Datasets:

sum0214
/

LOCI-revisit-data

sequence_id stringlengths 19 47	map_id stringclasses 7 values	subset stringclasses 1 value	split stringclasses 1 value	duration_seconds float64 60 600	frame_count int64 1.44k 14.4k	fps int64 24 24	width int64 1.28k 1.28k	height int64 720 720	scene_class stringclasses 3 values	trajectory_family stringclasses 5 values	license stringclasses 3 values	caption_en stringclasses 10 values	revisit_pair_count int64 661 7.2k	has_depth bool 2 classes	original_video stringlengths 48 76	compact_video stringlengths 47 75	preview_video stringlengths 44 72	metadata_dir stringlengths 40 68	depth_dir stringclasses 1 value	original_bytes int64 42.5M 11.5B	compact_bytes int64 42.5M 1.58B
bistro_exterior_route_001	bistro_exterior	cc-by	train	120	2,881	24	1,280	720	bright_day	closed_loop	CC-BY-4.0	A first-person camera explores NVIDIA Amazon Lumberyard Bistro Exterior along a collision-verified walkable route and then revisits the same places with exact camera poses.	1,381	false	original/cc-by/train/bistro_exterior_route_001/rgb.mp4	compact/cc-by/train/bistro_exterior_route_001/rgb.mp4	previews/cc-by/train/bistro_exterior_route_001.mp4	metadata/cc-by/train/bistro_exterior_route_001	null	276,333,263	276,333,263
bistro_exterior_route_002	bistro_exterior	cc-by	train	120	2,881	24	1,280	720	bright_day	crossing_or_figure_eight	CC-BY-4.0	A first-person camera explores NVIDIA Amazon Lumberyard Bistro Exterior along a collision-verified walkable route and then revisits the same places with exact camera poses.	1,381	false	original/cc-by/train/bistro_exterior_route_002/rgb.mp4	compact/cc-by/train/bistro_exterior_route_002/rgb.mp4	previews/cc-by/train/bistro_exterior_route_002.mp4	metadata/cc-by/train/bistro_exterior_route_002	null	278,955,079	278,955,079
bistro_exterior_route_003	bistro_exterior	cc-by	train	120	2,881	24	1,280	720	bright_day	random_walk_then_return	CC-BY-4.0	A first-person camera explores NVIDIA Amazon Lumberyard Bistro Exterior along a collision-verified walkable route and then revisits the same places with exact camera poses.	1,381	false	original/cc-by/train/bistro_exterior_route_003/rgb.mp4	compact/cc-by/train/bistro_exterior_route_003/rgb.mp4	previews/cc-by/train/bistro_exterior_route_003.mp4	metadata/cc-by/train/bistro_exterior_route_003	null	305,712,560	305,712,560
bistro_exterior_route_004	bistro_exterior	cc-by	train	600	14,401	24	1,280	720	bright_day	room_or_region_reentry	CC-BY-4.0	A first-person camera explores NVIDIA Amazon Lumberyard Bistro Exterior along a collision-verified walkable route and then revisits the same places with exact camera poses.	7,141	false	original/cc-by/train/bistro_exterior_route_004/rgb.mp4	compact/cc-by/train/bistro_exterior_route_004/rgb.mp4	previews/cc-by/train/bistro_exterior_route_004.mp4	metadata/cc-by/train/bistro_exterior_route_004	null	1,356,385,443	1,356,385,443
carla_town10_route_00	carla_town10hd_opt	cc-by	train	300	7,201	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected Town10 city route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	3,541	false	original/cc-by/train/carla_town10_route_00/rgb.mp4	compact/cc-by/train/carla_town10_route_00/rgb.mp4	previews/cc-by/train/carla_town10_route_00.mp4	metadata/cc-by/train/carla_town10_route_00	null	937,309,244	937,309,244
carla_town10_route_01	carla_town10hd_opt	cc-by	train	300	7,201	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected Town10 city route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	3,541	false	original/cc-by/train/carla_town10_route_01/rgb.mp4	compact/cc-by/train/carla_town10_route_01/rgb.mp4	previews/cc-by/train/carla_town10_route_01.mp4	metadata/cc-by/train/carla_town10_route_01	null	1,075,205,134	1,075,205,134
carla_town10_route_02	carla_town10hd_opt	cc-by	train	300	7,201	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected Town10 city route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	3,541	false	original/cc-by/train/carla_town10_route_02/rgb.mp4	compact/cc-by/train/carla_town10_route_02/rgb.mp4	previews/cc-by/train/carla_town10_route_02.mp4	metadata/cc-by/train/carla_town10_route_02	null	1,072,094,158	1,072,094,158
carla_town10_route_03	carla_town10hd_opt	cc-by	train	300	7,201	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected Town10 city route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	3,541	false	original/cc-by/train/carla_town10_route_03/rgb.mp4	compact/cc-by/train/carla_town10_route_03/rgb.mp4	previews/cc-by/train/carla_town10_route_03.mp4	metadata/cc-by/train/carla_town10_route_03	null	1,060,958,679	1,060,958,679
carla_town10_route_04	carla_town10hd_opt	cc-by	train	300	7,201	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected Town10 city route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	3,541	false	original/cc-by/train/carla_town10_route_04/rgb.mp4	compact/cc-by/train/carla_town10_route_04/rgb.mp4	previews/cc-by/train/carla_town10_route_04.mp4	metadata/cc-by/train/carla_town10_route_04	null	1,131,768,722	1,131,768,722
