhood with houses, trees, and occasional pedestrians. The scene contains 6 objects: a person at top-center (bbox: x=0.43, y=0.04, w=0.05, h=0.16); a building at middle-left (bbox: x=0.00, y=0.44, w=0.35, h=0.25); a building at top-left (bbox: x=0.00, y=0.09, w=0.28, h=0.28); a dog at center (bbox: x=0.49, y=0.49, w=0.09, h=0.06); a cat at top-center (bbox: x=0.39, y=0.10, w=0.06, h=0.05); a dog at bottom-right (bbox: x=0.74, y=0.85, w=0.06, h=0.06).", "objects": [ { "id": 0, "class_label": "person", "position": "top-center", "bbox": [ 0.4301, 0.0436, 0.0536, 0.1563 ] }, { "id": 1, "class_label": "building", "position": "middle-left", "bbox": [ 0.0, 0.4357, 0.3469, 0.2452 ] }, { "id": 2, "class_label": "building", "position": "top-left", "bbox": [ 0.0, 0.0898, 0.2843, 0.2782 ] }, { "id": 3, "class_label": "dog", "position": "center", "bbox": [ 0.4913, 0.4854, 0.0889, 0.0644 ] }, { "id": 4, "class_label": "cat", "position": "top-center", "bbox": [ 0.3895, 0.0977, 0.0611, 0.0517 ] }, { "id": 5, "class_label": "dog", "position": "bottom-right", "bbox": [ 0.7435, 0.8492, 0.0648, 0.0629 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.4301, 0.0436, 0.0536, 0.1563 ], "class_label": "person" }, { "id": 1, "bbox": [ 0.0, 0.4357, 0.3469, 0.2452 ], "class_label": "building" }, { "id": 2, "bbox": [ 0.0, 0.0898, 0.2843, 0.2782 ], "class_label": "building" }, { "id": 3, "bbox": [ 0.4913, 0.4854, 0.0889, 0.0644 ], "class_label": "dog" }, { "id": 4, "bbox": [ 0.3895, 0.0977, 0.0611, 0.0517 ], "class_label": "cat" }, { "id": 5, "bbox": [ 0.7435, 0.8492, 0.0648, 0.0629 ], "class_label": "dog" } ], "batch_id": 1, "task_id": "batch_audit", "difficulty": "hard", "seed": 3102 }, { "scene_id": "batch_audit_batch01_scene03", "scene_type": "parking_lot", "scene_description": "A parking lot with various vehicles and some pedestrians. The scene contains 11 objects: a building at bottom-center (bbox: x=0.31, y=0.67, w=0.17, h=0.29); a car at middle-right (bbox: x=0.80, y=0.46, w=0.18, h=0.08); a building at bottom-center (bbox: x=0.27, y=0.51, w=0.25, h=0.45); a car at top-left (bbox: x=0.18, y=0.15, w=0.19, h=0.13); a truck at bottom-left (bbox: x=0.14, y=0.76, w=0.16, h=0.18); a truck at top-left (bbox: x=0.10, y=0.10, w=0.24, h=0.15); a person at middle-left (bbox: x=0.16, y=0.31, w=0.07, h=0.14); a person at middle-left (bbox: x=0.10, y=0.26, w=0.08, h=0.25); a truck at middle-right (bbox: x=0.57, y=0.59, w=0.27, h=0.11); a bicycle at middle-right (bbox: x=0.81, y=0.55, w=0.10, h=0.07); a person at top-left (bbox: x=0.10, y=0.07, w=0.07, h=0.20).", "objects": [ { "id": 0, "class_label": "building", "position": "bottom-center", "bbox": [ 0.3075, 0.6676, 0.1674, 0.2865 ] }, { "id": 1, "class_label": "car", "position": "middle-right", "bbox": [ 0.7985, 0.4586, 0.1784, 0.0838 ] }, { "id": 2, "class_label": "building", "position": "bottom-center", "bbox": [ 0.2708, 0.5101, 0.2466, 0.4485 ] }, { "id": 3, "class_label": "car", "position": "top-left", "bbox": [ 0.1839, 0.1482, 0.186, 0.1314 ] }, { "id": 4, "class_label": "truck", "position": "bottom-left", "bbox": [ 0.1387, 0.7642, 0.1594, 0.1757 ] }, { "id": 5, "class_label": "truck", "position": "top-left", "bbox": [ 0.1001, 0.096, 0.2435, 0.1545 ] }, { "id": 6, "class_label": "person", "position": "middle-left", "bbox": [ 0.1627, 0.3081, 0.0669, 0.1396 ] }, { "id": 7, "class_label": "person", "position": "middle-left", "bbox": [ 0.096, 0.2594, 0.0799, 0.2466 ] }, { "id": 8, "class_label": "truck", "position": "middle-right", "bbox": [ 0.5675, 0.5894, 0.2674, 0.1136 ] }, { "id": 9, "class_label": "bicycle", "position": "middle-right", "bbox": [ 0.8146, 0.5498, 0.0965, 0.0731 ] }, { "id": 10, "class_label": "person", "position": "top-left", "bbox": [ 0.1038, 0.07, 0.0691, 0.2032 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.3075, 0.6676, 0.1674, 0.2865 ], "class_label": "building" }, { "id": 1, "bbox": [ 0.7985, 0.4586, 0.1784, 0.0838 ], "class_label": "car" }, { "id": 2, "bbox": [ 0.2708, 0.5101, 0.2466, 0.4485 ], "class_label": "building" }, { "id": 3, "bbox": [ 0.1839, 0.1482, 0.186, 0.1314 ], "class_label": "car" }, { "id": 4, "bbox": [ 0.1387, 0.7642, 0.1594, 0.1757 ], "class_label": "truck" }, { "id": 5, "bbox": [ 0.1001, 0.096, 0.2435, 0.1545 ], "class_label": "truck" }, { "id": 6, "bbox": [ 0.1627, 0.3081, 0.0669, 0.1396 ], "class_label": "person" }, { "id": 7, "bbox": [ 0.096, 0.2594, 0.0799, 0.2466 ], "class_label": "person" }, { "id": 8, "bbox": [ 0.5675, 0.5894, 0.2674, 0.1136 ], "class_label": "truck" }, { "id": 9, "bbox": [ 0.8146, 0.5498, 0.0965, 0.0731 ], "class_label": "bicycle" }, { "id": 10, "bbox": [ 0.1038, 0.07, 0.0691, 0.2032 ], "class_label": "person" } ], "batch_id": 1, "task_id": "batch_audit", "difficulty": "hard", "seed": 3103 }, { "scene_id": "batch_audit_batch01_scene04", "scene_type": "park", "scene_description": "A peaceful park setting with trees, benches, and people walking their pets. The scene contains 7 objects: a dog at middle-right (bbox: x=0.85, y=0.47, w=0.10, h=0.06); a tree at top-left (bbox: x=0.21, y=0.04, w=0.15, h=0.28); a dog at bottom-center (bbox: x=0.32, y=0.81, w=0.09, h=0.05); a person at middle-left (bbox: x=0.22, y=0.54, w=0.05, h=0.19); a cat at top-right (bbox: x=0.73, y=0.24, w=0.07, h=0.07); a cat at middle-left (bbox: x=0.12, y=0.45, w=0.05, h=0.05); a dog at bottom-center (bbox: x=0.57, y=0.80, w=0.06, h=0.07).", "objects": [ { "id": 0, "class_label": "dog", "position": "middle-right", "bbox": [ 0.8467, 0.4652, 0.0961, 0.061 ] }, { "id": 1, "class_label": "tree", "position": "top-left", "bbox": [ 0.2104, 0.0408, 0.1493, 0.2794 ] }, { "id": 2, "class_label": "dog", "position": "bottom-center", "bbox": [ 0.3174, 0.8055, 0.0869, 0.0511 ] }, { "id": 3, "class_label": "person", "position": "middle-left", "bbox": [ 0.2151, 0.5436, 0.0534, 0.1858 ] }, { "id": 4, "class_label": "cat", "position": "top-right", "bbox": [ 0.726, 0.2375, 0.0653, 0.07 ] }, { "id": 5, "class_label": "cat", "position": "middle-left", "bbox": [ 0.1178, 0.4474, 0.0543, 0.0507 ] }, { "id": 6, "class_label": "dog", "position": "bottom-center", "bbox": [ 0.5673, 0.7989, 0.0558, 0.0671 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.8467, 0.4652, 0.0961, 0.061 ], "class_label": "dog" }, { "id": 1, "bbox": [ 0.2104, 0.0408, 0.1493, 0.2794 ], "class_label": "tree" }, { "id": 2, "bbox": [ 0.3174, 0.8055, 0.0869, 0.0511 ], "class_label": "dog" }, { "id": 3, "bbox": [ 0.2151, 0.5436, 0.0534, 0.1858 ], "class_label": "person" }, { "id": 4, "bbox": [ 0.726, 0.2375, 0.0653, 0.07 ], "class_label": "cat" }, { "id": 5, "bbox": [ 0.1178, 0.4474, 0.0543, 0.0507 ], "class_label": "cat" }, { "id": 6, "bbox": [ 0.5673, 0.7989, 0.0558, 0.0671 ], "class_label": "dog" } ], "batch_id": 1, "task_id": "batch_audit", "difficulty": "hard", "seed": 3104 } ] }, { "batch_id": 2, "scenes": [ { "scene_id": "batch_audit_batch02_scene00", "scene_type": "intersection", "scene_description": "A road intersection with traffic lights, vehicles, and crossing pedestrians. The scene contains 11 objects: a bicycle at top-right (bbox: x=0.70, y=0.20, w=0.09, h=0.10); a bicycle at middle-left (bbox: x=0.24, y=0.42, w=0.09, h=0.07); a traffic_light at middle-left (bbox: x=0.09, y=0.57, w=0.03, h=0.08); a building at top-center (bbox: x=0.28, y=0.00, w=0.24, h=0.30); a person at middle-right (bbox: x=0.77, y=0.49, w=0.07, h=0.13); a building at top-center (bbox: x=0.30, y=0.00, w=0.24, h=0.37); a bicycle at middle-left (bbox: x=0.12, y=0.40, w=0.07, h=0.09); a person at middle-left (bbox: x=0.09, y=0.42, w=0.05, h=0.23); a person at center (bbox: x=0.53, y=0.30, w=0.07, h=0.21); a truck at middle-right (bbox: x=0.72, y=0.50, w=0.28, h=0.18); a truck at center (bbox: x=0.42, y=0.56, w=0.29, h=0.14).", "objects": [ { "id": 0, "class_label": "bicycle", "position": "top-right", "bbox": [ 0.6999, 0.1982, 0.0898, 0.0956 ] }, { "id": 1, "class_label": "bicycle", "position": "middle-left", "bbox": [ 0.2426, 0.4207, 0.0925, 0.0734 ] }, { "id": 2, "class_label": "traffic_light", "position": "middle-left", "bbox": [ 0.0894, 0.5744, 0.0317, 0.0824 ] }, { "id": 3, "class_label": "building", "position": "top-center", "bbox": [ 0.278, 0.0, 0.2429, 0.2964 ] }, { "id": 4, "class_label": "person", "position": "middle-right", "bbox": [ 0.7725, 0.495, 0.0672, 0.1293 ] }, { "id": 5, "class_label": "building", "position": "top-center", "bbox": [ 0.2991, 0.0, 0.2409, 0.3727 ] }, { "id": 6, "class_label": "bicycle", "position": "middle-left", "bbox": [ 0.1242, 0.3961, 0.0715, 0.0942 ] }, { "id": 7, "class_label": "person", "position": "middle-left", "bbox": [ 0.09, 0.4237, 0.0485, 0.2313 ] }, { "id": 8, "class_label": "person", "position": "center", "bbox": [ 0.5251, 0.2974, 0.0712, 0.2122 ] }, { "id": 9, "class_label": "truck", "position": "middle-right", "bbox": [ 0.7162, 0.5043, 0.2838, 0.1773 ] }, { "id": 10, "class_label": "truck", "position": "center", "bbox": [ 0.4213, 0.5561, 0.2922, 0.1385 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.6999, 0.1982, 0.0898, 0.0956 ], "class_label": "bicycle" }, { "id": 1, "bbox": [ 0.2426, 0.4207, 0.0925, 0.0734 ], "class_label": "bicycle" }, { "id": 2, "bbox": [ 0.0894, 0.5744, 0.0317, 0.0824 ], "class_label": "traffic_light" }, { "id": 3, "bbox": [ 0.278, 0.0, 0.2429, 0.2964 ], "class_label": "building" }, { "id": 4, "bbox": [ 0.7725, 0.495, 0.0672, 0.1293 ], "class_label": "person" }, { "id": 5, "bbox": [ 0.2991, 0.0, 0.2409, 0.3727 ], "class_label": "building" }, { "id": 6, "bbox": [ 0.1242, 0.3961, 0.0715, 0.0942 ], "class_label": "bicycle" }, { "id": 7, "bbox": [ 0.09, 0.4237, 0.0485, 0.2313 ], "class_label": "person" }, { "id": 8, "bbox": [ 0.5251, 0.2974, 0.0712, 0.2122 ], "class_label": "person" }, { "id": 9, "bbox": [ 0.7162, 0.5043, 0.2838, 0.1773 ], "class_label": "truck" }, { "id": 10, "bbox": [ 0.4213, 0.5561, 0.2922, 0.1385 ], "class_label": "truck" } ], "batch_id": 2, "task_id": "batch_audit", "difficulty": "hard", "seed": 3200 }, { "scene_id": "batch_audit_batch02_scene01", "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 middle-left (bbox: x=0.25, y=0.54, w=0.06, h=0.12); a tree at top-right (bbox: x=0.68, y=0.15, w=0.13, h=0.18); a cat at middle-left (bbox: x=0.11, y=0.46, w=0.08, h=0.06); a bench at middle-left (bbox: x=0.10, y=0.46, w=0.09, h=0.08); a cat at top-left (bbox: x=0.18, y=0.17, w=0.05, h=0.04).", "objects": [ { "id": 0, "class_label": "bicycle", "position": "middle-left", "bbox": [ 0.2475, 0.5446, 0.0615, 0.1161 ] }, { "id": 1, "class_label": "tree", "position": "top-right", "bbox": [ 0.6807, 0.1522, 0.1263, 0.1843 ] }, { "id": 2, "class_label": "cat", "position": "middle-left", "bbox": [ 0.1066, 0.4613, 0.0787, 0.0603 ] }, { "id": 3, "class_label": "bench", "position": "middle-left", "bbox": [ 0.1003, 0.4646, 0.0908, 0.0788 ] }, { "id": 4, "class_label": "cat", "position": "top-left", "bbox": [ 0.1795, 0.1701, 0.0504, 0.0418 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.2475, 0.5446, 0.0615, 0.1161 ], "class_label": "bicycle" }, { "id": 1, "bbox": [ 0.6807, 0.1522, 0.1263, 0.1843 ], "class_label": "tree" }, { "id": 2, "bbox": [ 0.1066, 0.4613, 0.0787, 0.0603 ], "class_label": "cat" }, { "id": 3, "bbox": [ 0.1003, 0.4646, 0.0908, 0.0788 ], "class_label": "bench" }, { "id": 4, "bbox": [ 0.1795, 0.1701, 0.0504, 0.0418 ], "class_label": "cat" } ], "batch_id": 2, "task_id": "batch_audit", "difficulty": "hard", "seed": 3201 }, { "scene_id": "batch_audit_batch02_scene02", "scene_type": "intersection", "scene_description": "A road intersection with traffic lights, vehicles, and crossing pedestrians. The scene contains 8 objects: a traffic_light at center (bbox: x=0.57, y=0.40, w=0.02, h=0.07); a traffic_light at bottom-left (bbox: x=0.15, y=0.82, w=0.03, h=0.06); a truck at bottom-left (bbox: x=0.12, y=0.71, w=0.21, h=0.12); a person at bottom-left (bbox: x=0.13, y=0.81, w=0.07, h=0.15); a person at top-right (bbox: x=0.76, y=0.13, w=0.07, h=0.23); a building at top-left (bbox: x=0.02, y=0.00, w=0.26, h=0.37); a traffic_light at bottom-center (bbox: x=0.39, y=0.73, w=0.02, h=0.09); a