carla_town10_route_05	carla_town10hd_opt	cc-by	train	300	7,201	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected Town10 city route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	3,541	false	original/cc-by/train/carla_town10_route_05/rgb.mp4	compact/cc-by/train/carla_town10_route_05/rgb.mp4	previews/cc-by/train/carla_town10_route_05.mp4	metadata/cc-by/train/carla_town10_route_05	null	1,063,469,250	1,063,469,250
carla_town10_route_06	carla_town10hd_opt	cc-by	train	300	7,201	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected Town10 city route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	3,541	false	original/cc-by/train/carla_town10_route_06/rgb.mp4	compact/cc-by/train/carla_town10_route_06/rgb.mp4	previews/cc-by/train/carla_town10_route_06.mp4	metadata/cc-by/train/carla_town10_route_06	null	906,463,257	906,463,257
carla_town12_route040_multiaxis_height_rgbd_v30	carla_town12	cc-by	train	600	14,401	24	1,280	720	bright_day_dense_city	out_and_back	CC-BY-4.0? + MIT	A bright daytime revisit trajectory through a dense CARLA Town12 city street with storefronts, intersections, buildings, road markings, trees and street furniture, including full yaw, strong upward/downward views, signed roll and height variation.	7,201	true	original/cc-by/train/carla_town12_route040_multiaxis_height_rgbd_v30/rgb.mp4	compact/cc-by/train/carla_town12_route040_multiaxis_height_rgbd_v30/rgb.mp4	previews/cc-by/train/carla_town12_route040_multiaxis_height_rgbd_v30.mp4	metadata/cc-by/train/carla_town12_route040_multiaxis_height_rgbd_v30	depth/cc-by/train/carla_town12_route040_multiaxis_height_rgbd_v30	11,525,057,806	461,530,986
carla_town12_route_000	carla_town12	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town12 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town12_route_000/rgb.mp4	compact/cc-by/train/carla_town12_route_000/rgb.mp4	previews/cc-by/train/carla_town12_route_000.mp4	metadata/cc-by/train/carla_town12_route_000	null	435,945,453	435,945,453
carla_town12_route_004	carla_town12	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town12 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town12_route_004/rgb.mp4	compact/cc-by/train/carla_town12_route_004/rgb.mp4	previews/cc-by/train/carla_town12_route_004.mp4	metadata/cc-by/train/carla_town12_route_004	null	548,073,196	548,073,196
carla_town12_route_005	carla_town12	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town12 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town12_route_005/rgb.mp4	compact/cc-by/train/carla_town12_route_005/rgb.mp4	previews/cc-by/train/carla_town12_route_005.mp4	metadata/cc-by/train/carla_town12_route_005	null	542,409,566	542,409,566
carla_town12_route_008	carla_town12	cc-by	train	60	1,441	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town12 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	661	false	original/cc-by/train/carla_town12_route_008/rgb.mp4	compact/cc-by/train/carla_town12_route_008/rgb.mp4	previews/cc-by/train/carla_town12_route_008.mp4	metadata/cc-by/train/carla_town12_route_008	null	42,478,315	42,478,315
carla_town12_route_009	carla_town12	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town12 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town12_route_009/rgb.mp4	compact/cc-by/train/carla_town12_route_009/rgb.mp4	previews/cc-by/train/carla_town12_route_009.mp4	metadata/cc-by/train/carla_town12_route_009	null	491,875,106	491,875,106
carla_town12_route_010	carla_town12	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town12 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town12_route_010/rgb.mp4	compact/cc-by/train/carla_town12_route_010/rgb.mp4	previews/cc-by/train/carla_town12_route_010.mp4	metadata/cc-by/train/carla_town12_route_010	null	595,965,231	595,965,231
carla_town12_route_018	carla_town12	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town12 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town12_route_018/rgb.mp4	compact/cc-by/train/carla_town12_route_018/rgb.mp4	previews/cc-by/train/carla_town12_route_018.mp4	metadata/cc-by/train/carla_town12_route_018	null	550,844,042	550,844,042
carla_town12_route_019	carla_town12	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town12 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town12_route_019/rgb.mp4	compact/cc-by/train/carla_town12_route_019/rgb.mp4	previews/cc-by/train/carla_town12_route_019.mp4	metadata/cc-by/train/carla_town12_route_019	null	571,074,263	571,074,263
carla_town12_route_020	carla_town12	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town12 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town12_route_020/rgb.mp4	compact/cc-by/train/carla_town12_route_020/rgb.mp4	previews/cc-by/train/carla_town12_route_020.mp4	metadata/cc-by/train/carla_town12_route_020	null	560,681,247	560,681,247
carla_town12_route_021	carla_town12	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town12 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town12_route_021/rgb.mp4	compact/cc-by/train/carla_town12_route_021/rgb.mp4	previews/cc-by/train/carla_town12_route_021.mp4	metadata/cc-by/train/carla_town12_route_021	null	580,647,520	580,647,520
carla_town12_route_022	carla_town12	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town12 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town12_route_022/rgb.mp4	compact/cc-by/train/carla_town12_route_022/rgb.mp4	previews/cc-by/train/carla_town12_route_022.mp4	metadata/cc-by/train/carla_town12_route_022	null	658,877,351	658,877,351
carla_town12_route_023	carla_town12	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town12 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town12_route_023/rgb.mp4	compact/cc-by/train/carla_town12_route_023/rgb.mp4	previews/cc-by/train/carla_town12_route_023.mp4	metadata/cc-by/train/carla_town12_route_023	null	427,521,543	427,521,543
carla_town12_route_025	carla_town12	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town12 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town12_route_025/rgb.mp4	compact/cc-by/train/carla_town12_route_025/rgb.mp4	previews/cc-by/train/carla_town12_route_025.mp4	metadata/cc-by/train/carla_town12_route_025	null	547,050,462	547,050,462
carla_town12_route_027	carla_town12	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town12 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town12_route_027/rgb.mp4	compact/cc-by/train/carla_town12_route_027/rgb.mp4	previews/cc-by/train/carla_town12_route_027.mp4	metadata/cc-by/train/carla_town12_route_027	null	490,832,380	490,832,380
carla_town12_route_028	carla_town12	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town12 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town12_route_028/rgb.mp4	compact/cc-by/train/carla_town12_route_028/rgb.mp4	previews/cc-by/train/carla_town12_route_028.mp4	metadata/cc-by/train/carla_town12_route_028	null	585,799,729	585,799,729
carla_town12_route_029	carla_town12	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town12 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town12_route_029/rgb.mp4	compact/cc-by/train/carla_town12_route_029/rgb.mp4	previews/cc-by/train/carla_town12_route_029.mp4	metadata/cc-by/train/carla_town12_route_029	null	536,821,442	536,821,442
carla_town12_route_030	carla_town12	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town12 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town12_route_030/rgb.mp4	compact/cc-by/train/carla_town12_route_030/rgb.mp4	previews/cc-by/train/carla_town12_route_030.mp4	metadata/cc-by/train/carla_town12_route_030	null	454,309,689	454,309,689
carla_town13_route_006	carla_town13	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town13 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town13_route_006/rgb.mp4	compact/cc-by/train/carla_town13_route_006/rgb.mp4	previews/cc-by/train/carla_town13_route_006.mp4	metadata/cc-by/train/carla_town13_route_006	null	758,311,379	758,311,379
carla_town13_route_007	carla_town13	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town13 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town13_route_007/rgb.mp4	compact/cc-by/train/carla_town13_route_007/rgb.mp4	previews/cc-by/train/carla_town13_route_007.mp4	metadata/cc-by/train/carla_town13_route_007	null	772,530,456	772,530,456
carla_town13_route_008	carla_town13	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town13 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town13_route_008/rgb.mp4	compact/cc-by/train/carla_town13_route_008/rgb.mp4	previews/cc-by/train/carla_town13_route_008.mp4	metadata/cc-by/train/carla_town13_route_008	null	664,625,355	664,625,355
carla_town13_route_009	carla_town13	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town13 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town13_route_009/rgb.mp4	compact/cc-by/train/carla_town13_route_009/rgb.mp4	previews/cc-by/train/carla_town13_route_009.mp4	metadata/cc-by/train/carla_town13_route_009	null	806,101,535	806,101,535
carla_town13_route_013	carla_town13	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town13 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town13_route_013/rgb.mp4	compact/cc-by/train/carla_town13_route_013/rgb.mp4	previews/cc-by/train/carla_town13_route_013.mp4	metadata/cc-by/train/carla_town13_route_013	null	657,673,982	657,673,982
carla_town13_route_020	carla_town13	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town13 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town13_route_020/rgb.mp4	compact/cc-by/train/carla_town13_route_020/rgb.mp4	previews/cc-by/train/carla_town13_route_020.mp4	metadata/cc-by/train/carla_town13_route_020	null	691,972,848	691,972,848
carla_town13_route_022	carla_town13	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town13 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town13_route_022/rgb.mp4	compact/cc-by/train/carla_town13_route_022/rgb.mp4	previews/cc-by/train/carla_town13_route_022.mp4	metadata/cc-by/train/carla_town13_route_022	null	650,470,526	650,470,526
carla_town13_route_026	carla_town13	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town13 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town13_route_026/rgb.mp4	compact/cc-by/train/carla_town13_route_026/rgb.mp4	previews/cc-by/train/carla_town13_route_026.mp4	metadata/cc-by/train/carla_town13_route_026	null	618,521,642	618,521,642