traffic_light at top-left (bbox: x=0.23, y=0.20, w=0.03, h=0.06).", "objects": [ { "id": 0, "class_label": "traffic_light", "position": "center", "bbox": [ 0.5749, 0.3951, 0.0213, 0.0678 ] }, { "id": 1, "class_label": "traffic_light", "position": "bottom-left", "bbox": [ 0.1461, 0.8235, 0.0283, 0.0634 ] }, { "id": 2, "class_label": "truck", "position": "bottom-left", "bbox": [ 0.1234, 0.7136, 0.212, 0.1211 ] }, { "id": 3, "class_label": "person", "position": "bottom-left", "bbox": [ 0.1339, 0.8076, 0.0723, 0.1488 ] }, { "id": 4, "class_label": "person", "position": "top-right", "bbox": [ 0.7596, 0.132, 0.0742, 0.231 ] }, { "id": 5, "class_label": "building", "position": "top-left", "bbox": [ 0.0228, 0.0, 0.2592, 0.3672 ] }, { "id": 6, "class_label": "traffic_light", "position": "bottom-center", "bbox": [ 0.392, 0.7253, 0.0205, 0.0887 ] }, { "id": 7, "class_label": "traffic_light", "position": "top-left", "bbox": [ 0.2313, 0.2003, 0.0268, 0.062 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.5749, 0.3951, 0.0213, 0.0678 ], "class_label": "traffic_light" }, { "id": 1, "bbox": [ 0.1461, 0.8235, 0.0283, 0.0634 ], "class_label": "traffic_light" }, { "id": 2, "bbox": [ 0.1234, 0.7136, 0.212, 0.1211 ], "class_label": "truck" }, { "id": 3, "bbox": [ 0.1339, 0.8076, 0.0723, 0.1488 ], "class_label": "person" }, { "id": 4, "bbox": [ 0.7596, 0.132, 0.0742, 0.231 ], "class_label": "person" }, { "id": 5, "bbox": [ 0.0228, 0.0, 0.2592, 0.3672 ], "class_label": "building" }, { "id": 6, "bbox": [ 0.392, 0.7253, 0.0205, 0.0887 ], "class_label": "traffic_light" }, { "id": 7, "bbox": [ 0.2313, 0.2003, 0.0268, 0.062 ], "class_label": "traffic_light" } ], "batch_id": 2, "task_id": "batch_audit", "difficulty": "hard", "seed": 3202 }, { "scene_id": "batch_audit_batch02_scene03", "scene_type": "parking_lot", "scene_description": "A parking lot with various vehicles and some pedestrians. The scene contains 10 objects: a person at bottom-right (bbox: x=0.83, y=0.70, w=0.04, h=0.18); a bicycle at top-center (bbox: x=0.46, y=0.10, w=0.07, h=0.07); a car at middle-right (bbox: x=0.64, y=0.59, w=0.16, h=0.10); a car at top-left (bbox: x=0.14, y=0.20, w=0.21, h=0.12); a building at middle-left (bbox: x=0.05, y=0.21, w=0.25, h=0.45); a truck at bottom-right (bbox: x=0.77, y=0.79, w=0.21, h=0.14); a truck at middle-right (bbox: x=0.74, y=0.50, w=0.17, h=0.10); a car at bottom-center (bbox: x=0.32, y=0.71, w=0.24, h=0.09); a car at bottom-center (bbox: x=0.27, y=0.77, w=0.17, h=0.08); a car at bottom-left (bbox: x=0.15, y=0.69, w=0.22, h=0.12).", "objects": [ { "id": 0, "class_label": "person", "position": "bottom-right", "bbox": [ 0.8276, 0.6998, 0.0408, 0.1792 ] }, { "id": 1, "class_label": "bicycle", "position": "top-center", "bbox": [ 0.4618, 0.1023, 0.0707, 0.0696 ] }, { "id": 2, "class_label": "car", "position": "middle-right", "bbox": [ 0.6388, 0.5857, 0.1572, 0.0979 ] }, { "id": 3, "class_label": "car", "position": "top-left", "bbox": [ 0.1445, 0.1982, 0.2097, 0.125 ] }, { "id": 4, "class_label": "building", "position": "middle-left", "bbox": [ 0.0549, 0.209, 0.2486, 0.4451 ] }, { "id": 5, "class_label": "truck", "position": "bottom-right", "bbox": [ 0.7663, 0.7855, 0.21, 0.1442 ] }, { "id": 6, "class_label": "truck", "position": "middle-right", "bbox": [ 0.7412, 0.5032, 0.1683, 0.1016 ] }, { "id": 7, "class_label": "car", "position": "bottom-center", "bbox": [ 0.3203, 0.7127, 0.2373, 0.0932 ] }, { "id": 8, "class_label": "car", "position": "bottom-center", "bbox": [ 0.2704, 0.7745, 0.1677, 0.0812 ] }, { "id": 9, "class_label": "car", "position": "bottom-left", "bbox": [ 0.1461, 0.6923, 0.2153, 0.1156 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.8276, 0.6998, 0.0408, 0.1792 ], "class_label": "person" }, { "id": 1, "bbox": [ 0.4618, 0.1023, 0.0707, 0.0696 ], "class_label": "bicycle" }, { "id": 2, "bbox": [ 0.6388, 0.5857, 0.1572, 0.0979 ], "class_label": "car" }, { "id": 3, "bbox": [ 0.1445, 0.1982, 0.2097, 0.125 ], "class_label": "car" }, { "id": 4, "bbox": [ 0.0549, 0.209, 0.2486, 0.4451 ], "class_label": "building" }, { "id": 5, "bbox": [ 0.7663, 0.7855, 0.21, 0.1442 ], "class_label": "truck" }, { "id": 6, "bbox": [ 0.7412, 0.5032, 0.1683, 0.1016 ], "class_label": "truck" }, { "id": 7, "bbox": [ 0.3203, 0.7127, 0.2373, 0.0932 ], "class_label": "car" }, { "id": 8, "bbox": [ 0.2704, 0.7745, 0.1677, 0.0812 ], "class_label": "car" }, { "id": 9, "bbox": [ 0.1461, 0.6923, 0.2153, 0.1156 ], "class_label": "car" } ], "batch_id": 2, "task_id": "batch_audit", "difficulty": "hard", "seed": 3203 }, { "scene_id": "batch_audit_batch02_scene04", "scene_type": "park", "scene_description": "A peaceful park setting with trees, benches, and people walking their pets. The scene contains 7 objects: a cat at middle-left (bbox: x=0.10, y=0.33, w=0.06, h=0.06); a tree at top-center (bbox: x=0.38, y=0.13, w=0.10, h=0.32); a bicycle at middle-right (bbox: x=0.66, y=0.39, w=0.09, h=0.10); a dog at top-right (bbox: x=0.70, y=0.09, w=0.09, h=0.07); a person at top-right (bbox: x=0.86, y=0.07, w=0.06, h=0.22); a tree at top-right (bbox: x=0.64, y=0.07, w=0.14, h=0.21); a bench at bottom-left (bbox: x=0.11, y=0.80, w=0.10, h=0.07).", "objects": [ { "id": 0, "class_label": "cat", "position": "middle-left", "bbox": [ 0.0985, 0.3276, 0.0605, 0.0611 ] }, { "id": 1, "class_label": "tree", "position": "top-center", "bbox": [ 0.3807, 0.1255, 0.1013, 0.3199 ] }, { "id": 2, "class_label": "bicycle", "position": "middle-right", "bbox": [ 0.6614, 0.3866, 0.09, 0.0964 ] }, { "id": 3, "class_label": "dog", "position": "top-right", "bbox": [ 0.7006, 0.086, 0.0935, 0.0713 ] }, { "id": 4, "class_label": "person", "position": "top-right", "bbox": [ 0.8597, 0.0736, 0.0605, 0.2171 ] }, { "id": 5, "class_label": "tree", "position": "top-right", "bbox": [ 0.6365, 0.0678, 0.1363, 0.2089 ] }, { "id": 6, "class_label": "bench", "position": "bottom-left", "bbox": [ 0.1115, 0.7971, 0.0955, 0.0726 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.0985, 0.3276, 0.0605, 0.0611 ], "class_label": "cat" }, { "id": 1, "bbox": [ 0.3807, 0.1255, 0.1013, 0.3199 ], "class_label": "tree" }, { "id": 2, "bbox": [ 0.6614, 0.3866, 0.09, 0.0964 ], "class_label": "bicycle" }, { "id": 3, "bbox": [ 0.7006, 0.086, 0.0935, 0.0713 ], "class_label": "dog" }, { "id": 4, "bbox": [ 0.8597, 0.0736, 0.0605, 0.2171 ], "class_label": "person" }, { "id": 5, "bbox": [ 0.6365, 0.0678, 0.1363, 0.2089 ], "class_label": "tree" }, { "id": 6, "bbox": [ 0.1115, 0.7971, 0.0955, 0.0726 ], "class_label": "bench" } ], "batch_id": 2, "task_id": "batch_audit", "difficulty": "hard", "seed": 3204 } ] }, { "batch_id": 3, "scenes": [ { "scene_id": "batch_audit_batch03_scene00", "scene_type": "residential_area", "scene_description": "A quiet residential neighborhood with houses, trees, and occasional pedestrians. The scene contains 9 objects: a cat at middle-left (bbox: x=0.21, y=0.37, w=0.08, h=0.05); a cat at middle-right (bbox: x=0.70, y=0.57, w=0.05, h=0.06); a bench at bottom-left (bbox: x=0.09, y=0.68, w=0.12, h=0.07); a tree at middle-left (bbox: x=0.09, y=0.38, w=0.11, h=0.25); a tree at top-left (bbox: x=0.14, y=0.07, w=0.10, h=0.18); a person at top-center (bbox: x=0.56, y=0.14, w=0.07, h=0.25); a building at middle-left (bbox: x=0.00, y=0.26, w=0.32, h=0.25); a bench at top-center (bbox: x=0.52, y=0.19, w=0.10, h=0.05); a building at top-left (bbox: x=0.05, y=0.06, w=0.23, h=0.38).", "objects": [ { "id": 0, "class_label": "cat", "position": "middle-left", "bbox": [ 0.2093, 0.3737, 0.0797, 0.0455 ] }, { "id": 1, "class_label": "cat", "position": "middle-right", "bbox": [ 0.7013, 0.5696, 0.0528, 0.0641 ] }, { "id": 2, "class_label": "bench", "position": "bottom-left", "bbox": [ 0.0876, 0.6789, 0.1154, 0.0689 ] }, { "id": 3, "class_label": "tree", "position": "middle-left", "bbox": [ 0.0937, 0.3787, 0.113, 0.2477 ] }, { "id": 4, "class_label": "tree", "position": "top-left", "bbox": [ 0.1389, 0.0694, 0.0982, 0.1849 ] }, { "id": 5, "class_label": "person", "position": "top-center", "bbox": [ 0.5628, 0.1447, 0.0664, 0.2482 ] }, { "id": 6, "class_label": "building", "position": "middle-left", "bbox": [ 0.0, 0.2597, 0.3174, 0.2507 ] }, { "id": 7, "class_label": "bench", "position": "top-center", "bbox": [ 0.5195, 0.188, 0.1046, 0.0528 ] }, { "id": 8, "class_label": "building", "position": "top-left", "bbox": [ 0.0486, 0.0605, 0.228, 0.3759 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.2093, 0.3737, 0.0797, 0.0455 ], "class_label": "cat" }, { "id": 1, "bbox": [ 0.7013, 0.5696, 0.0528, 0.0641 ], "class_label": "cat" }, { "id": 2, "bbox": [ 0.0876, 0.6789, 0.1154, 0.0689 ], "class_label": "bench" }, { "id": 3, "bbox": [ 0.0937, 0.3787, 0.113, 0.2477 ], "class_label": "tree" }, { "id": 4, "bbox": [ 0.1389, 0.0694, 0.0982, 0.1849 ], "class_label": "tree" }, { "id": 5, "bbox": [ 0.5628, 0.1447, 0.0664, 0.2482 ], "class_label": "person" }, { "id": 6, "bbox": [ 0.0, 0.2597, 0.3174, 0.2507 ], "class_label": "building" }, { "id": 7, "bbox": [ 0.5195, 0.188, 0.1046, 0.0528 ], "class_label": "bench" }, { "id": 8, "bbox": [ 0.0486, 0.0605, 0.228, 0.3759 ], "class_label": "building" } ], "batch_id": 3, "task_id": "batch_audit", "difficulty": "hard", "seed": 3300 }, { "scene_id": "batch_audit_batch03_scene01", "scene_type": "intersection", "scene_description": "A road intersection with traffic lights, vehicles, and crossing pedestrians. The scene contains 6 objects: a person at bottom-left (bbox: x=0.16, y=0.74, w=0.06, h=0.25); a truck at center (bbox: x=0.31, y=0.46, w=0.16, h=0.15); a bicycle at top-right (bbox: x=0.72, y=0.05, w=0.12, h=0.10); a bicycle at top-right (bbox: x=0.76, y=0.25, w=0.06, h=0.06); a building at bottom-left (bbox: x=0.03, y=0.58, w=0.16, h=0.36); a truck at bottom-center (bbox: x=0.31, y=0.74, w=0.29, h=0.12).", "objects": [ { "id": 0, "class_label": "person", "position": "bottom-left", "bbox": [ 0.1554, 0.736, 0.0633, 0.2451 ] }, { "id": 1, "class_label": "truck", "position": "center", "bbox": [ 0.3101, 0.4574, 0.1614, 0.155 ] }, { "id": 2, "class_label": "bicycle", "position": "top-right", "bbox": [ 0.7174, 0.0519, 0.1168, 0.1007 ] }, { "id": 3, "class_label": "bicycle", "position": "top-right", "bbox": [ 0.7619, 0.249, 0.0607, 0.0631 ] }, { "id": 4, "class_label": "building", "position": "bottom-left", "bbox": [ 0.033, 0.5761, 0.1645, 0.3634 ] }, { "id": 5, "class_label": "truck", "position": "bottom-center", "bbox": [ 0.3111, 0.7387, 0.2945, 0.1169 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.1554, 0.736, 0.0633, 0.2451 ], "class_label": "person" }, { "id": 1, "bbox": [ 0.3101, 0.4574, 0.1614, 0.155 ], "class_label": "truck" }, { "id": 2, "bbox": [ 0.7174, 0.0519, 0.1168, 0.1007 ], "class_label": "bicycle" }, { "id": 3, "bbox": [ 0.7619, 0.249, 0.0607, 0.0631 ], "class_label": "bicycle" }, { "id": 4, "bbox": [ 0.033, 0.5761, 0.1645, 0.3634 ], "class_label": "building" }, { "id": 5, "bbox": [ 0.3111, 0.7387, 0.2945, 0.1169 ], "class_label": "truck" } ], "batch_id": 3, "task_id": "batch_audit", "difficulty": "hard", "seed": 3301 }, { "scene_id": "batch_audit_batch03_scene02", "scene_type": "parking_lot", "scene_description": "A parking lot with various vehicles and some pedestrians. The scene contains 10 objects: a building at bottom-center (bbox: x=0.23, y=0.71, w=0.30, h=0.24); a building at bottom-center (bbox: x=0.55, y=0.62, w=0.17, h=0.38); a building at bottom-right (bbox: x=0.64, y=0.57, w=0.30, h=0.43); a truck at top-center (bbox: x=0.50, y=0.16, w=0.19, h=0.17); a car at top-right (bbox: x=0.61, y=0.13, w=0.23, h=0.08); a car at middle-right (bbox: x=0.62, y=0.33, w=0.18, h=0.10); a truck at middle-right (bbox: x=0.74, y=0.54, w=0.16, h=0.12); a truck at bottom-left (bbox: x=0.05, y=0.64, w=0.18, h=0.17); a car at bottom-right (bbox: x=0.80, y=0.69, w=0.11, h=0.09); a truck at center (bbox: x=0.49, y=0.48, w=0.17, h=0.12).", "objects": [ { "id": 0, "class_label": "building", "position": "bottom-center", "bbox": [ 0.2347, 0.7112, 0.3048, 0.2401 ] }, { "id": 1, "class_label": "building", "position": "bottom-center", "bbox": [ 