carla_town13_route_028	carla_town13	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town13 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town13_route_028/rgb.mp4	compact/cc-by/train/carla_town13_route_028/rgb.mp4	previews/cc-by/train/carla_town13_route_028.mp4	metadata/cc-by/train/carla_town13_route_028	null	715,784,769	715,784,769
carla_town13_route_029	carla_town13	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town13 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town13_route_029/rgb.mp4	compact/cc-by/train/carla_town13_route_029/rgb.mp4	previews/cc-by/train/carla_town13_route_029.mp4	metadata/cc-by/train/carla_town13_route_029	null	749,626,669	749,626,669
carla_town13_route_035	carla_town13	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town13 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town13_route_035/rgb.mp4	compact/cc-by/train/carla_town13_route_035/rgb.mp4	previews/cc-by/train/carla_town13_route_035.mp4	metadata/cc-by/train/carla_town13_route_035	null	663,171,495	663,171,495
carla_town13_route_036	carla_town13	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town13 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town13_route_036/rgb.mp4	compact/cc-by/train/carla_town13_route_036/rgb.mp4	previews/cc-by/train/carla_town13_route_036.mp4	metadata/cc-by/train/carla_town13_route_036	null	654,951,229	654,951,229
carla_town13_route_037	carla_town13	cc-by	train	600	14,401	24	1,280	720	null	out_and_back	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows the center of a connected CARLA Town13 route, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town13_route_037/rgb.mp4	compact/cc-by/train/carla_town13_route_037/rgb.mp4	previews/cc-by/train/carla_town13_route_037.mp4	metadata/cc-by/train/carla_town13_route_037	null	712,492,593	712,492,593
carla_town13_route_040	carla_town13	cc-by	train	600	14,401	24	1,280	720	null	random_walk_then_return	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows a deterministically seeded sequence of road branches in CARLA Town13, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town13_route_040/rgb.mp4	compact/cc-by/train/carla_town13_route_040/rgb.mp4	previews/cc-by/train/carla_town13_route_040.mp4	metadata/cc-by/train/carla_town13_route_040	null	749,402,166	749,402,166
carla_town13_route_061	carla_town13	cc-by	train	600	14,401	24	1,280	720	null	random_walk_then_return	CC-BY-4.0? + MIT	A static-clear-day first-person camera follows a deterministically seeded sequence of road branches in CARLA Town13, passes several road segments and intersections, then returns along the same path for exact spatial revisits.	7,141	false	original/cc-by/train/carla_town13_route_061/rgb.mp4	compact/cc-by/train/carla_town13_route_061/rgb.mp4	previews/cc-by/train/carla_town13_route_061.mp4	metadata/cc-by/train/carla_town13_route_061	null	829,969,384	829,969,384
classroom_route_000	classroom	cc-by	train	60	1,441	24	1,280	720	bright_interior	room_or_region_reentry	CC0-1.0	A first-person camera explores Blender Classroom along a collision-verified walkable route and then revisits the same places with exact camera poses.	661	false	original/cc-by/train/classroom_route_000/rgb.mp4	compact/cc-by/train/classroom_route_000/rgb.mp4	previews/cc-by/train/classroom_route_000.mp4	metadata/cc-by/train/classroom_route_000	null	70,316,346	70,316,346
dogwalk_winter_forest_route_000	dogwalk_winter_forest	cc-by	train	600	14,401	24	1,280	720	bright_day	room_or_region_reentry	CC-BY-4.0	A bright stylized winter-forest tour revisits the same six regions after a continuous return bridge, with broad natural camera yaw, pitch, roll and height variation.	6,240	false	original/cc-by/train/dogwalk_winter_forest_route_000/rgb.mp4	compact/cc-by/train/dogwalk_winter_forest_route_000/rgb.mp4	previews/cc-by/train/dogwalk_winter_forest_route_000.mp4	metadata/cc-by/train/dogwalk_winter_forest_route_000	null	1,580,680,674	1,580,680,674
dogwalk_winter_forest_route_001	dogwalk_winter_forest	cc-by	train	600	14,401	24	1,280	720	bright_day	room_or_region_reentry	CC-BY-4.0	A bright stylized winter-forest closed loop revisits five landmark regions after a continuous return bridge, with broad natural camera yaw, pitch, roll and height variation.	6,240	false	original/cc-by/train/dogwalk_winter_forest_route_001/rgb.mp4	compact/cc-by/train/dogwalk_winter_forest_route_001/rgb.mp4	previews/cc-by/train/dogwalk_winter_forest_route_001.mp4	metadata/cc-by/train/dogwalk_winter_forest_route_001	null	1,579,088,427	1,579,088,427
polyhaven_smugglers_cove_west_fort_route_000	poly_haven_smugglers_cove_west_fort_static_cloud_v9	cc-by	train	567	13,609	24	1,280	720	null	closed_loop	CC0-1.0	A daylight first-person camera explores the authored stone fort, arches, towers, moss and rock of Poly Haven's Smugglers Cove along a continuous supported-surface closed route, then exactly replays every camera pose for long-gap visual-memory revisits. The camera covers every yaw sector, strong upward and downward view...	6,804	false	original/cc-by/train/polyhaven_smugglers_cove_west_fort_route_000/rgb.mp4	compact/cc-by/train/polyhaven_smugglers_cove_west_fort_route_000/rgb.mp4	previews/cc-by/train/polyhaven_smugglers_cove_west_fort_route_000.mp4	metadata/cc-by/train/polyhaven_smugglers_cove_west_fort_route_000	null	1,036,180,753	1,036,180,753