0.5536, 0.6209, 0.1695, 0.3791 ] }, { "id": 2, "class_label": "building", "position": "bottom-right", "bbox": [ 0.6362, 0.5655, 0.303, 0.4345 ] }, { "id": 3, "class_label": "truck", "position": "top-center", "bbox": [ 0.4994, 0.1608, 0.1921, 0.1723 ] }, { "id": 4, "class_label": "car", "position": "top-right", "bbox": [ 0.606, 0.1251, 0.2265, 0.0824 ] }, { "id": 5, "class_label": "car", "position": "middle-right", "bbox": [ 0.6209, 0.3308, 0.1831, 0.1044 ] }, { "id": 6, "class_label": "truck", "position": "middle-right", "bbox": [ 0.7366, 0.5437, 0.1562, 0.1165 ] }, { "id": 7, "class_label": "truck", "position": "bottom-left", "bbox": [ 0.0481, 0.6435, 0.1805, 0.167 ] }, { "id": 8, "class_label": "car", "position": "bottom-right", "bbox": [ 0.7963, 0.6928, 0.1103, 0.094 ] }, { "id": 9, "class_label": "truck", "position": "center", "bbox": [ 0.4855, 0.481, 0.1707, 0.1169 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.2347, 0.7112, 0.3048, 0.2401 ], "class_label": "building" }, { "id": 1, "bbox": [ 0.5536, 0.6209, 0.1695, 0.3791 ], "class_label": "building" }, { "id": 2, "bbox": [ 0.6362, 0.5655, 0.303, 0.4345 ], "class_label": "building" }, { "id": 3, "bbox": [ 0.4994, 0.1608, 0.1921, 0.1723 ], "class_label": "truck" }, { "id": 4, "bbox": [ 0.606, 0.1251, 0.2265, 0.0824 ], "class_label": "car" }, { "id": 5, "bbox": [ 0.6209, 0.3308, 0.1831, 0.1044 ], "class_label": "car" }, { "id": 6, "bbox": [ 0.7366, 0.5437, 0.1562, 0.1165 ], "class_label": "truck" }, { "id": 7, "bbox": [ 0.0481, 0.6435, 0.1805, 0.167 ], "class_label": "truck" }, { "id": 8, "bbox": [ 0.7963, 0.6928, 0.1103, 0.094 ], "class_label": "car" }, { "id": 9, "bbox": [ 0.4855, 0.481, 0.1707, 0.1169 ], "class_label": "truck" } ], "batch_id": 3, "task_id": "batch_audit", "difficulty": "hard", "seed": 3302 }, { "scene_id": "batch_audit_batch03_scene03", "scene_type": "residential_area", "scene_description": "A quiet residential neighborhood with houses, trees, and occasional pedestrians. The scene contains 8 objects: a car at top-center (bbox: x=0.24, y=0.22, w=0.23, h=0.11); a building at top-left (bbox: x=0.00, y=0.03, w=0.33, h=0.45); a tree at bottom-center (bbox: x=0.48, y=0.74, w=0.10, h=0.26); a dog at center (bbox: x=0.57, y=0.61, w=0.09, h=0.06); a tree at middle-right (bbox: x=0.65, y=0.24, w=0.13, h=0.26); a person at bottom-left (bbox: x=0.20, y=0.72, w=0.06, h=0.17); a building at middle-left (bbox: x=0.06, y=0.37, w=0.21, h=0.31); a bench at top-left (bbox: x=0.17, y=0.24, w=0.11, h=0.06).", "objects": [ { "id": 0, "class_label": "car", "position": "top-center", "bbox": [ 0.2372, 0.2239, 0.2292, 0.1143 ] }, { "id": 1, "class_label": "building", "position": "top-left", "bbox": [ 0.0, 0.0321, 0.3259, 0.4472 ] }, { "id": 2, "class_label": "tree", "position": "bottom-center", "bbox": [ 0.484, 0.7396, 0.0953, 0.2604 ] }, { "id": 3, "class_label": "dog", "position": "center", "bbox": [ 0.5696, 0.613, 0.0932, 0.0611 ] }, { "id": 4, "class_label": "tree", "position": "middle-right", "bbox": [ 0.6468, 0.2422, 0.1304, 0.2572 ] }, { "id": 5, "class_label": "person", "position": "bottom-left", "bbox": [ 0.1984, 0.721, 0.0574, 0.1665 ] }, { "id": 6, "class_label": "building", "position": "middle-left", "bbox": [ 0.0625, 0.3667, 0.2148, 0.3139 ] }, { "id": 7, "class_label": "bench", "position": "top-left", "bbox": [ 0.1717, 0.2419, 0.1102, 0.0627 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.2372, 0.2239, 0.2292, 0.1143 ], "class_label": "car" }, { "id": 1, "bbox": [ 0.0, 0.0321, 0.3259, 0.4472 ], "class_label": "building" }, { "id": 2, "bbox": [ 0.484, 0.7396, 0.0953, 0.2604 ], "class_label": "tree" }, { "id": 3, "bbox": [ 0.5696, 0.613, 0.0932, 0.0611 ], "class_label": "dog" }, { "id": 4, "bbox": [ 0.6468, 0.2422, 0.1304, 0.2572 ], "class_label": "tree" }, { "id": 5, "bbox": [ 0.1984, 0.721, 0.0574, 0.1665 ], "class_label": "person" }, { "id": 6, "bbox": [ 0.0625, 0.3667, 0.2148, 0.3139 ], "class_label": "building" }, { "id": 7, "bbox": [ 0.1717, 0.2419, 0.1102, 0.0627 ], "class_label": "bench" } ], "batch_id": 3, "task_id": "batch_audit", "difficulty": "hard", "seed": 3303 }, { "scene_id": "batch_audit_batch03_scene04", "scene_type": "residential_area", "scene_description": "A quiet residential neighborhood with houses, trees, and occasional pedestrians. The scene contains 4 objects: a car at top-left (bbox: x=0.19, y=0.11, w=0.20, h=0.14); a dog at middle-left (bbox: x=0.11, y=0.58, w=0.10, h=0.06); a building at top-right (bbox: x=0.75, y=0.02, w=0.19, h=0.43); a car at middle-left (bbox: x=0.07, y=0.33, w=0.23, h=0.14).", "objects": [ { "id": 0, "class_label": "car", "position": "top-left", "bbox": [ 0.1866, 0.1127, 0.1956, 0.1383 ] }, { "id": 1, "class_label": "dog", "position": "middle-left", "bbox": [ 0.1072, 0.5843, 0.0987, 0.0606 ] }, { "id": 2, "class_label": "building", "position": "top-right", "bbox": [ 0.7546, 0.0248, 0.1909, 0.4281 ] }, { "id": 3, "class_label": "car", "position": "middle-left", "bbox": [ 0.0662, 0.3287, 0.2269, 0.1425 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.1866, 0.1127, 0.1956, 0.1383 ], "class_label": "car" }, { "id": 1, "bbox": [ 0.1072, 0.5843, 0.0987, 0.0606 ], "class_label": "dog" }, { "id": 2, "bbox": [ 0.7546, 0.0248, 0.1909, 0.4281 ], "class_label": "building" }, { "id": 3, "bbox": [ 0.0662, 0.3287, 0.2269, 0.1425 ], "class_label": "car" } ], "batch_id": 3, "task_id": "batch_audit", "difficulty": "hard", "seed": 3304 } ] }, { "batch_id": 4, "scenes": [ { "scene_id": "batch_audit_batch04_scene00", "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=0.06, y=0.39, w=0.16, h=0.13); a dog at bottom-right (bbox: x=0.67, y=0.73, w=0.09, h=0.05); a car at bottom-right (bbox: x=0.77, y=0.77, w=0.23, h=0.15); a car at top-right (bbox: x=0.70, y=0.09, w=0.17, h=0.13); a building at top-left (bbox: x=0.04, y=0.08, w=0.29, h=0.44); a bench at middle-left (bbox: x=0.18, y=0.60, w=0.11, h=0.06); a cat at top-center (bbox: x=0.52, y=0.24, w=0.06, h=0.07); a building at middle-left (bbox: x=0.03, y=0.24, w=0.18, h=0.34); a building at top-right (bbox: x=0.67, y=0.05, w=0.26, h=0.40).", "objects": [ { "id": 0, "class_label": "car", "position": "middle-left", "bbox": [ 0.0603, 0.3855, 0.1599, 0.1282 ] }, { "id": 1, "class_label": "dog", "position": "bottom-right", "bbox": [ 0.6736, 0.7309, 0.0879, 0.0451 ] }, { "id": 2, "class_label": "car", "position": "bottom-right", "bbox": [ 0.7677, 0.7698, 0.2315, 0.1467 ] }, { "id": 3, "class_label": "car", "position": "top-right", "bbox": [ 0.7045, 0.0923, 0.1735, 0.1327 ] }, { "id": 4, "class_label": "building", "position": "top-left", "bbox": [ 0.0362, 0.0757, 0.2932, 0.4416 ] }, { "id": 5, "class_label": "bench", "position": "middle-left", "bbox": [ 0.1814, 0.604, 0.1054, 0.0641 ] }, { "id": 6, "class_label": "cat", "position": "top-center", "bbox": [ 0.5158, 0.2373, 0.0569, 0.0681 ] }, { "id": 7, "class_label": "building", "position": "middle-left", "bbox": [ 0.0329, 0.2409, 0.1778, 0.3434 ] }, { "id": 8, "class_label": "building", "position": "top-right", "bbox": [ 0.6658, 0.0534, 0.2584, 0.4 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.0603, 0.3855, 0.1599, 0.1282 ], "class_label": "car" }, { "id": 1, "bbox": [ 0.6736, 0.7309, 0.0879, 0.0451 ], "class_label": "dog" }, { "id": 2, "bbox": [ 0.7677, 0.7698, 0.2315, 0.1467 ], "class_label": "car" }, { "id": 3, "bbox": [ 0.7045, 0.0923, 0.1735, 0.1327 ], "class_label": "car" }, { "id": 4, "bbox": [ 0.0362, 0.0757, 0.2932, 0.4416 ], "class_label": "building" }, { "id": 5, "bbox": [ 0.1814, 0.604, 0.1054, 0.0641 ], "class_label": "bench" }, { "id": 6, "bbox": [ 0.5158, 0.2373, 0.0569, 0.0681 ], "class_label": "cat" }, { "id": 7, "bbox": [ 0.0329, 0.2409, 0.1778, 0.3434 ], "class_label": "building" }, { "id": 8, "bbox": [ 0.6658, 0.0534, 0.2584, 0.4 ], "class_label": "building" } ], "batch_id": 4, "task_id": "batch_audit", "difficulty": "hard", "seed": 3400 }, { "scene_id": "batch_audit_batch04_scene01", "scene_type": "parking_lot", "scene_description": "A parking lot with various vehicles and some pedestrians. The scene contains 9 objects: a building at center (bbox: x=0.38, y=0.35, w=0.17, h=0.40); a bicycle at bottom-left (bbox: x=0.17, y=0.70, w=0.07, h=0.07); a bicycle at bottom-left (bbox: x=0.24, y=0.72, w=0.08, h=0.10); a bicycle at bottom-left (bbox: x=0.20, y=0.83, w=0.11, h=0.06); a bicycle at bottom-right (bbox: x=0.67, y=0.71, w=0.12, h=0.09); a building at top-left (bbox: x=0.12, y=0.03, w=0.29, h=0.27); a truck at middle-right (bbox: x=0.74, y=0.57, w=0.18, h=0.13); a car at middle-left (bbox: x=0.16, y=0.30, w=0.10, h=0.13); a truck at top-center (bbox: x=0.35, y=0.07, w=0.22, h=0.15).", "objects": [ { "id": 0, "class_label": "building", "position": "center", "bbox": [ 0.3845, 0.3549, 0.1707, 0.3955 ] }, { "id": 1, "class_label": "bicycle", "position": "bottom-left", "bbox": [ 0.1698, 0.6989, 0.066, 0.072 ] }, { "id": 2, "class_label": "bicycle", "position": "bottom-left", "bbox": [ 0.2448, 0.7176, 0.0779, 0.0956 ] }, { "id": 3, "class_label": "bicycle", "position": "bottom-left", "bbox": [ 0.2043, 0.829, 0.1076, 0.0619 ] }, { "id": 4, "class_label": "bicycle", "position": "bottom-right", "bbox": [ 0.6727, 0.7075, 0.1153, 0.0935 ] }, { "id": 5, "class_label": "building", "position": "top-left", "bbox": [ 0.1238, 0.0322, 0.2949, 0.2681 ] }, { "id": 6, "class_label": "truck", "position": "middle-right", "bbox": [ 0.7407, 0.569, 0.1817, 0.1325 ] }, { "id": 7, "class_label": "car", "position": "middle-left", "bbox": [ 0.1648, 0.3048, 0.1001, 0.1262 ] }, { "id": 8, "class_label": "truck", "position": "top-center", "bbox": [ 0.3537, 0.0741, 0.2166, 0.1477 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.3845, 0.3549, 0.1707, 0.3955 ], "class_label": "building" }, { "id": 1, "bbox": [ 0.1698, 0.6989, 0.066, 0.072 ], "class_label": "bicycle" }, { "id": 2, "bbox": [ 0.2448, 0.7176, 0.0779, 0.0956 ], "class_label": "bicycle" }, { "id": 3, "bbox": [ 0.2043, 0.829, 0.1076, 0.0619 ], "class_label": "bicycle" }, { "id": 4, "bbox": [ 0.6727, 0.7075, 0.1153, 0.0935 ], "class_label": "bicycle" }, { "id": 5, "bbox": [ 0.1238, 0.0322, 0.2949, 0.2681 ], "class_label": "building" }, { "id": 6, "bbox": [ 0.7407, 0.569, 0.1817, 0.1325 ], "class_label": "truck" }, { "id": 7, "bbox": [ 0.1648, 0.3048, 0.1001, 0.1262 ], "class_label": "car" }, { "id": 8, "bbox": [ 0.3537, 0.0741, 0.2166, 0.1477 ], "class_label": "truck" } ], "batch_id": 4, "task_id": "batch_audit", "difficulty": "hard", "seed": 3401 }, { "scene_id": "batch_audit_batch04_scene02", "scene_type": "urban_street", "scene_description": "A busy urban street scene with vehicles, pedestrians, and city infrastructure. The scene contains 10 objects: a building at top-right (bbox: x=0.58, y=0.06, w=0.33, h=0.42); a traffic_light at bottom-right (bbox: x=0.88, y=0.80, w=0.02, h=0.09); a bicycle at top-center (bbox: x=0.42, y=0.09, w=0.08, h=0.07); a person at bottom-left (bbox: x=0.09, y=0.69, w=0.05, h=0.23); a person at center (bbox: x=0.45, y=0.53, w=0.07, h=0.19); a truck at center (bbox: x=0.26, y=0.55, w=0.20, h=0.13); a bench at bottom-left (bbox: x=0.19, y=0.75, w=0.08, h=0.07); a tree at middle-right (bbox: x=0.67, y=0.40, w=0.11, h=0.28); a truck at bottom-center (bbox: x=0.35, y=0.81, w=0.18, h=0.14); a building at top-left (bbox: x=0.03, y=0.10, w=0.17, h=0.27).", "objects": [ { "id": 0, "class_label": "building", "position": "top-right", "bbox": [ 0.582, 0.0628, 0.3263, 0.417 ] }, { "id": 1, "class_label": "traffic_light", "position": "bottom-right", "bbox": [ 0.8835, 0.7961, 0.0201, 0.0916 ] }, { "id": 2, "class_label": "bicycle", "position": "top-center", "bbox": [ 0.4193, 0.0919, 0.0849, 0.0664 ] }, { "id": 3, "class_label": "person", "position": "bottom-left", "bbox": [ 0.0933, 0.6903, 0.0489, 0.2298 ] }, { "id": 4, "class_label": "person", "position": "center", "bbox": [ 0.4534, 0.5312, 0.0689, 0.1864 ] }, { "id": 5, "class_label": "truck", "position": "center", "bbox": [ 0.2586, 0.5511, 0.195, 0.1347 ] }, { "id": 6, "class_label": "bench", "position": "bottom-left", "bbox": [ 0.1909, 0.7455, 0.0801, 0.0725 ] }, { "id": 7, "class_label": "tree", "position": "middle-right", "bbox": [ 0.6741, 0.3963, 0.1145, 0.2799 ] }, { "id": 8, "class_label": "truck", "position": "bottom-center", "bbox": [ 0.3472, 0.8104, 0.1778, 0.1392 ] }, { "id": 9, "class_label": "building", "position": "top-left", "bbox": [ 0.0339, 0.0963, 0.1703, 0.2685 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.582, 0.0628, 0.3263, 0.417 ], "class_label": "building" }, { "id": 1, "bbox": [ 0.8835, 0.7961, 0.0201, 0.0916 ], "class_label": "traffic_light" }, { "id": 2, "bbox": [ 0.4193, 0.0919, 0.0849, 0.0664 ], "class_label": "bicycle" }, { "id": 3, "bbox": [ 0.0933, 0.6903, 0.0489, 0.2298 ], "class_label": "person" }, { "id": 4, "bbox": [ 0.4534, 0.5312, 0.0689, 0.1864 ], "class_label": "person" }, { "id": 5, "bbox": [ 0.2586, 0.5511, 0.195, 0.1347 ], "class_label": "truck" }, { "id": 6, "bbox": [ 0.1909, 0.7455, 0.0801, 0.0725 ], "class_label": "bench" }, { "id": 7, "bbox": [ 0.6741, 0.3963, 0.1145, 0.2799 ], "class_label": "tree" }, { "id": 8, "bbox": [ 0.3472, 0.8104, 0.1778, 0.1392 ], "class_label": "truck" }, { "id": 9, "bbox": [ 0.0339, 0.0963, 0.1703, 0.2685 ], "class_label": "building" } ], "batch_id": 4, "task_id": "batch_audit", "difficulty": "hard", "seed": 3402 }, { "scene_id": "batch_audit_batch04_scene03", "scene_type": "park", "scene_description": "A peaceful park setting with trees, benches, and people walking their pets. The scene contains 6 objects: a person at top-right (bbox: x=0.67, y=0.12, w=0.06, h=0.22); a bench at middle-right (bbox: x=0.70, y=0.42, w=0.13, h=0.06); a person at top-right (bbox: x=0.76, y=0.13, w=0.06, h=0.12); a cat at middle-left (bbox: x=0.21, y=0.37, w=0.05, h=0.04); a cat at bottom-left (bbox: x=0.24, y=0.70, w=0.07, h=0.04); a tree at bottom-center (bbox: x=0.47, y=0.70, w=0.10, h=0.20).", "objects": [ { "id": 0, "class_label": "person", "position": "top-right", "bbox": [ 0.6748, 0.1168, 0.0593, 0.2229 ] }, { "id": 1, "class_label": "bench", "position": "middle-right", "bbox": [ 0.7, 0.4222, 0.1292, 0.0632 ] }, { "id": 2, "class_label": "person", "position": "top-right", "bbox": [ 0.758, 0.1301, 0.0557, 0.1168 ] }, { "id": 3, "class_label": "cat", "position": "middle-left", "bbox": [ 0.2124, 0.3705, 0.0525, 0.0401 ] }, { "id": 4, "class_label": "cat", "position": "bottom-left", "bbox": [ 0.2443, 0.6952, 0.0705, 0.0414 ] }, { "id": 5, "class_label": "tree", "position": "bottom-center", "bbox": [ 0.4714, 0.6958, 0.1017, 0.2046 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.6748, 0.1168, 0.0593, 0.2229 ], "class_label": "person" }, { "id": 1, "bbox": [ 0.7, 0.4222, 0.1292, 0.0632 ], "class_label": "bench" }, { "id": 2, "bbox": [ 0.758, 0.1301, 0.0557, 0.1168 ], "class_label": "person" }, { "id": 3, "bbox": [ 0.2124, 0.3705, 0.0525, 0.0401 ], "class_label": "cat" }, { "id": 4, "bbox": [ 0.2443, 0.6952, 0.0705, 0.0414 ], "class_label": "cat" }, { "id": 5, "bbox": [ 0.4714, 0.6958, 0.1017, 0.2046 ], "class_label": "tree" } ], "batch_id": 4, "task_id": "batch_audit", "difficulty": "hard", "seed": 3403 }, { "scene_id": "batch_audit_batch04_scene04", "scene_type": "park", "scene_description": "A peaceful park setting with trees, benches, and people walking their pets. The scene contains 4 objects: a dog at bottom-right (bbox: x=0.78, y=0.75, w=0.05, h=0.05); a dog at middle-right (bbox: x=0.71, y=0.46, w=0.05, h=0.07); a dog at top-right (bbox: x=0.73, y=0.25, w=0.09, h=0.08); a bicycle at middle-right (bbox: x=0.75, y=0.42, w=0.09, h=0.10).", "objects": [ { "id": 0, "class_label": "dog", "position": "bottom-right", "bbox": [ 0.7808, 0.7516, 0.0527, 0.047 ] }, { "id": 1, "class_label": "dog", "position": "middle-right", "bbox": [ 0.7104, 0.4564, 0.0549, 0.074 ] }, { "id": 2, "class_label": "dog", "position": "top-right", "bbox": [ 0.7307, 0.2481, 0.0858, 0.0786 ] }, { "id": 3, "class_label": "bicycle", "position": "middle-right", "bbox": [ 0.7525, 0.4181, 0.0899, 0.0989 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.7808, 0.7516, 0.0527, 0.047 ], "class_label": "dog" }, { "id": 1, "bbox": [ 0.7104, 0.4564, 0.0549, 0.074 ], "class_label": "dog" }, { "id": 2, "bbox": [ 0.7307, 0.2481, 0.0858, 0.0786 ], "class_label": "dog" }, { "id": 3, "bbox": [ 0.7525, 0.4181, 0.0899, 0.0989 ], "class_label": "bicycle" } ], "batch_id": 4, "task_id": "batch_audit", "difficulty": "hard", "seed": 3404 } ] }, { "batch_id": 5, "scenes": [ { "scene_id": "batch_audit_batch05_scene00", "scene_type": "residential_area", "scene_description": "A quiet residential neighborhood with houses, trees, and occasional pedestrians. The scene contains 8 objects: a dog at top-right (bbox: x=0.83, y=0.14, w=0.09, h=0.07); a cat at top-right (bbox: x=0.67, y=0.15, w=0.08, h=0.05); a cat at bottom-center (bbox: x=0.54, y=0.77, w=0.05, h=0.04); a bench at bottom-center (bbox: x=0.48, y=0.76, w=0.12, h=0.07); a building at top-right (bbox: x=0.62, y=0.13, w=0.34, h=0.23); a dog at bottom-center (bbox: x=0.50, y=0.83, w=0.09, h=0.07); a dog at bottom-left (bbox: x=0.08, y=0.67, w=0.08, h=0.08); a car at middle-left (bbox: x=0.21, y=0.44, w=0.14, h=0.14).", "objects": [ { "id": 0, "class_label": "dog", "position": "top-right", "bbox": [ 0.8261, 0.137, 0.0894, 0.0702 ] }, { "id": 1, "class_label": "cat", "position": "top-right", "bbox": [ 0.6673, 0.1548, 0.0767, 0.0463 ] }, { "id": 2, "class_label": "cat", "position": "bottom-center", "bbox": [ 0.5373, 0.7714, 0.0524, 0.0409 ] }, { "id": 3, "class_label": "bench", "position": "bottom-center", "bbox": [ 0.4775, 0.756, 0.1169, 0.0736 ] }, { "id": 4, "class_label": "building", "position": "top-right", "bbox": [ 0.6243, 0.13, 0.3448, 0.2257 ] }, { "id": 5, "class_label": "dog", "position": "bottom-center", "bbox": [ 0.5028, 0.8329, 0.0914, 0.068 ] }, { "id": 6, "class_label": "dog", "position": "bottom-left", "bbox": [ 0.0785, 0.6705, 0.0763, 0.0762 ] }, { "id": 7, "class_label": "car", "position": "middle-left", "bbox": [ 0.2122, 0.4402, 0.1354, 0.1429 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.8261, 0.137, 0.0894, 0.0702 ], "class_label": "dog" }, { "id": 1, "bbox": [ 0.6673, 0.1548, 0.0767, 0.0463 ], "class_label": "cat" }, { "id": 2, "bbox": [ 0.5373, 0.7714, 0.0524, 0.0409 ], "class_label": "cat" }, { "id": 3, "bbox": [ 0.4775, 0.756, 0.1169, 0.0736 ], "class_label": "bench" }, { "id": 4, "bbox": [ 0.6243, 0.13, 0.3448, 0.2257 ], "class_label": "building" }, { "id": 5, "bbox": [ 0.5028, 0.8329, 0.0914, 0.068 ], "class_label": "dog" }, { "id": 6, "bbox": [ 0.0785, 0.6705, 0.0763, 0.0762 ], "class_label": "dog" }, { "id": 7, "bbox": [ 0.2122, 0.4402, 0.1354, 0.1429 ], "class_label": "car" } ], "batch_id": 5, "task_id": "batch_audit", "difficulty": "hard", "seed": 3500 }, { "scene_id": "batch_audit_batch05_scene01", "scene_type": "urban_street", "scene_description": "A busy urban street scene with vehicles, pedestrians, and city infrastructure. The scene contains 6 objects: a car at top-left (bbox: x=0.17, y=0.14, w=0.12, h=0.11); a person at bottom-right (bbox: x=0.71, y=0.64, w=0.08, h=0.15); a car at top-right (bbox: x=0.73, y=0.08, w=0.18, h=0.14); a truck at center (bbox: x=0.30, y=0.31, w=0.19, h=0.12); a truck at top-left (bbox: x=0.04, y=0.14, w=0.25, h=0.14); a building at top-left (bbox: x=0.07, y=0.00, w=0.33, h=0.42).", "objects": [ { "id": 0, "class_label": "car", "position": "top-left", "bbox": [ 0.1698, 0.1422, 0.1154, 0.1108 ] }, { "id": 1, "class_label": "person", "position": "bottom-right", "bbox": [ 0.7146, 0.6374, 0.0782, 0.1503 ] }, { "id": 2, "class_label": "car", "position": "top-right", "bbox": [ 0.7332, 0.0766, 0.1774, 0.1354 ] }, { "id": 3, "class_label": "truck", "position": "center", "bbox": [ 0.2967, 0.3094, 0.1914, 0.1235 ] }, { "id": 4, "class_label": "truck", "position": "top-left", "bbox": [ 0.0351, 0.1352, 0.2507, 0.135 ] }, { "id": 5, "class_label": "building", "position": "top-left", "bbox": [ 0.0696, 0.0, 0.3349, 0.4213 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.1698, 0.1422, 0.1154, 0.1108 ], "class_label": "car" }, { "id": 1, "bbox": [ 0.7146, 0.6374, 0.0782, 0.1503 ], "class_label": "person" }, { "id": 2, "bbox": [ 0.7332, 0.0766, 0.1774, 0.1354 ], "class_label": "car" }, { "id": 3, "bbox": [ 0.2967, 0.3094, 0.1914, 0.1235 ], "class_label": "truck" }, { "id": 4, "bbox": [ 0.0351, 0.1352, 0.2507, 0.135 ], "class_label": "truck" }, { "id": 5, "bbox": [ 0.0696, 0.0, 0.3349, 0.4213 ], "class_label": "building" } ], "batch_id": 5, "task_id": "batch_audit", "difficulty": "hard", "seed": 3501 }, { "scene_id": "batch_audit_batch05_scene02", "scene_type": "urban_street", "scene_description": "A busy urban street scene with vehicles, pedestrians, and city infrastructure. The scene contains 5 objects: a person at top-center (bbox: x=0.50, y=0.06, w=0.06, h=0.15); a car at bottom-center (bbox: x=0.34, y=0.70, w=0.20, h=0.09); a building at top-center (bbox: x=0.34, y=0.08, w=0.32, h=0.28); a bench at bottom-left (bbox: x=0.13, y=0.77, w=0.10, h=0.06); a person at middle-right (bbox: x=0.74, y=0.41, w=0.06, h=0.20).", "objects": [ { "id": 0, "class_label": "person", "position": "top-center", "bbox": [ 0.5019, 0.0596, 0.0594, 0.1481 ] }, { "id": 1, "class_label": "car", "position": "bottom-center", "bbox": [ 0.3436, 0.7049, 0.2044, 0.0929 ] }, { "id": 2, "class_label": "building", "position": "top-center", "bbox": [ 0.3361, 0.0793, 0.3241, 0.2752 ] }, { "id": 3, "class_label": "bench", "position": "bottom-left", "bbox": [ 0.1345, 0.7676, 0.0972, 0.0585 ] }, { "id": 4, "class_label": "person", "position": "middle-right", "bbox": [ 0.744, 0.4106, 0.0599, 0.1958 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.5019, 0.0596, 0.0594, 0.1481 ], "class_label": "person" }, { "id": 1, "bbox": [ 0.3436, 0.7049, 0.2044, 0.0929 ], "class_label": "car" }, { "id": 2, "bbox": [ 0.3361, 0.0793, 0.3241, 0.2752 ], "class_label": "building" }, { "id": 3, "bbox": [ 0.1345, 0.7676, 0.0972, 0.0585 ], "class_label": "bench" }, { "id": 4, "bbox": [ 0.744, 0.4106, 0.0599, 0.1958 ], "class_label": "person" } ], "batch_id": 5, "task_id": "batch_audit", "difficulty": "hard", "seed": 3502 }, { "scene_id": "batch_audit_batch05_scene03", "scene_type": "parking_lot", "scene_description": "A parking lot with various vehicles and some pedestrians. The scene contains 9 objects: a person at middle-left (bbox: x=0.13, y=0.52, w=0.08, h=0.13); a building at bottom-center (bbox: x=0.35, y=0.50, w=0.32, h=0.42); a truck at top-center (bbox: x=0.37, y=0.11, w=0.29, h=0.10); a car at bottom-left (bbox: x=0.09, y=0.71, w=0.14, h=0.14); a truck at top-center (bbox: x=0.43, y=0.13, w=0.17, h=0.16); a person at top-right (bbox: x=0.81, y=0.09, w=0.06, h=0.18); a truck at bottom-center (bbox: x=0.47, y=0.75, w=0.29, h=0.15); a building at bottom-right (bbox: x=0.61, y=0.62, w=0.28, h=0.25); a building at bottom-right (bbox: x=0.79, y=0.73, w=0.21, h=0.22).", "objects": [ { "id": 0, "class_label": "person", "position": "middle-left", "bbox": [ 0.1306, 0.519, 0.0774, 0.1269 ] }, { "id": 1, "class_label": "building", "position": "bottom-center", "bbox": [ 0.3476, 0.5039, 0.3215, 0.4174 ] }, { "id": 2, "class_label": "truck", "position": "top-center", "bbox": [ 0.3707, 0.1148, 0.295, 0.1007 ] }, { "id": 3, "class_label": "car", "position": "bottom-left", "bbox": [ 0.0926, 0.7099, 0.1368, 0.1388 ] }, { "id": 4, "class_label": "truck", "position": "top-center", "bbox": [ 0.4333, 0.1297, 0.1654, 0.1567 ] }, { "id": 5, "class_label": "person", "position": "top-right", "bbox": [ 0.8089, 0.0852, 0.0605, 0.1815 ] }, { "id": 6, "class_label": "truck", "position": "bottom-center", "bbox": [ 0.4671, 0.748, 0.2899, 0.1524 ] }, { "id": 7, "class_label": "building", "position": "bottom-right", "bbox": [ 0.6147, 0.6226, 0.2814, 0.2501 ] }, { "id": 8, "class_label": "building", "position": "bottom-right", "bbox": [ 0.7873, 0.7337, 0.2127, 0.2207 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.1306, 0.519, 0.0774, 0.1269 ], "class_label": "person" }, { "id": 1, "bbox": [ 0.3476, 0.5039, 0.3215, 0.4174 ], "class_label": "building" }, { "id": 2, "bbox": [ 0.3707, 0.1148, 0.295, 0.1007 ], "class_label": "truck" }, { "id": 3, "bbox": [ 0.0926, 0.7099, 0.1368, 0.1388 ], "class_label": "car" }, { "id": 4, "bbox": [ 0.4333, 0.1297, 0.1654, 0.1567 ], "class_label": "truck" }, { "id": 5, "bbox": [ 0.8089, 0.0852, 0.0605, 0.1815 ], "class_label": "person" }, { "id": 6, "bbox": [ 0.4671, 0.748, 0.2899, 0.1524 ], "class_label": "truck" }, { "id": 7, "bbox": [ 0.6147, 0.6226, 0.2814, 0.2501 ], "class_label": "building" }, { "id": 8, "bbox": [ 0.7873, 0.7337, 0.2127, 0.2207 ], "class_label": "building" } ], "batch_id": 5, "task_id": "batch_audit", "difficulty": "hard", "seed": 3503 }, { "scene_id": "batch_audit_batch05_scene04", "scene_type": "residential_area", "scene_description": "A quiet residential neighborhood with houses, trees, and occasional pedestrians. The scene contains 6 objects: a bench at top-center (bbox: x=0.55, y=0.10, w=0.15, h=0.05); a cat at bottom-left (bbox: x=0.11, y=0.86, w=0.05, h=0.04); a cat at top-left (bbox: x=0.23, y=0.22, w=0.07, h=0.05); a bench at center (bbox: x=0.55, y=0.58, w=0.13, h=0.08); a car at top-left (bbox: x=0.04, y=0.23, w=0.24, h=0.14); a car at center (bbox: x=0.28, y=0.40, w=0.17, h=0.13).", "objects": [ { "id": 0, "class_label": "bench", "position": "top-center", "bbox": [ 0.5475, 0.1048, 0.1468, 0.0528 ] }, { "id": 1, "class_label": "cat", "position": "bottom-left", "bbox": [ 0.1131, 0.8564, 0.0475, 0.0427 ] }, { "id": 2, "class_label": "cat", "position": "top-left", "bbox": [ 0.2317, 0.2155, 0.0746, 0.0482 ] }, { "id": 3, "class_label": "bench", "position": "center", "bbox": [ 0.5495, 0.583, 0.1283, 0.0767 ] }, { "id": 4, "class_label": "car", "position": "top-left", "bbox": [ 0.0404, 0.226, 0.2398, 0.1367 ] }, { "id": 5, "class_label": "car", "position": "center", "bbox": [ 0.2768, 0.3984, 0.1746, 0.1273 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.5475, 0.1048, 0.1468, 0.0528 ], "class_label": "bench" }, { "id": 1, "bbox": [ 0.1131, 0.8564, 0.0475, 0.0427 ], "class_label": "cat" }, { "id": 2, "bbox": [ 0.2317, 0.2155, 0.0746, 0.0482 ], "class_label": "cat" }, { "id": 3, "bbox": [ 0.5495, 0.583, 0.1283, 0.0767 ], "class_label": "bench" }, { "id": 4, "bbox": [ 0.0404, 0.226, 0.2398, 0.1367 ], "class_label": "car" }, { "id": 5, "bbox": [ 0.2768, 0.3984, 0.1746, 0.1273 ], "class_label": "car" } ], "batch_id": 5, "task_id": "batch_audit", "difficulty": "hard", "seed": 3504 } ] }, { "batch_id": 6, "scenes": [ { "scene_id": "batch_audit_batch06_scene00", "scene_type": "park", "scene_description": "A peaceful park setting with trees, benches, and people walking their pets. The scene contains 8 objects: a bench at middle-left (bbox: x=0.18, y=0.59, w=0.14, h=0.06); a bicycle at center (bbox: x=0.31, y=0.36, w=0.10, h=0.06); a dog at middle-left (bbox: x=0.15, y=0.45, w=0.09, h=0.07); a person at top-center (bbox: x=0.40, y=0.04, w=0.05, h=0.22); a person at top-left (bbox: x=0.15, y=0.18, w=0.06, h=0.21); a bicycle at center (bbox: x=0.39, y=0.52, w=0.10, h=0.11); a bicycle at top-center (bbox: x=0.54, y=0.15, w=0.10, h=0.09); a bench at middle-right (bbox: x=0.83, y=0.46, w=0.12, h=0.06).", "objects": [ { "id": 0, "class_label": "bench", "position": "middle-left", "bbox": [ 0.185, 0.5899, 0.1357, 0.0572 ] }, { "id": 1, "class_label": "bicycle", "position": "center", "bbox": [ 0.3142, 0.3588, 0.1038, 0.0647 ] }, { "id": 2, "class_label": "dog", "position": "middle-left", "bbox": [ 0.1515, 0.4505, 0.0936, 0.0708 ] }, { "id": 3, "class_label": "person", "position": "top-center", "bbox": [ 0.4044, 0.0439, 0.0487, 0.2227 ] }, { "id": 4, "class_label": "person", "position": "top-left", "bbox": [ 0.153, 0.1771, 0.0633, 0.2116 ] }, { "id": 5, "class_label": "bicycle", "position": "center", "bbox": [ 0.3866, 0.5224, 0.0997, 0.1126 ] }, { "id": 6, "class_label": "bicycle", "position": "top-center", "bbox": [ 0.5354, 0.1479, 0.1034, 0.0905 ] }, { "id": 7, "class_label": "bench", "position": "middle-right", "bbox": [ 0.8341, 0.4607, 0.1248, 0.0579 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.185, 0.5899, 0.1357, 0.0572 ], "class_label": "bench" }, { "id": 1, "bbox": [ 0.3142, 0.3588, 0.1038, 0.0647 ], "class_label": "bicycle" }, { "id": 2, "bbox": [ 0.1515, 0.4505, 0.0936, 0.0708 ], "class_label": "dog" }, { "id": 3, "bbox": [ 0.4044, 0.0439, 0.0487, 0.2227 ], "class_label": "person" }, { "id": 4, "bbox": [ 0.153, 0.1771, 0.0633, 0.2116 ], "class_label": "person" }, { "id": 5, "bbox": [ 0.3866, 0.5224, 0.0997, 0.1126 ], "class_label": "bicycle" }, { "id": 6, "bbox": [ 0.5354, 0.1479, 0.1034, 0.0905 ], "class_label": "bicycle" }, { "id": 7, "bbox": [ 0.8341, 0.4607, 0.1248, 0.0579 ], "class_label": "bench" } ], "batch_id": 6, "task_id": "batch_audit", "difficulty": "hard", "seed": 3600 }, { "scene_id": "batch_audit_batch06_scene01", "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-left (bbox: x=0.09, y=0.60, w=0.12, h=0.23); a building at center (bbox: x=0.33, y=0.31, w=0.29, h=0.35); a person at center (bbox: x=0.41, y=0.44, w=0.05, h=0.17); a car at top-right (bbox: x=0.76, y=0.17, w=0.24, h=0.12); a bicycle at middle-right (bbox: x=0.71, y=0.50, w=0.06, h=0.07); a bicycle at bottom-left (bbox: x=0.11, y=0.68, w=0.09, h=0.12); a person at top-left (bbox: x=0.22, y=0.00, w=0.05, h=0.23); a traffic_light at bottom-left (bbox: x=0.13, y=0.78, w=0.03, h=0.10); a tree at top-center (bbox: x=0.41, y=0.00, w=0.09, h=0.23); a traffic_light at center (bbox: x=0.52, y=0.59, w=0.03, h=0.08).", "objects": [ { "id": 0, "class_label": "tree", "position": "bottom-left", "bbox": [ 0.0898, 0.6022, 0.1234, 0.2261 ] }, { "id": 1, "class_label": "building", "position": "center", "bbox": [ 0.3279, 0.3071, 0.2936, 0.3499 ] }, { "id": 2, "class_label": "person", "position": "center", "bbox": [ 0.4144, 0.4366, 0.0453, 0.1714 ] }, { "id": 3, "class_label": "car", "position": "top-right", "bbox": [ 0.7614, 0.168, 0.2386, 0.1227 ] }, { "id": 4, "class_label": "bicycle", "position": "middle-right", "bbox": [ 0.7092, 0.5007, 0.0633, 0.0713 ] }, { "id": 5, "class_label": "bicycle", "position": "bottom-left", "bbox": [ 0.1074, 0.6833, 0.0935, 0.1197 ] }, { "id": 6, "class_label": "person", "position": "top-left", "bbox": [ 0.2236, 0.0, 0.0459, 0.2318 ] }, { "id": 7, "class_label": "traffic_light", "position": "bottom-left", "bbox": [ 0.1252, 0.7771, 0.0282, 0.0952 ] }, { "id": 8, "class_label": "tree", "position": "top-center", "bbox": [ 0.4118, 0.0, 0.0884, 0.2336 ] }, { "id": 9, "class_label": "traffic_light", "position": "center", "bbox": [ 0.5247, 0.5919, 0.0267, 0.0778 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.0898, 0.6022, 0.1234, 0.2261 ], "class_label": "tree" }, { "id": 1, "bbox": [ 0.3279, 0.3071, 0.2936, 0.3499 ], "class_label": "building" }, { "id": 2, "bbox": [ 0.4144, 0.4366, 0.0453, 0.1714 ], "class_label": "person" }, { "id": 3, "bbox": [ 0.7614, 0.168, 0.2386, 0.1227 ], "class_label": "car" }, { "id": 4, "bbox": [ 0.7092, 0.5007, 0.0633, 0.0713 ], "class_label": "bicycle" }, { "id": 5, "bbox": [ 0.1074, 0.6833, 0.0935, 0.1197 ], "class_label": "bicycle" }, { "id": 6, "bbox": [ 0.2236, 0.0, 0.0459, 0.2318 ], "class_label": "person" }, { "id": 7, "bbox": [ 0.1252, 0.7771, 0.0282, 0.0952 ], "class_label": "traffic_light" }, { "id": 8, "bbox": [ 0.4118, 0.0, 0.0884, 0.2336 ], "class_label": "tree" }, { "id": 9, "bbox": [ 0.5247, 0.5919, 0.0267, 0.0778 ], "class_label": "traffic_light" } ], "batch_id": 6, "task_id": "batch_audit", "difficulty": "hard", "seed": 3601 }, { "scene_id": "batch_audit_batch06_scene02", "scene_type": "intersection", "scene_description": "A road intersection with traffic lights, vehicles, and crossing pedestrians. The scene contains 9 objects: a truck at bottom-left (bbox: x=0.16, y=0.76, w=0.25, h=0.17); a truck at middle-left (bbox: x=0.05, y=0.41, w=0.18, h=0.15); a person at bottom-right (bbox: x=0.79, y=0.73, w=0.05, h=0.25); a car at bottom-right (bbox: x=0.69, y=0.70, w=0.21, h=0.12); a building at middle-left (bbox: x=0.00, y=0.19, w=0.27, h=0.44); a building at bottom-right (bbox: x=0.67, y=0.59, w=0.31, h=0.35); a traffic_light at center (bbox: x=0.35, y=0.42, w=0.03, h=0.07); a car at bottom-center (bbox: x=0.55, y=0.80, w=0.16, h=0.11); a truck at middle-left (bbox: x=0.06, y=0.43, w=0.20, h=0.11).", "objects": [ { "id": 0, "class_label": "truck", "position": "bottom-left", "bbox": [ 0.1566, 0.7561, 0.2516, 0.1691 ] }, { "id": 1, "class_label": "truck", "position": "middle-left", "bbox": [ 0.0542, 0.4123, 0.1778, 0.1453 ] }, { "id": 2, "class_label": "person", "position": "bottom-right", "bbox": [ 0.7884, 0.7274, 0.0527, 0.2451 ] }, { "id": 3, "class_label": "car", "position": "bottom-right", "bbox": [ 0.6904, 0.6959, 0.213, 0.12 ] }, { "id": 4, "class_label": "building", "position": "middle-left", "bbox": [ 0.0, 0.1882, 0.2675, 0.4375 ] }, { "id": 5, "class_label": "building", "position": "bottom-right", "bbox": [ 0.6698, 0.5908, 0.3065, 0.3533 ] }, { "id": 6, "class_label": "traffic_light", "position": "center", "bbox": [ 0.3485, 0.4231, 0.0323, 0.0701 ] }, { "id": 7, "class_label": "car", "position": "bottom-center", "bbox": [ 0.5452, 0.8035, 0.1568, 0.1071 ] }, { "id": 8, "class_label": "truck", "position": "middle-left", "bbox": [ 0.0637, 0.4287, 0.197, 0.1096 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.1566, 0.7561, 0.2516, 0.1691 ], "class_label": "truck" }, { "id": 1, "bbox": [ 0.0542, 0.4123, 0.1778, 0.1453 ], "class_label": "truck" }, { "id": 2, "bbox": [ 0.7884, 0.7274, 0.0527, 0.2451 ], "class_label": "person" }, { "id": 3, "bbox": [ 0.6904, 0.6959, 0.213, 0.12 ], "class_label": "car" }, { "id": 4, "bbox": [ 0.0, 0.1882, 0.2675, 0.4375 ], "class_label": "building" }, { "id": 5, "bbox": [ 0.6698, 0.5908, 0.3065, 0.3533 ], "class_label": "building" }, { "id": 6, "bbox": [ 0.3485, 0.4231, 0.0323, 0.0701 ], "class_label": "traffic_light" }, { "id": 7, "bbox": [ 0.5452, 0.8035, 0.1568, 0.1071 ], "class_label": "car" }, { "id": 8, "bbox": [ 0.0637, 0.4287, 0.197, 0.1096 ], "class_label": "truck" } ], "batch_id": 6, "task_id": "batch_audit", "difficulty": "hard", "seed": 3602 }, { "scene_id": "batch_audit_batch06_scene03", "scene_type": "parking_lot", "scene_description": "A parking lot with various vehicles and some pedestrians. The scene contains 7 objects: a building at bottom-left (bbox: x=0.06, y=0.54, w=0.31, h=0.31); a person at bottom-left (bbox: x=0.18, y=0.83, w=0.05, h=0.12); a person at middle-left (bbox: x=0.09, y=0.47, w=0.05, h=0.14); a building at middle-right (bbox: x=0.72, y=0.45, w=0.28, h=0.38); a bicycle at top-right (bbox: x=0.80, y=0.23, w=0.10, h=0.07); a bicycle at top-center (bbox: x=0.51, y=0.06, w=0.12, h=0.10); a building at top-center (bbox: x=0.49, y=0.00, w=0.26, h=0.42).", "objects": [ { "id": 0, "class_label": "building", "position": "bottom-left", "bbox": [ 0.0627, 0.5448, 0.3076, 0.3118 ] }, { "id": 1, "class_label": "person", "position": "bottom-left", "bbox": [ 0.1815, 0.8337, 0.0525, 0.1216 ] }, { "id": 2, "class_label": "person", "position": "middle-left", "bbox": [ 0.0891, 0.4662, 0.046, 0.1386 ] }, { "id": 3, "class_label": "building", "position": "middle-right", "bbox": [ 0.7173, 0.4514, 0.2827, 0.3785 ] }, { "id": 4, "class_label": "bicycle", "position": "top-right", "bbox": [ 0.8047, 0.2255, 0.1007, 0.0749 ] }, { "id": 5, "class_label": "bicycle", "position": "top-center", "bbox": [ 0.506, 0.06, 0.1191, 0.0976 ] }, { "id": 6, "class_label": "building", "position": "top-center", "bbox": [ 0.4864, 0.0, 0.2596, 0.4221 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.0627, 0.5448, 0.3076, 0.3118 ], "class_label": "building" }, { "id": 1, "bbox": [ 0.1815, 0.8337, 0.0525, 0.1216 ], "class_label": "person" }, { "id": 2, "bbox": [ 0.0891, 0.4662, 0.046, 0.1386 ], "class_label": "person" }, { "id": 3, "bbox": [ 0.7173, 0.4514, 0.2827, 0.3785 ], "class_label": "building" }, { "id": 4, "bbox": [ 0.8047, 0.2255, 0.1007, 0.0749 ], "class_label": "bicycle" }, { "id": 5, "bbox": [ 0.506, 0.06, 0.1191, 0.0976 ], "class_label": "bicycle" }, { "id": 