Dataset of LOCI: Spatial Linear Memory for Streaming World Models — arXiv:2609.40222 · HF paper page · project page · code

LOCI UE-Revisit v1

Long, continuous first-person / free-flight videos rendered in game engines (Unreal Engine 5, CARLA, Stunt Rally / OgreNext, Xonotic, Red Eclipse, Unvanquished, Godot) with per-frame camera poses and intrinsics and exact-pose revisit pairs: the camera returns to places it has already seen, and every return frame is linked to the earlier frame rendered from the identical (or a recorded nearby) camera pose. It was built to train and evaluate long-horizon spatial memory in video world models.

This v1 release contains 138 sequences from 49 maps, 17.60 hours, 1,520,463 frames at 1280x720 / 24 fps, and 741,064 revisit pairs, split into license subsets.

	size
`original/` - the delivered videos, byte-identical (H.264; lossless RGB `gbrp` High 4:4:4 for 54 sequences, CRF 12-16 yuv420p for the rest)	738.0 GB
`compact/` - visually-lossless H.264 yuv420p (lossless sources re-encoded at CRF 16 / preset slow; already-lossy sources are the identical file)	111.1 GB
`depth/` - per-frame depth where it was captured	21.0 GB
`metadata/` - poses, intrinsics, timestamps, revisit pairs, captions, licenses, source records	3.4 GB
`previews/` - 384x216 12 fps browsing clips	1.0 GB

Pick either original/ or compact/; metadata, frame indices and frame counts are identical for both (every video was checked: decoded frame count == metadata frame count).

Subsets

Each sequence is assigned to exactly one subset by the license of the scene it was rendered from (see LICENSES.md). Each subset is a separate config; splits are by map (no map appears in two splits), inherited from the original dataset split, so per-subset split sizes are uneven.

subset	maps	sequences	hours	train / val / test seq	original	compact	depth	revisit pairs
`cc-by`	10	67	9.57	48 / 18 / 1	60.5 GB	49.5 GB	12.8 GB (1 seq)	407,968
`share-alike`	38	69	7.69	62 / 5 / 2	673.1 GB	57.3 GB	8.2 GB (9 seq)	318,814
`non-commercial`	1	2	0.33	0 / 2 / 0	4.3 GB	4.3 GB	0.0 GB (0 seq)	14,282

cc-by: CC0 and CC BY (attribution-only) scenes: CARLA towns, Poly Haven, Blender demo, Amazon Lumberyard Bistro, Blender Studio DOGWALK.
share-alike: copyleft scenes: GPL-3.0 (Xonotic, Stunt Rally tracks), CC BY-SA (Red Eclipse, Unvanquished, Hallwyl gallery, SuperTuxKart), and Stunt Rally tracks whose rendered media mix CC0 / CC BY / CC BY-SA / GPL.
non-commercial: CC BY-NC-SA 3.0 (NVIDIA Emerald Square).

21 further maps (61 sequences, 6.69 h) from the source delivery are withheld from v1 because their asset licenses are unclear or restrictive (UnrealZoo/UE-Marketplace packages pending asset audit, AirSim precompiled environments, ASWF-licensed scenes, Intel Sponza add-ons).

Loading

from datasets import load_dataset
ds = load_dataset("<org>/<repo>", "cc-by")          # index rows; videos are referenced by path
row = ds["train"][0]; row["compact_video"], row["metadata_dir"]

from huggingface_hub import snapshot_download        # fetch files: e.g. compact videos + metadata of one subset
snapshot_download("<org>/<repo>", repo_type="dataset",
                  allow_patterns=["index/cc-by/*", "metadata/cc-by/*", "compact/cc-by/*", "licenses/*", "*.md"])

Layout

index/<subset>/<split>.jsonl                 one row per sequence (HF configs point here)
original/<subset>/<split>/<seq>/rgb.mp4      delivered video, byte-identical (sha256 in sequence.json + MANIFEST)
compact/<subset>/<split>/<seq>/rgb.mp4       H.264 yuv420p, 1280x720, 24 fps, same frames
depth/<subset>/<split>/<seq>/                depth_exr/NNNNNN.exr (float32 OpenEXR) or depth.mkv (1 CARLA sequence)
previews/<subset>/<split>/<seq>.mp4          small browsing clip
metadata/<subset>/<split>/<seq>/
    sequence.json     release summary (map, split, fps, frames, videos, sha256, license, coordinate system)
    frames.jsonl      normalized per-frame camera (one line per video frame)
    revisits.jsonl    normalized revisit pairs
    captions.json     {en, zh, segments}
    license.json      exact license(s), source URL, required attribution, original license record
    source/           the original per-sequence records (camera.jsonl, camera.json, trajectory.json,
                      revisits.json, sequence.json, acceptance.json, QA json, SOURCE.md, depth index)
licenses/notices/   upstream license / source notices      licenses/audits/   Stunt Rally per-asset license audits
LICENSES.md  ATTRIBUTION.md  MANIFEST.sha256  manifest.jsonl

Schema

frames.jsonl (exactly frame_count lines, frame_index 0..N-1 = decoded video frame index):

field	meaning
`frame_index`, `timestamp_s`	video frame index; time = index / 24
`position_m`	camera centre `[x, y, z]` in the sequence's canonical right-handed, z-up world. Metres, except Stunt Rally (1 world unit treated as 1 m) and Red Eclipse (16 engine units = 1 nominal metre); see `sequence.json -> poses.coordinate_systems`
`c2w_opencv`	3x3 camera-to-world rotation, OpenCV camera axes (x right, y down, z forward)
`intrinsics`	`fx, fy, cx, cy, width, height` in pixels (pinhole, no distortion)
`phase`	route phase from the source (e.g. `first_visit`, `outbound`, `return`, `revisit`)
`paired_revisit_frame_index`	for mirrored out-and-back routes, the frame rendered from the identical pose; else null

c2w_opencv is derived from each engine's native record (UE rotator, Quake/Tesseract angles, Ogre camera-to-world matrix, Godot basis); the conversion per family is listed in sequence.json -> poses. It was checked against the images: for frame pairs taken from the same position with different orientation the predicted rotation homography must align the two frames better than wrong-convention alternatives (sequence.json -> poses.rotation_image_check). The untouched engine records are kept in source/camera.jsonl.

revisits.jsonl: history_frame, current_frame (later frame), gap_seconds, position_distance_m, view_angle_deg (angle between the two viewing directions; 0 for exact-pose pairs), event_type, ground_truth_level.