6, "bbox": [ 0.4864, 0.0, 0.2596, 0.4221 ], "class_label": "building" } ], "batch_id": 6, "task_id": "batch_audit", "difficulty": "hard", "seed": 3603 }, { "scene_id": "batch_audit_batch06_scene04", "scene_type": "parking_lot", "scene_description": "A parking lot with various vehicles and some pedestrians. The scene contains 11 objects: a bicycle at bottom-right (bbox: x=0.69, y=0.79, w=0.06, h=0.06); a building at top-right (bbox: x=0.79, y=0.03, w=0.21, h=0.42); a bicycle at middle-right (bbox: x=0.85, y=0.43, w=0.09, h=0.11); a person at top-right (bbox: x=0.78, y=0.08, w=0.08, h=0.24); a car at bottom-right (bbox: x=0.61, y=0.69, w=0.21, h=0.10); a car at middle-right (bbox: x=0.77, y=0.57, w=0.14, h=0.10); a building at bottom-center (bbox: x=0.28, y=0.64, w=0.23, h=0.27); a building at middle-right (bbox: x=0.74, y=0.17, w=0.16, h=0.45); a person at center (bbox: x=0.60, y=0.39, w=0.05, h=0.17); a building at bottom-left (bbox: x=0.12, y=0.64, w=0.21, h=0.36); a bicycle at bottom-right (bbox: x=0.74, y=0.70, w=0.12, h=0.08).", "objects": [ { "id": 0, "class_label": "bicycle", "position": "bottom-right", "bbox": [ 0.6917, 0.792, 0.0639, 0.0641 ] }, { "id": 1, "class_label": "building", "position": "top-right", "bbox": [ 0.7867, 0.0315, 0.2087, 0.4212 ] }, { "id": 2, "class_label": "bicycle", "position": "middle-right", "bbox": [ 0.8463, 0.4283, 0.0939, 0.106 ] }, { "id": 3, "class_label": "person", "position": "top-right", "bbox": [ 0.7823, 0.0804, 0.0781, 0.2394 ] }, { "id": 4, "class_label": "car", "position": "bottom-right", "bbox": [ 0.6085, 0.6918, 0.2085, 0.0995 ] }, { "id": 5, "class_label": "car", "position": "middle-right", "bbox": [ 0.7678, 0.5727, 0.1435, 0.101 ] }, { "id": 6, "class_label": "building", "position": "bottom-center", "bbox": [ 0.2825, 0.6399, 0.2311, 0.2655 ] }, { "id": 7, "class_label": "building", "position": "middle-right", "bbox": [ 0.7361, 0.1735, 0.1592, 0.4462 ] }, { "id": 8, "class_label": "person", "position": "center", "bbox": [ 0.5974, 0.3915, 0.0465, 0.1654 ] }, { "id": 9, "class_label": "building", "position": "bottom-left", "bbox": [ 0.1167, 0.6436, 0.2074, 0.3554 ] }, { "id": 10, "class_label": "bicycle", "position": "bottom-right", "bbox": [ 0.7375, 0.6983, 0.1158, 0.0846 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.6917, 0.792, 0.0639, 0.0641 ], "class_label": "bicycle" }, { "id": 1, "bbox": [ 0.7867, 0.0315, 0.2087, 0.4212 ], "class_label": "building" }, { "id": 2, "bbox": [ 0.8463, 0.4283, 0.0939, 0.106 ], "class_label": "bicycle" }, { "id": 3, "bbox": [ 0.7823, 0.0804, 0.0781, 0.2394 ], "class_label": "person" }, { "id": 4, "bbox": [ 0.6085, 0.6918, 0.2085, 0.0995 ], "class_label": "car" }, { "id": 5, "bbox": [ 0.7678, 0.5727, 0.1435, 0.101 ], "class_label": "car" }, { "id": 6, "bbox": [ 0.2825, 0.6399, 0.2311, 0.2655 ], "class_label": "building" }, { "id": 7, "bbox": [ 0.7361, 0.1735, 0.1592, 0.4462 ], "class_label": "building" }, { "id": 8, "bbox": [ 0.5974, 0.3915, 0.0465, 0.1654 ], "class_label": "person" }, { "id": 9, "bbox": [ 0.1167, 0.6436, 0.2074, 0.3554 ], "class_label": "building" }, { "id": 10, "bbox": [ 0.7375, 0.6983, 0.1158, 0.0846 ], "class_label": "bicycle" } ], "batch_id": 6, "task_id": "batch_audit", "difficulty": "hard", "seed": 3604 } ] }, { "batch_id": 7, "scenes": [ { "scene_id": "batch_audit_batch07_scene00", "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 middle-left (bbox: x=0.22, y=0.54, w=0.11, h=0.08); a building at top-left (bbox: x=0.01, y=0.00, w=0.26, h=0.35); a car at center (bbox: x=0.37, y=0.32, w=0.22, h=0.09); a bicycle at bottom-center (bbox: x=0.45, y=0.84, w=0.10, h=0.09); a car at top-center (bbox: x=0.54, y=0.18, w=0.16, h=0.11); a traffic_light at bottom-left (bbox: x=0.16, y=0.78, w=0.03, h=0.10); a car at bottom-left (bbox: x=0.15, y=0.68, w=0.21, h=0.11).", "objects": [ { "id": 0, "class_label": "bench", "position": "middle-left", "bbox": [ 0.2243, 0.5444, 0.1077, 0.0781 ] }, { "id": 1, "class_label": "building", "position": "top-left", "bbox": [ 0.012, 0.0, 0.255, 0.3458 ] }, { "id": 2, "class_label": "car", "position": "center", "bbox": [ 0.3686, 0.3231, 0.222, 0.087 ] }, { "id": 3, "class_label": "bicycle", "position": "bottom-center", "bbox": [ 0.4536, 0.8385, 0.1034, 0.0919 ] }, { "id": 4, "class_label": "car", "position": "top-center", "bbox": [ 0.5447, 0.1759, 0.1596, 0.1148 ] }, { "id": 5, "class_label": "traffic_light", "position": "bottom-left", "bbox": [ 0.1639, 0.7813, 0.0348, 0.0992 ] }, { "id": 6, "class_label": "car", "position": "bottom-left", "bbox": [ 0.1544, 0.6825, 0.2101, 0.1082 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.2243, 0.5444, 0.1077, 0.0781 ], "class_label": "bench" }, { "id": 1, "bbox": [ 0.012, 0.0, 0.255, 0.3458 ], "class_label": "building" }, { "id": 2, "bbox": [ 0.3686, 0.3231, 0.222, 0.087 ], "class_label": "car" }, { "id": 3, "bbox": [ 0.4536, 0.8385, 0.1034, 0.0919 ], "class_label": "bicycle" }, { "id": 4, "bbox": [ 0.5447, 0.1759, 0.1596, 0.1148 ], "class_label": "car" }, { "id": 5, "bbox": [ 0.1639, 0.7813, 0.0348, 0.0992 ], "class_label": "traffic_light" }, { "id": 6, "bbox": [ 0.1544, 0.6825, 0.2101, 0.1082 ], "class_label": "car" } ], "batch_id": 7, "task_id": "batch_audit", "difficulty": "hard", "seed": 3700 }, { "scene_id": "batch_audit_batch07_scene01", "scene_type": "urban_street", "scene_description": "A busy urban street scene with vehicles, pedestrians, and city infrastructure. The scene contains 9 objects: a traffic_light at bottom-left (bbox: x=0.10, y=0.70, w=0.03, h=0.08); a car at bottom-left (bbox: x=0.05, y=0.66, w=0.10, h=0.15); a building at middle-right (bbox: x=0.55, y=0.29, w=0.32, h=0.37); a bench at bottom-center (bbox: x=0.31, y=0.83, w=0.10, h=0.07); a truck at bottom-left (bbox: x=0.15, y=0.79, w=0.22, h=0.12); a bench at middle-left (bbox: x=0.22, y=0.52, w=0.10, h=0.07); a person at bottom-left (bbox: x=0.22, y=0.75, w=0.07, h=0.15); a tree at bottom-left (bbox: x=0.03, y=0.70, w=0.14, h=0.27); a traffic_light at top-left (bbox: x=0.22, y=0.22, w=0.04, h=0.08).", "objects": [ { "id": 0, "class_label": "traffic_light", "position": "bottom-left", "bbox": [ 0.1008, 0.6958, 0.0341, 0.0797 ] }, { "id": 1, "class_label": "car", "position": "bottom-left", "bbox": [ 0.054, 0.661, 0.1027, 0.1472 ] }, { "id": 2, "class_label": "building", "position": "middle-right", "bbox": [ 0.5547, 0.2891, 0.3162, 0.3673 ] }, { "id": 3, "class_label": "bench", "position": "bottom-center", "bbox": [ 0.3124, 0.8301, 0.1027, 0.0674 ] }, { "id": 4, "class_label": "truck", "position": "bottom-left", "bbox": [ 0.1485, 0.7896, 0.2202, 0.1171 ] }, { "id": 5, "class_label": "bench", "position": "middle-left", "bbox": [ 0.2199, 0.5175, 0.1018, 0.0674 ] }, { "id": 6, "class_label": "person", "position": "bottom-left", "bbox": [ 0.2242, 0.7509, 0.0704, 0.1492 ] }, { "id": 7, "class_label": "tree", "position": "bottom-left", "bbox": [ 0.0318, 0.6992, 0.1394, 0.2734 ] }, { "id": 8, "class_label": "traffic_light", "position": "top-left", "bbox": [ 0.2236, 0.2151, 0.0351, 0.0767 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.1008, 0.6958, 0.0341, 0.0797 ], "class_label": "traffic_light" }, { "id": 1, "bbox": [ 0.054, 0.661, 0.1027, 0.1472 ], "class_label": "car" }, { "id": 2, "bbox": [ 0.5547, 0.2891, 0.3162, 0.3673 ], "class_label": "building" }, { "id": 3, "bbox": [ 0.3124, 0.8301, 0.1027, 0.0674 ], "class_label": "bench" }, { "id": 4, "bbox": [ 0.1485, 0.7896, 0.2202, 0.1171 ], "class_label": "truck" }, { "id": 5, "bbox": [ 0.2199, 0.5175, 0.1018, 0.0674 ], "class_label": "bench" }, { "id": 6, "bbox": [ 0.2242, 0.7509, 0.0704, 0.1492 ], "class_label": "person" }, { "id": 7, "bbox": [ 0.0318, 0.6992, 0.1394, 0.2734 ], "class_label": "tree" }, { "id": 8, "bbox": [ 0.2236, 0.2151, 0.0351, 0.0767 ], "class_label": "traffic_light" } ], "batch_id": 7, "task_id": "batch_audit", "difficulty": "hard", "seed": 3701 }, { "scene_id": "batch_audit_batch07_scene02", "scene_type": "intersection", "scene_description": "A road intersection with traffic lights, vehicles, and crossing pedestrians. The scene contains 8 objects: a car at bottom-left (bbox: x=0.13, y=0.68, w=0.12, h=0.10); a traffic_light at bottom-right (bbox: x=0.69, y=0.75, w=0.02, h=0.09); a traffic_light at top-left (bbox: x=0.17, y=0.17, w=0.03, h=0.08); a bicycle at bottom-center (bbox: x=0.49, y=0.73, w=0.12, h=0.12); a bicycle at top-left (bbox: x=0.14, y=0.11, w=0.09, h=0.11); a bicycle at top-left (bbox: x=0.13, y=0.06, w=0.08, h=0.11); a traffic_light at bottom-left (bbox: x=0.18, y=0.65, w=0.04, h=0.10); a truck at bottom-center (bbox: x=0.45, y=0.67, w=0.18, h=0.17).", "objects": [ { "id": 0, "class_label": "car", "position": "bottom-left", "bbox": [ 0.1278, 0.6751, 0.119, 0.1032 ] }, { "id": 1, "class_label": "traffic_light", "position": "bottom-right", "bbox": [ 0.6948, 0.7514, 0.0241, 0.0887 ] }, { "id": 2, "class_label": "traffic_light", "position": "top-left", "bbox": [ 0.1728, 0.1673, 0.0293, 0.0848 ] }, { "id": 3, "class_label": "bicycle", "position": "bottom-center", "bbox": [ 0.4868, 0.7265, 0.1171, 0.1165 ] }, { "id": 4, "class_label": "bicycle", "position": "top-left", "bbox": [ 0.1354, 0.1134, 0.0944, 0.1064 ] }, { "id": 5, "class_label": "bicycle", "position": "top-left", "bbox": [ 0.1287, 0.0559, 0.0821, 0.1096 ] }, { "id": 6, "class_label": "traffic_light", "position": "bottom-left", "bbox": [ 0.1841, 0.6509, 0.035, 0.1 ] }, { "id": 7, "class_label": "truck", "position": "bottom-center", "bbox": [ 0.452, 0.6726, 0.1765, 0.1718 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.1278, 0.6751, 0.119, 0.1032 ], "class_label": "car" }, { "id": 1, "bbox": [ 0.6948, 0.7514, 0.0241, 0.0887 ], "class_label": "traffic_light" }, { "id": 2, "bbox": [ 0.1728, 0.1673, 0.0293, 0.0848 ], "class_label": "traffic_light" }, { "id": 3, "bbox": [ 0.4868, 0.7265, 0.1171, 0.1165 ], "class_label": "bicycle" }, { "id": 4, "bbox": [ 0.1354, 0.1134, 0.0944, 0.1064 ], "class_label": "bicycle" }, { "id": 5, "bbox": [ 0.1287, 0.0559, 0.0821, 0.1096 ], "class_label": "bicycle" }, { "id": 6, "bbox": [ 0.1841, 0.6509, 0.035, 0.1 ], "class_label": "traffic_light" }, { "id": 7, "bbox": [ 0.452, 0.6726, 0.1765, 0.1718 ], "class_label": "truck" } ], "batch_id": 7, "task_id": "batch_audit", "difficulty": "hard", "seed": 3702 }, { "scene_id": "batch_audit_batch07_scene03", "scene_type": "park", "scene_description": "A peaceful park setting with trees, benches, and people walking their pets. The scene contains 5 objects: a tree at top-right (bbox: x=0.67, y=0.06, w=0.08, h=0.34); a bicycle at center (bbox: x=0.43, y=0.59, w=0.06, h=0.06); a cat at middle-right (bbox: x=0.67, y=0.46, w=0.06, h=0.04); a bicycle at top-center (bbox: x=0.42, y=0.11, w=0.07, h=0.09); a person at bottom-left (bbox: x=0.25, y=0.71, w=0.04, h=0.13).", "objects": [ { "id": 0, "class_label": "tree", "position": "top-right", "bbox": [ 0.6733, 0.064, 0.0836, 0.3354 ] }, { "id": 1, "class_label": "bicycle", "position": "center", "bbox": [ 0.4286, 0.5943, 0.0647, 0.0607 ] }, { "id": 2, "class_label": "cat", "position": "middle-right", "bbox": [ 0.6736, 0.4648, 0.0582, 0.0406 ] }, { "id": 3, "class_label": "bicycle", "position": "top-center", "bbox": [ 0.4235, 0.1078, 0.0672, 0.0864 ] }, { "id": 4, "class_label": "person", "position": "bottom-left", "bbox": [ 0.2463, 0.7052, 0.0447, 0.1329 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.6733, 0.064, 0.0836, 0.3354 ], "class_label": "tree" }, { "id": 1, "bbox": [ 0.4286, 0.5943, 0.0647, 0.0607 ], "class_label": "bicycle" }, { "id": 2, "bbox": [ 0.6736, 0.4648, 0.0582, 0.0406 ], "class_label": "cat" }, { "id": 3, "bbox": [ 0.4235, 0.1078, 0.0672, 0.0864 ], "class_label": "bicycle" }, { "id": 4, "bbox": [ 0.2463, 0.7052, 0.0447, 0.1329 ], "class_label": "person" } ], "batch_id": 7, "task_id": "batch_audit", "difficulty": "hard", "seed": 3703 }, { "scene_id": "batch_audit_batch07_scene04", "scene_type": "intersection", "scene_description": "A road intersection with traffic lights, vehicles, and crossing pedestrians. The scene contains 7 objects: a car at top-center (bbox: x=0.39, y=0.11, w=0.24, h=0.15); a bicycle at top-right (bbox: x=0.71, y=0.12, w=0.11, h=0.11); a truck at bottom-right (bbox: x=0.57, y=0.76, w=0.27, h=0.10); a person at center (bbox: x=0.51, y=0.47, w=0.07, h=0.10); a bicycle at top-left (bbox: x=0.16, y=0.18, w=0.11, h=0.07); a traffic_light at bottom-right (bbox: x=0.70, y=0.77, w=0.04, h=0.07); a truck at top-right (bbox: x=0.57, y=0.07, w=0.26, h=0.12).", "objects": [ { "id": 0, "class_label": "car", "position": "top-center", "bbox": [ 0.386, 0.1106, 0.2387, 0.1473 ] }, { "id": 1, "class_label": "bicycle", "position": "top-right", "bbox": [ 0.713, 0.1225, 0.1137, 0.1121 ] }, { "id": 2, "class_label": "truck", "position": "bottom-right", "bbox": [ 0.5701, 0.7581, 0.2747, 0.1033 ] }, { "id": 3, "class_label": "person", "position": "center", "bbox": [ 0.5074, 0.4722, 0.0715, 0.1048 ] }, { "id": 4, "class_label": "bicycle", "position": "top-left", "bbox": [ 0.1586, 0.1787, 0.1105, 0.0714 ] }, { "id": 5, "class_label": "traffic_light", "position": "bottom-right", "bbox": [ 0.705, 0.767, 0.0394, 0.0679 ] }, { "id": 6, "class_label": "truck", "position": "top-right", "bbox": [ 0.5743, 0.0742, 0.2607, 0.1197 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.386, 0.1106, 0.2387, 0.1473 ], "class_label": "car" }, { "id": 1, "bbox": [ 0.713, 0.1225, 0.1137, 0.1121 ], "class_label": "bicycle" }, { "id": 2, "bbox": [ 0.5701, 0.7581, 0.2747, 0.1033 ], "class_label": "truck" }, { "id": 3, "bbox": [ 0.5074, 0.4722, 0.0715, 0.1048 ], "class_label": "person" }, { "id": 4, "bbox": [ 0.1586, 0.1787, 0.1105, 0.0714 ], "class_label": "bicycle" }, { "id": 5, "bbox": [ 0.705, 0.767, 0.0394, 0.0679 ], "class_label": "traffic_light" }, { "id": 6, "bbox": [ 0.5743, 0.0742, 0.2607, 0.1197 ], "class_label": "truck" } ], "batch_id": 7, "task_id": "batch_audit", "difficulty": "hard", "seed": 3704 } ] }, { "batch_id": 8, "scenes": [ { "scene_id": "batch_audit_batch08_scene00", "scene_type": "parking_lot", "scene_description": "A parking lot with various vehicles and some pedestrians. The scene contains 11 objects: a person at top-left (bbox: x=0.25, y=0.18, w=0.07, h=0.14); a truck at top-center (bbox: x=0.38, y=0.11, w=0.27, h=0.17); a bicycle at bottom-center (bbox: x=0.43, y=0.75, w=0.11, h=0.08); a car at top-center (bbox: x=0.39, y=0.13, w=0.23, h=0.14); a bicycle at bottom-left (bbox: x=0.11, y=0.67, w=0.08, h=0.10); a person at bottom-right (bbox: x=0.86, y=0.60, w=0.07, h=0.24); a person at middle-left (bbox: x=0.11, y=0.44, w=0.06, h=0.13); a bicycle at center (bbox: x=0.31, y=0.55, w=0.12, h=0.09); a bicycle at middle-left (bbox: x=0.25, y=0.42, w=0.10, h=0.08); a building at middle-left (bbox: x=0.10, y=0.19, w=0.28, h=0.40); a truck at center (bbox: x=0.36, y=0.35, w=0.25, h=0.16).", "objects": [ { "id": 0, "class_label": "person", "position": "top-left", "bbox": [ 0.2473, 0.1814, 0.0665, 0.1367 ] }, { "id": 1, "class_label": "truck", "position": "top-center", "bbox": [ 0.376, 0.1089, 0.2713, 0.166 ] }, { "id": 2, "class_label": "bicycle", "position": "bottom-center", "bbox": [ 0.4262, 0.7462, 0.1143, 0.0768 ] }, { "id": 3, "class_label": "car", "position": "top-center", "bbox": [ 0.3851, 0.1297, 0.2251, 0.1354 ] }, { "id": 4, "class_label": "bicycle", "position": "bottom-left", "bbox": [ 0.1123, 0.6728, 0.0774, 0.1005 ] }, { "id": 5, "class_label": "person", "position": "bottom-right", "bbox": [ 0.8609, 0.5983, 0.0746, 0.2428 ] }, { "id": 6, "class_label": "person", "position": "middle-left", "bbox": [ 0.1108, 0.4449, 0.0584, 0.1265 ] }, { "id": 7, "class_label": "bicycle", "position": "center", "bbox": [ 0.3114, 0.5479, 0.1195, 0.0943 ] }, { "id": 8, "class_label": "bicycle", "position": "middle-left", "bbox": [ 0.2487, 0.4236, 0.0962, 0.076 ] }, { "id": 9, "class_label": "building", "position": "middle-left", "bbox": [ 0.102, 0.1938, 0.2838, 0.4028 ] }, { "id": 10, "class_label": "truck", "position": "center", "bbox": [ 0.3632, 0.3544, 0.2478, 0.1627 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.2473, 0.1814, 0.0665, 0.1367 ], "class_label": "person" }, { "id": 1, "bbox": [ 0.376, 0.1089, 0.2713, 0.166 ], "class_label": "truck" }, { "id": 2, "bbox": [ 0.4262, 0.7462, 0.1143, 0.0768 ], "class_label": "bicycle" }, { "id": 3, "bbox": [ 0.3851, 0.1297, 0.2251, 0.1354 ], "class_label": "car" }, { "id": 4, "bbox": [ 0.1123, 0.6728, 0.0774, 0.1005 ], "class_label": "bicycle" }, { "id": 5, "bbox": [ 0.8609, 0.5983, 0.0746, 0.2428 ], "class_label": "person" }, { "id": 6, "bbox": [ 0.1108, 0.4449, 0.0584, 0.1265 ], "class_label": "person" }, { "id": 7, "bbox": [ 0.3114, 0.5479, 0.1195, 0.0943 ], "class_label": "bicycle" }, { "id": 8, "bbox": [ 0.2487, 0.4236, 0.0962, 0.076 ], "class_label": "bicycle" }, { "id": 9, "bbox": [ 0.102, 0.1938, 0.2838, 0.4028 ], "class_label": "building" }, { "id": 10, "bbox": [ 0.3632, 0.3544, 0.2478, 0.1627 ], "class_label": "truck" } ], "batch_id": 8, "task_id": "batch_audit", "difficulty": "hard", "seed": 3800 }, { "scene_id": "batch_audit_batch08_scene01", "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 bottom-right (bbox: x=0.77, y=0.68, w=0.11, h=0.17); a bicycle at center (bbox: x=0.42, y=0.34, w=0.07, h=0.11); a person at bottom-center (bbox: x=0.35, y=0.77, w=0.05, h=0.23); a cat at top-center (bbox: x=0.45, y=0.17, w=0.08, h=0.04).", "objects": [ { "id": 0, "class_label": "tree", "position": "bottom-right", "bbox": [ 0.7695, 0.6837, 0.1087, 0.1679 ] }, { "id": 1, "class_label": "bicycle", "position": "center", "bbox": [ 0.4161, 0.3445, 0.0687, 0.1097 ] }, { "id": 2, "class_label": "person", "position": "bottom-center", "bbox": [ 0.3491, 0.7713, 0.0456, 0.2287 ] }, { "id": 3, "class_label": "cat", "position": "top-center", "bbox": [ 0.4483, 0.1742, 0.0759, 0.0411 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.7695, 0.6837, 0.1087, 0.1679 ], "class_label": "tree" }, { "id": 1, "bbox": [ 0.4161, 0.3445, 0.0687, 0.1097 ], "class_label": "bicycle" }, { "id": 2, "bbox": [ 0.3491, 0.7713, 0.0456, 0.2287 ], "class_label": "person" }, { "id": 3, "bbox": [ 0.4483, 0.1742, 0.0759, 0.0411 ], "class_label": "cat" } ], "batch_id": 8, "task_id": "batch_audit", "difficulty": "hard", "seed": 3801 }, { "scene_id": "batch_audit_batch08_scene02", "scene_type": "residential_area", "scene_description": "A quiet residential neighborhood with houses, trees, and occasional pedestrians. The scene contains 9 objects: a cat at center (bbox: x=0.53, y=0.50, w=0.06, h=0.05); a person at bottom-right (bbox: x=0.86, y=0.70, w=0.06, h=0.25); a building at bottom-center (bbox: x=0.46, y=0.55, w=0.16, h=0.33); a tree at bottom-center (bbox: x=0.42, y=0.70, w=0.13, h=0.21); a dog at center (bbox: x=0.36, y=0.58, w=0.05, h=0.08); a person at middle-left (bbox: x=0.15, y=0.52, w=0.05, h=0.17); a building at top-center (bbox: x=0.25, y=0.07, w=0.24, h=0.39); a person at top-left (bbox: x=0.23, y=0.24, w=0.04, h=0.12); a building at top-center (bbox: x=0.32, y=0.01, w=0.34, h=0.35).", "objects": [ { "id": 0, "class_label": "cat", "position": "center", "bbox": [ 0.5294, 0.4964, 0.058, 0.0472 ] }, { "id": 1, "class_label": "person", "position": "bottom-right", "bbox": [ 0.8642, 0.6966, 0.0575, 0.2485 ] }, { "id": 2, "class_label": "building", "position": "bottom-center", "bbox": [ 0.4645, 0.5453, 0.1584, 0.3262 ] }, { "id": 3, "class_label": "tree", "position": "bottom-center", "bbox": [ 0.4205, 0.6986, 0.1325, 0.2069 ] }, { "id": 4, "class_label": "dog", "position": "center", "bbox": [ 0.3554, 0.577, 0.0527, 0.079 ] }, { "id": 5, "class_label": "person", "position": "middle-left", "bbox": [ 0.1493, 0.523, 0.0484, 0.1734 ] }, { "id": 6, "class_label": "building", "position": "top-center", "bbox": [ 0.2478, 0.0714, 0.2432, 0.394 ] }, { "id": 7, "class_label": "person", "position": "top-left", "bbox": [ 0.2281, 0.2353, 0.0411, 0.1178 ] }, { "id": 8, "class_label": "building", "position": "top-center", "bbox": [ 0.3217, 0.006, 0.3407, 0.3461 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.5294, 0.4964, 0.058, 0.0472 ], "class_label": "cat" }, { "id": 1, "bbox": [ 0.8642, 0.6966, 0.0575, 0.2485 ], "class_label": "person" }, { "id": 2, "bbox": [ 0.4645, 0.5453, 0.1584, 0.3262 ], "class_label": "building" }, { "id": 3, "bbox": [ 0.4205, 0.6986, 0.1325, 0.2069 ], "class_label": "tree" }, { "id": 4, "bbox": [ 0.3554, 0.577, 0.0527, 0.079 ], "class_label": "dog" }, { "id": 5, "bbox": [ 0.1493, 0.523, 0.0484, 0.1734 ], "class_label": "person" }, { "id": 6, "bbox": [ 0.2478, 0.0714, 0.2432, 0.394 ], "class_label": "building" }, { "id": 7, "bbox": [ 0.2281, 0.2353, 0.0411, 0.1178 ], "class_label": "person" }, { "id": 8, "bbox": [ 0.3217, 0.006, 0.3407, 0.3461 ], "class_label": "building" } ], "batch_id": 8, "task_id": "batch_audit", "difficulty": "hard", "seed": 3802 }, { "scene_id": "batch_audit_batch08_scene03", "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 center (bbox: x=0.50, y=0.50, w=0.12, h=0.07); a car at bottom-center (bbox: x=0.56, y=0.74, w=0.11, h=0.15); a car at bottom-right (bbox: x=0.76, y=0.79, w=0.11, h=0.10); a traffic_light at middle-right (bbox: x=0.75, y=0.61, w=0.03, h=0.08); a traffic_light at top-center (bbox: x=0.43, y=0.23, w=0.03, h=0.07); a truck at bottom-center (bbox: x=0.37, y=0.73, w=0.15, h=0.13); a building at top-left (bbox: x=0.16, y=0.04, w=0.16, h=0.24).", "objects": [ { "id": 0, "class_label": "bench", "position": "center", "bbox": [ 0.5034, 0.5007, 0.1184, 0.0663 ] }, { "id": 1, "class_label": "car", "position": "bottom-center", "bbox": [ 0.5595, 0.7412, 0.1079, 0.1485 ] }, { "id": 2, "class_label": "car", "position": "bottom-right", "bbox": [ 0.7632, 0.791, 0.1073, 0.0992 ] }, { "id": 3, "class_label": "traffic_light", "position": "middle-right", "bbox": [ 0.7534, 0.6092, 0.0296, 0.0755 ] }, { "id": 4, "class_label": "traffic_light", "position": "top-center", "bbox": [ 0.4273, 0.2323, 0.033, 0.0712 ] }, { "id": 5, "class_label": "truck", "position": "bottom-center", "bbox": [ 0.3712, 0.732, 0.1508, 0.1271 ] }, { "id": 6, "class_label": "building", "position": "top-left", "bbox": [ 0.1554, 0.0441, 0.1625, 0.2384 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.5034, 0.5007, 0.1184, 0.0663 ], "class_label": "bench" }, { "id": 1, "bbox": [ 0.5595, 0.7412, 0.1079, 0.1485 ], "class_label": "car" }, { "id": 2, "bbox": [ 0.7632, 0.791, 0.1073, 0.0992 ], "class_label": "car" }, { "id": 3, "bbox": [ 0.7534, 0.6092, 0.0296, 0.0755 ], "class_label": "traffic_light" }, { "id": 4, "bbox": [ 0.4273, 0.2323, 0.033, 0.0712 ], "class_label": "traffic_light" }, { "id": 5, "bbox": [ 0.3712, 0.732, 0.1508, 0.1271 ], "class_label": "truck" }, { "id": 6, "bbox": [ 0.1554, 0.0441, 0.1625, 0.2384 ], "class_label": "building" } ], "batch_id": 8, "task_id": "batch_audit", "difficulty": "hard", "seed": 3803 }, { "scene_id": "batch_audit_batch08_scene04", "scene_type": "intersection", "scene_description": "A road intersection with traffic lights, vehicles, and crossing pedestrians. The scene contains 9 objects: a person at middle-right (bbox: x=0.71, y=0.40, w=0.06, h=0.22); a truck at middle-left (bbox: x=0.08, y=0.45, w=0.23, h=0.17); a traffic_light at top-left (bbox: x=0.14, y=0.10, w=0.03, h=0.09); a traffic_light at middle-left (bbox: x=0.15, y=0.41, w=0.02, h=0.06); a traffic_light at bottom-left (bbox: x=0.26, y=0.74, w=0.03, h=0.07); a building at bottom-center (bbox: x=0.44, y=0.67, w=0.16, h=0.32); a traffic_light at top-right (bbox: x=0.70, y=0.09, w=0.03, h=0.09); a car at bottom-left (bbox: x=0.12, y=0.65, w=0.18, h=0.10); a traffic_light at middle-left (bbox: x=0.12, y=0.36, w=0.04, h=0.08).", "objects": [ { "id": 0, "class_label": "person", "position": "middle-right", "bbox": [ 0.7068, 0.4033, 0.0637, 0.2179 ] }, { "id": 1, "class_label": "truck", "position": "middle-left", "bbox": [ 0.0809, 0.4478, 0.234, 0.1691 ] }, { "id": 2, "class_label": "traffic_light", "position": "top-left", "bbox": [ 0.136, 0.0985, 0.0342, 0.0922 ] }, { "id": 3, "class_label": "traffic_light", "position": "middle-left", "bbox": [ 0.1453, 0.405, 0.0246, 0.0608 ] }, { "id": 4, "class_label": "traffic_light", "position": "bottom-left", "bbox": [ 0.2604, 0.7401, 0.0316, 0.0709 ] }, { "id": 5, "class_label": "building", "position": "bottom-center", "bbox": [ 0.4411, 0.6734, 0.1592, 0.317 ] }, { "id": 6, "class_label": "traffic_light", "position": "top-right", "bbox": [ 0.6967, 0.0932, 0.0316, 0.0875 ] }, { "id": 7, "class_label": "car", "position": "bottom-left", "bbox": [ 0.1234, 0.6548, 0.1816, 0.0954 ] }, { "id": 8, "class_label": "traffic_light", "position": "middle-left", "bbox": [ 0.1219, 0.3636, 0.0392, 0.0846 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.7068, 0.4033, 0.0637, 0.2179 ], "class_label": "person" }, { "id": 1, "bbox": [ 0.0809, 0.4478, 0.234, 0.1691 ], "class_label": "truck" }, { "id": 2, "bbox": [ 0.136, 0.0985, 0.0342, 0.0922 ], "class_label": "traffic_light" }, { "id": 3, "bbox": [ 0.1453, 0.405, 0.0246, 0.0608 ], "class_label": "traffic_light" }, { "id": 4, "bbox": [ 0.2604, 0.7401, 0.0316, 0.0709 ], "class_label": "traffic_light" }, { "id": 5, "bbox": [ 0.4411, 0.6734, 0.1592, 0.317 ], "class_label": "building" }, { "id": 6, "bbox": [ 0.6967, 0.0932, 0.0316, 0.0875 ], "class_label": "traffic_light" }, { "id": 7, "bbox": [ 0.1234, 0.6548, 0.1816, 0.0954 ], "class_label": "car" }, { "id": 8, "bbox": [ 0.1219, 0.3636, 0.0392, 0.0846 ], "class_label": "traffic_light" } ], "batch_id": 8, "task_id": "batch_audit", "difficulty": "hard", "seed": 3804 } ] }, { "batch_id": 9, "scenes": [ { "scene_id": "batch_audit_batch09_scene00", "scene_type": "urban_street", "scene_description": "A busy urban street scene with vehicles, pedestrians, and city infrastructure. The scene contains 7 objects: a person at top-center (bbox: x=0.57, y=0.13, w=0.06, h=0.13); a bench at top-right (bbox: x=0.65, y=0.24, w=0.15, h=0.05); a person at bottom-left (bbox: x=0.14, y=0.76, w=0.06, h=0.16); a person at center (bbox: x=0.56, y=0.27, w=0.04, h=0.19); a car at bottom-left (bbox: x=0.23, y=0.71, w=0.12, h=0.11); a traffic_light at middle-right (bbox: x=0.84, y=0.35, w=0.03, h=0.10); a traffic_light at middle-left (bbox: x=0.13, y=0.38, w=0.04, h=0.08).", "objects": [ { "id": 0, "class_label": "person", "position": "top-center", "bbox": [ 0.5735, 0.1255, 0.0594, 0.1288 ] }, { "id": 1, "class_label": "bench", "position": "top-right", "bbox": [ 0.6463, 0.2372, 0.1458, 0.0503 ] }, { "id": 2, "class_label": "person", "position": "bottom-left", "bbox": [ 0.145, 0.7593, 0.0594, 0.1622 ] }, { "id": 3, "class_label": "person", "position": "center", "bbox": [ 0.5637, 0.2652, 0.0414, 0.1903 ] }, { "id": 4, "class_label": "car", "position": "bottom-left", "bbox": [ 0.2269, 0.7135, 0.1224, 0.1121 ] }, { "id": 5, "class_label": "traffic_light", "position": "middle-right", "bbox": [ 0.8442, 0.3543, 0.0338, 0.0974 ] }, { "id": 6, "class_label": "traffic_light", "position": "middle-left", "bbox": [ 0.1279, 0.3786, 0.0393, 0.0811 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.5735, 0.1255, 0.0594, 0.1288 ], "class_label": "person" }, { "id": 1, "bbox": [ 0.6463, 0.2372, 0.1458, 0.0503 ], "class_label": "bench" }, { "id": 2, "bbox": [ 0.145, 0.7593, 0.0594, 0.1622 ], "class_label": "person" }, { "id": 3, "bbox": [ 0.5637, 0.2652, 0.0414, 0.1903 ], "class_label": "person" }, { "id": 4, "bbox": [ 0.2269, 0.7135, 0.1224, 0.1121 ], "class_label": "car" }, { "id": 5, "bbox": [ 0.8442, 0.3543, 0.0338, 0.0974 ], "class_label": "traffic_light" }, { "id": 6, "bbox": [ 0.1279, 0.3786, 0.0393, 0.0811 ], "class_label": "traffic_light" } ], "batch_id": 9, "task_id": "batch_audit", "difficulty": "hard", "seed": 3900 }, { "scene_id": "batch_audit_batch09_scene01", "scene_type": "intersection", "scene_description": "A road intersection with traffic lights, vehicles, and crossing pedestrians. The scene contains 9 objects: a traffic_light at middle-right (bbox: x=0.82, y=0.43, w=0.03, h=0.10); a person at top-right (bbox: x=0.78, y=0.16, w=0.07, h=0.17); a bicycle at top-right (bbox: x=0.83, y=0.19, w=0.09, h=0.08); a car at top-right (bbox: x=0.72, y=0.22, w=0.20, h=0.13); a car at middle-left (bbox: x=0.21, y=0.57, w=0.11, h=0.09); a traffic_light at top-left (bbox: x=0.18, y=0.25, w=0.04, h=0.10); a person at top-left (bbox: x=0.10, y=0.06, w=0.07, h=0.24); a person at bottom-right (bbox: x=0.77, y=0.69, w=0.05, h=0.18); a traffic_light at middle-left (bbox: x=0.14, y=0.37, w=0.03, h=0.09).", "objects": [ { "id": 0, "class_label": "traffic_light", "position": "middle-right", "bbox": [ 0.8151, 0.4321, 0.0301, 0.0961 ] }, { "id": 1, "class_label": "person", "position": "top-right", "bbox": [ 0.7771, 0.1584, 0.0674, 0.174 ] }, { "id": 2, "class_label": "bicycle", "position": "top-right", "bbox": [ 0.832, 0.1905, 0.0885, 0.0754 ] }, { "id": 3, "class_label": "car", "position": "top-right", "bbox": [ 0.7216, 0.219, 0.1972, 0.1333 ] }, { "id": 4, "class_label": "car", "position": "middle-left", "bbox": [ 0.2069, 0.5679, 0.106, 0.0908 ] }, { "id": 5, "class_label": "traffic_light", "position": "top-left", "bbox": [ 0.1797, 0.2495, 0.0379, 0.0978 ] }, { "id": 6, "class_label": "person", "position": "top-left", "bbox": [ 0.1012, 0.0573, 0.0748, 0.2392 ] }, { "id": 7, "class_label": "person", "position": "bottom-right", "bbox": [ 0.7683, 0.6882, 0.0501, 0.1756 ] }, { "id": 8, "class_label": "traffic_light", "position": "middle-left", "bbox": [ 0.1423, 0.372, 0.0273, 0.0852 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.8151, 0.4321, 0.0301, 0.0961 ], "class_label": "traffic_light" }, { "id": 1, "bbox": [ 0.7771, 0.1584, 0.0674, 0.174 ], "class_label": "person" }, { "id": 2, "bbox": [ 0.832, 0.1905, 0.0885, 0.0754 ], "class_label": "bicycle" }, { "id": 3, "bbox": [ 0.7216, 0.219, 0.1972, 0.1333 ], "class_label": "car" }, { "id": 4, "bbox": [ 0.2069, 0.5679, 0.106, 0.0908 ], "class_label": "car" }, { "id": 5, "bbox": [ 0.1797, 0.2495, 0.0379, 0.0978 ], "class_label": "traffic_light" }, { "id": 6, "bbox": [ 0.1012, 0.0573, 0.0748, 0.2392 ], "class_label": "person" }, { "id": 7, "bbox": [ 0.7683, 0.6882, 0.0501, 0.1756 ], "class_label": "person" }, { "id": 8, "bbox": [ 0.1423, 0.372, 0.0273, 0.0852 ], "class_label": "traffic_light" } ], "batch_id": 9, "task_id": "batch_audit", "difficulty": "hard", "seed": 3901 }, { "scene_id": "batch_audit_batch09_scene02", "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=0.21, y=0.21, w=0.06, h=0.05); a tree at top-center (bbox: x=0.57, y=0.11, w=0.14, h=0.16); a cat at top-right (bbox: x=0.87, y=0.13, w=0.06, h=0.07); a person at middle-left (bbox: x=0.15, y=0.40, w=0.05, h=0.14); a bench at bottom-left (bbox: x=0.19, y=0.82, w=0.09, h=0.06).", "objects": [ { "id": 0, "class_label": "cat", "position": "top-left", "bbox": [ 0.2146, 0.208, 0.0613, 0.0476 ] }, { "id": 1, "class_label": "tree", "position": "top-center", "bbox": [ 0.5691, 0.1072, 0.1389, 0.1626 ] }, { "id": 2, "class_label": "cat", "position": "top-right", "bbox": [ 0.866, 0.1276, 0.0575, 0.0688 ] }, { "id": 3, "class_label": "person", "position": "middle-left", "bbox": [ 0.1529, 0.401, 0.0486, 0.1434 ] }, { "id": 4, "class_label": "bench", "position": "bottom-left", "bbox": [ 0.1889, 0.8219, 0.087, 0.0594 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.2146, 0.208, 0.0613, 0.0476 ], "class_label": "cat" }, { "id": 1, "bbox": [ 0.5691, 0.1072, 0.1389, 0.1626 ], "class_label": "tree" }, { "id": 2, "bbox": [ 0.866, 0.1276, 0.0575, 0.0688 ], "class_label": "cat" }, { "id": 3, "bbox": [ 0.1529, 0.401, 0.0486, 0.1434 ], "class_label": "person" }, { "id": 4, "bbox": [ 0.1889, 0.8219, 0.087, 0.0594 ], "class_label": "bench" } ], "batch_id": 9, "task_id": "batch_audit", "difficulty": "hard", "seed": 3902 }, { "scene_id": "batch_audit_batch09_scene03", "scene_type": "urban_street", "scene_description": "A busy urban street scene with vehicles, pedestrians, and city infrastructure. The scene contains 9 objects: a tree at top-left (bbox: x=0.24, y=0.12, w=0.11, h=0.35); a bench at bottom-center (bbox: x=0.30, y=0.80, w=0.12, h=0.06); a bench at top-center (bbox: x=0.48, y=0.20, w=0.13, h=0.07); a bench at center (bbox: x=0.37, y=0.50, w=0.10, h=0.08); a tree at bottom-left (bbox: x=0.06, y=0.69, w=0.12, h=0.22); a bench at top-center (bbox: x=0.55, y=0.11, w=0.13, h=0.07); a tree at bottom-left (bbox: x=0.16, y=0.74, w=0.09, h=0.17); a person at top-left (bbox: x=0.16, y=0.07, w=0.07, h=0.14); a building at bottom-right (bbox: x=0.62, y=0.67, w=0.29, h=0.27).", "objects": [ { "id": 0, "class_label": "tree", "position": "top-left", "bbox": [ 0.2385, 0.1183, 0.107, 0.3473 ] }, { "id": 1, "class_label": "bench", "position": "bottom-center", "bbox": [ 0.2959, 0.8048, 0.1174, 0.0602 ] }, { "id": 2, "class_label": "bench", "position": "top-center", "bbox": [ 0.4848, 0.1974, 0.1252, 0.065 ] }, { "id": 3, "class_label": "bench", "position": "center", "bbox": [ 0.3742, 0.4966, 0.0967, 0.0774 ] }, { "id": 4, "class_label": "tree", "position": "bottom-left", "bbox": [ 0.0557, 0.6881, 0.1188, 0.216 ] }, { "id": 5, "class_label": "bench", "position": "top-center", "bbox": [ 0.5466, 0.1055, 0.1324, 0.0666 ] }, { "id": 6, "class_label": "tree", "position": "bottom-left", "bbox": [ 0.1567, 0.7396, 0.0889, 0.1669 ] }, { "id": 7, "class_label": "person", "position": "top-left", "bbox": [ 0.1568, 0.0744, 0.0697, 0.1408 ] }, { "id": 8, "class_label": "building", "position": "bottom-right", "bbox": [ 0.6182, 0.6697, 0.292, 0.2739 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.2385, 0.1183, 0.107, 0.3473 ], "class_label": "tree" }, { "id": 1, "bbox": [ 0.2959, 0.8048, 0.1174, 0.0602 ], "class_label": "bench" }, { "id": 2, "bbox": [ 0.4848, 0.1974, 0.1252, 0.065 ], "class_label": "bench" }, { "id": 3, "bbox": [ 0.3742, 0.4966, 0.0967, 0.0774 ], "class_label": "bench" }, { "id": 4, "bbox": [ 0.0557, 0.6881, 0.1188, 0.216 ], "class_label": "tree" }, { "id": 5, "bbox": [ 0.5466, 0.1055, 0.1324, 0.0666 ], "class_label": "bench" }, { "id": 6, "bbox": [ 0.1567, 0.7396, 0.0889, 0.1669 ], "class_label": "tree" }, { "id": 7, "bbox": [ 0.1568, 0.0744, 0.0697, 0.1408 ], "class_label": "person" }, { "id": 8, "bbox": [ 0.6182, 0.6697, 0.292, 0.2739 ], "class_label": "building" } ], "batch_id": 9, "task_id": "batch_audit", "difficulty": "hard", "seed": 3903 }, { "scene_id": "batch_audit_batch09_scene04", "scene_type": "intersection", "scene_description": "A road intersection with traffic lights, vehicles, and crossing pedestrians. The scene contains 6 objects: a truck at top-left (bbox: x=0.02, y=0.16, w=0.17, h=0.15); a person at top-left (bbox: x=0.24, y=0.07, w=0.07, h=0.11); a person at bottom-right (bbox: x=0.82, y=0.68, w=0.05, h=0.23); a car at middle-right (bbox: x=0.59, y=0.48, w=0.25, h=0.13); a building at middle-left (bbox: x=0.07, y=0.31, w=0.21, h=0.41); a traffic_light at middle-left (bbox: x=0.14, y=0.51, w=0.03, h=0.09).", "objects": [ { "id": 0, "class_label": "truck", "position": "top-left", "bbox": [ 0.0205, 0.1573, 0.165, 0.1488 ] }, { "id": 1, "class_label": "person", "position": "top-left", "bbox": [ 0.2413, 0.0698, 0.0672, 0.1137 ] }, { "id": 2, "class_label": "person", "position": "bottom-right", "bbox": [ 0.8217, 0.6759, 0.0478, 0.2333 ] }, { "id": 3, "class_label": "car", "position": "middle-right", "bbox": [ 0.5899, 0.4849, 0.2465, 0.1267 ] }, { "id": 4, "class_label": "building", "position": "middle-left", "bbox": [ 0.0711, 0.3115, 0.2105, 0.4143 ] }, { "id": 5, "class_label": "traffic_light", "position": "middle-left", "bbox": [ 0.14, 0.5094, 0.0272, 0.0946 ] } ], "gold_annotations": [ { "id": 0, "bbox": [ 0.0205, 0.1573, 0.165, 0.1488 ], "class_label": "truck" }, { "id": 1, "bbox": [ 0.2413, 0.0698, 0.0672, 0.1137 ], "class_label": "person" }, { "id": 2, "bbox": [ 0.8217, 0.6759, 0.0478, 0.2333 ], "class_label": "person" }, { "id": 3, "bbox": [ 0.5899, 0.4849, 0.2465, 0.1267 ], "class_label": "car" }, { "id": 4, "bbox": [ 0.0711, 0.3115, 0.2105, 0.4143 ], "class_label": "building" }, { "id": 5, "bbox": [ 0.14, 0.5094, 0.0272, 0.0946 ], "class_label": "traffic_light" } ], "batch_id": 9, "task_id": "batch_audit", "difficulty": "hard", "seed": 3904 } ] } ]