How revisits work

Every route is planned so that the camera re-enters places it has already seen. Families: out_and_back / exact-reverse (the second half replays the first half's poses in reverse order, so frame N-1-i has exactly the pose of frame i), closed_loop, crossing_or_figure_eight, room_or_region_reentry, delayed same-order replays, and typed re-entries that return to the same position with a different orientation. Pairs are emitted only where the source pipeline verified the pose (engine transform readback) and, for many sequences, photometric agreement of the paired frames. Minimum gap is 5 s for most families.

Intended use

Research on long-horizon memory, camera-controlled video generation, novel-view / revisit consistency, pose estimation and SLAM-style evaluation in static scenes. Respect each subset's license; the non-commercial subset may not be used commercially; share-alike derivatives must be shared alike.

Limitations

Game-engine visuals. Rendered, not captured: clean lighting, no sensor noise or motion blur, finite asset detail, repeated textures; some maps (Quake-family arenas, Stunt Rally tracks) have a stylized look.
Static worlds. No pedestrians, vehicles or animated foliage (dynamics disabled for exact repeats).
Exact-reverse trajectories. Many sequences are out-and-back replays: the second half is the first half reversed. Models can exploit this (copy mirrored frames) instead of learning memory; hold out or sub-sample reverse halves accordingly.
Scale. Stunt Rally / Red Eclipse units are nominal, not calibrated metres.
Uneven splits. Splits are by map and inherited from the source dataset; some subsets have few or no test maps (non-commercial has validation only).
Pose records. In 10 Unreal-native sequences (Bistro, Pawn Shop, Classroom, Hallwyl Gallery, Emerald Square, Namaqualand) the source camera.jsonl rotation_degrees fields are mislabeled (UE roll/pitch/yaw stored in yaw/pitch/roll slots without the handedness sign flip); frames.jsonl is derived from the raw UE rotator in source/camera.json instead. Use frames.jsonl.
English captions are template-like for many sequences.

Licenses

license: other because every sequence carries the license of the scene it was rendered from. Per-map:

map	subset	seq	minutes	splits	license(s)	source
`bistro_exterior`	`cc-by`	4	16.0	train	CC-BY-4.0	Amazon Lumberyard Bistro v5.2 (NVIDIA ORCA)
`carla_town10hd_opt`	`cc-by`	7	35.0	train	CC-BY (CARLA-specific assets; Creative Commons version not stated upstream); MIT (CARLA code)	CARLA Simulator 0.10.0 (Computer Vision Center, UAB)
`carla_town12`	`cc-by`	18	171.0	train	CC-BY (CARLA-specific assets; Creative Commons version not stated upstream); MIT (CARLA code)	CARLA Simulator 0.9.16 (Computer Vision Center, UAB)
`carla_town13`	`cc-by`	15	150.0	train	CC-BY (CARLA-specific assets; Creative Commons version not stated upstream); MIT (CARLA code)	CARLA Simulator 0.9.16 (Computer Vision Center, UAB)
`carla_town15`	`cc-by`	17	170.0	validation	CC-BY (CARLA-specific assets; Creative Commons version not stated upstream); MIT (CARLA code)	CARLA Simulator 0.9.16 (Computer Vision Center, UAB)
`classroom`	`cc-by`	1	1.0	train	CC0-1.0	Blender demo file "Classroom" by Christophe Seux
`dogwalk_winter_forest`	`cc-by`	2	20.0	train	CC-BY-4.0	Blender Studio DOGWALK v1.0.4 (Godot project)
`namaqualand`	`cc-by`	1	1.0	validation	CC0-1.0	Poly Haven - Namaqualand
`pawn_shop`	`cc-by`	1	1.0	test	CC0-1.0	Poly Haven - Pawn Shop
`poly_haven_smugglers_cove_west_fort_static_cloud_v9`	`cc-by`	1	9.4	train	CC0-1.0	Poly Haven - Smuggler's Cove
`mcguire_gallery`	`share-alike`	1	1.0	train	CC-BY-SA-4.0	The Picture Gallery, The Hallwyl Museum (via Morgan McGuire's Computer Graphics Archive)
`redeclipse_institute`	`share-alike`	1	1.0	train	CC-BY-SA-4.0-or-later	Red Eclipse 2.0.0 official map "Institute"
`redeclipse_octavus`	`share-alike`	1	1.0	train	CC-BY-SA-3.0	Red Eclipse 2.0.0 official map "Octavus"
`stk_candela_city`	`share-alike`	3	32.4	validation	CC-BY-SA-4.0; CC-BY-SA-3.0; GPL-3.0-or-later	SuperTuxKart track "Candela City" by Jean-Manuel Clemencon (static-scene derivative)
`stuntrally_atm10_lakecity_sanitized`	`share-alike`	2	19.8	train	GPL-3.0-only (track); CC-BY-4.0 (render media); CC-BY-SA-3.0 (render media); CC-BY-SA-3.0 WITH CC-BY-3.0 AND CC0 COMPONENTS (render media); CC0-1.0 (render media); GPL-3.0-only (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`stuntrally_aus3_canyon_sanitized_collision_runtime_v1`	`share-alike`	1	15.0	train	GPL-3.0-only (track); CC-BY-3.0 (render media); CC-BY-4.0 (render media); CC-BY-SA-3.0 (render media); CC-BY-SA-3.0 WITH CC-BY-3.0 AND CC0 COMPONENTS (render media); CC0-1.0 (render media); CUSTOM-PERMISSIVE-FREE-USE (render media); GPL-3.0-only (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`stuntrally_can8_redopoka_bright_sanitized_collision_runtime_v2`	`share-alike`	1	10.0	train	GPL-3.0-only (track); CC-BY-4.0 (render media); CC-BY-SA-3.0 (render media); CC0-1.0 (render media); CUSTOM-PERMISSIVE-FREE-USE (render media); GPL-3.0-only (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`stuntrally_des11_persiancity_sanitized`	`share-alike`	3	17.0	train	GPL-3.0-only (track); CC-BY-3.0 (render media); CC-BY-4.0 (render media); CC-BY-SA-3.0 (render media); CC0-1.0 (render media); GPL-3.0-only (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`stuntrally_des15_citywall_sanitized`	`share-alike`	2	16.0	train	GPL-3.0-only (track); CC-BY-3.0 (render media); CC-BY-4.0 (render media); CC-BY-SA-3.0 (render media); CC-BY-SA-3.0 WITH CC-BY-3.0 AND CC0 COMPONENTS (render media); CC0-1.0 (render media); GPL-3.0-only (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`stuntrally_des17_nilecity_sanitized`	`share-alike`	2	16.0	train	GPL-3.0-only (track); CC-BY-3.0 (render media); CC-BY-4.0 (render media); CC-BY-SA-3.0 (render media); CC-BY-SA-3.0 WITH CC-BY-3.0 AND CC0 COMPONENTS (render media); CC0-1.0 (render media); GPL-3.0-only (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`stuntrally_fin1_lakes_sanitized`	`share-alike`	1	15.0	train	GPL-3.0-only (track); CC-BY-3.0 (render media); CC-BY-4.0 (render media); CC-BY-SA-3.0 WITH CC-BY-3.0 AND CC0 COMPONENTS (render media); CC0-1.0 (render media); CUSTOM-PERMISSIVE-FREE-USE (render media); GPL-3.0-only (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`stuntrally_for13_castle_sanitized`	`share-alike`	2	20.0	train	GPL-3.0-only (track); CC-BY-3.0 (render media); CC-BY-4.0 (render media); CC0-1.0 (render media); CUSTOM-PERMISSIVE-FREE-USE (render media); GPL-3.0-only (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`stuntrally_for18_mountcaro_sanitized`	`share-alike`	1	5.0	train	GPL-3.0-only (track); CC-BY-3.0 (render media); CC-BY-4.0 (render media); CC0-1.0 (render media); GPL-3.0-only (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`stuntrally_for19_crolkarstmntn_sanitized`	`share-alike`	2	20.0	train	GPL-3.0-only (track); CC-BY-3.0 (render media); CC-BY-4.0 (render media); CC-BY-SA-3.0 WITH CC-BY-3.0 AND CC0 COMPONENTS (render media); CC0-1.0 (render media); CUSTOM-PERMISSIVE-FREE-USE (render media); GPL-3.0-only (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`stuntrally_grc10_beachcity_sanitized`	`share-alike`	1	1.0	train	GPL-3.0-only (track); CC-BY-3.0 (render media); CC-BY-4.0 (render media); CC-BY-SA-3.0 (render media); CC-BY-SA-3.0 WITH CC-BY-3.0 AND CC0 COMPONENTS (render media); CC0-1.0 (render media); GPL-3.0-only (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`stuntrally_grc12_kapaba_rivers_bright_open_bridge_collision_runtime_v3`	`share-alike`	1	15.0	train	GPL-3.0-only (track); CC-BY-3.0 (render media); CC-BY-4.0 (render media); CC-BY-SA-3.0 WITH CC-BY-3.0 AND CC0 COMPONENTS (render media); CC0-1.0 (render media); GPL-3.0-only (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`stuntrally_grc14_squazelo_lakes_sanitized_collision_runtime_v2`	`share-alike`	1	15.0	train	GPL-3.0-only (track); CC-BY-4.0 (render media); CC-BY-SA-3.0 WITH CC-BY-3.0 AND CC0 COMPONENTS (render media); CC0-1.0 (render media); GPL-3.0-only (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`stuntrally_grc8_slopecity_sanitized`	`share-alike`	3	17.0	train	GPL-3.0-only (track); CC-BY-4.0 (render media); CC-BY-SA-3.0 (render media); CC-BY-SA-3.0 WITH CC-BY-3.0 AND CC0 COMPONENTS (render media); CC0-1.0 (render media); GPL-3.0-only (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`stuntrally_grc9_oasis_sanitized`	`share-alike`	2	18.1	train	GPL-3.0-only (track); CC-BY-3.0 (render media); CC-BY-4.0 (render media); CC-BY-SA-3.0 (render media); CC-BY-SA-3.0 WITH CC-BY-3.0 AND CC0 COMPONENTS (render media); CC0-1.0 (render media); GPL-3.0-only (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`stuntrally_isl13_beauty_sanitized`	`share-alike`	1	15.0	train	GPL-3.0-only (track); CC-BY-3.0 (render media); CC-BY-4.0 (render media); CC0-1.0 (render media); CUSTOM-PERMISSIVE-FREE-USE (render media); GPL-3.0-only (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`stuntrally_isl15_islandvillages_sanitized`	`share-alike`	1	15.0	train	GPL-3.0-only (track); CC-BY-3.0 (render media); CC-BY-SA-3.0 (render media); CC0-1.0 (render media); CUSTOM-PERMISSIVE-FREE-USE (render media); GPL-3.0-only (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`stuntrally_isl17_adapterislands_sanitized`	`share-alike`	2	16.0	train	GPL-3.0-only (track); CC-BY-3.0 (render media); CC-BY-4.0 (render media); CC-BY-SA-3.0 (render media); CC-BY-SA-3.0 WITH CC-BY-3.0 AND CC0 COMPONENTS (render media); CC0-1.0 (render media); GPL-3.0-only (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`stuntrally_isl2_sandy_sanitized`	`share-alike`	3	20.0	train	GPL-3.0-only (track); CC-BY-4.0 (render media); CC-BY-SA-3.0 (render media); CC0-1.0 (render media); GPL-3.0-only (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`stuntrally_jng11_temple_bright_sanitized_collision_runtime_v1`	`share-alike`	1	15.0	train	GPL-3.0-only (track); CC-BY-3.0 (render media); CC-BY-4.0 (render media); CC-BY-SA-3.0 (render media); CC0-1.0 (render media); GPL-3.0-only (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`stuntrally_jng21_paradise_bright_sanitized_collision_runtime_v1`	`share-alike`	1	15.0	train	GPL-3.0-only (track); CC-BY-3.0 (render media); CC-BY-4.0 (render media); CC0-1.0 (render media); GPL-3.0-only (render media); MULTIPLE-COMPATIBLE-ASSET-SOURCES (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`stuntrally_jng26_milkyway_sanitized`	`share-alike`	2	15.4	train	GPL-3.0-only (track); CC-BY-3.0 (render media); CC-BY-4.0 (render media); CC-BY-SA-3.0 (render media); CC0-1.0 (render media); CUSTOM-PERMISSIVE-FREE-USE (render media); GPL-3.0-only (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`stuntrally_wnt11_glacier_day_sanitized`	`share-alike`	1	20.0	train	GPL-3.0-only (track); CC-BY-3.0 (render media); CC-BY-4.0 (render media); CC0-1.0 (render media); GPL-3.0-only (render media); MULTIPLE-COMPATIBLE-ASSET-SOURCES (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`stuntrally_wnt20_firstsnow_sanitized`	`share-alike`	1	15.0	train	GPL-3.0-only (track); CC-BY-3.0 (render media); CC-BY-4.0 (render media); CC-BY-SA-3.0 WITH CC-BY-3.0 AND CC0 COMPONENTS (render media); CC0-1.0 (render media); GPL-3.0-only (render media); MULTIPLE-COMPATIBLE-ASSET-SOURCES (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`stuntrally_wnt6_snowyeights_day_sanitized_collision_runtime_v2`	`share-alike`	1	20.0	train	GPL-3.0-only (track); CC-BY-4.0 (render media); CC0-1.0 (render media); GPL-3.0-only (render media)	Stunt Rally 3.3 (stuntrally3 + tracks3), license-sanitized track derivative
`unvanquished_forlorn`	`share-alike`	1	1.5	train	CC-BY-SA-3.0	Unvanquished 0.56.2 official map "Forlorn"
`unvanquished_vega`	`share-alike`	1	1.5	test	CC-BY-SA-3.0	Unvanquished 0.56.2 official map "Vega Industries: Research Outpost 13"
`xonotic_afterslime`	`share-alike`	2	3.3	validation	GPL-3.0-or-later	Xonotic 0.8.6 official map "afterslime"
`xonotic_go`	`share-alike`	2	3.3	train	GPL-3.0-or-later	Xonotic 0.8.6 official map "go"
`xonotic_runningman`	`share-alike`	3	5.0	train	GPL-3.0-or-later	Xonotic 0.8.6 official map "runningman"
`xonotic_runningmanctf`	`share-alike`	4	6.7	train	GPL-3.0-or-later	Xonotic 0.8.6 official map "runningmanctf"
`xonotic_solarium`	`share-alike`	4	6.7	train	GPL-3.0-or-later	Xonotic 0.8.6 official map "solarium"
`xonotic_trident`	`share-alike`	1	1.7	test	GPL-3.0-or-later	Xonotic 0.8.6 official map "trident"
`xonotic_vorix`	`share-alike`	6	10.0	train	GPL-3.0-or-later	Xonotic 0.8.6 official map "vorix"
`emerald_square`	`non-commercial`	2	20.0	validation	CC-BY-NC-SA-3.0	NVIDIA Emerald Square v4.1 (NVIDIA ORCA)

Full terms, attribution lines and notices: LICENSES.md, ATTRIBUTION.md, licenses/, and metadata/.../license.json. Engines, runtimes and raw scene assets are not redistributed.

Splits by map

train: bistro_exterior, carla_town10hd_opt, carla_town12, carla_town13, classroom, dogwalk_winter_forest, mcguire_gallery, poly_haven_smugglers_cove_west_fort_static_cloud_v9, redeclipse_institute, redeclipse_octavus, stuntrally_atm10_lakecity_sanitized, stuntrally_aus3_canyon_sanitized_collision_runtime_v1, stuntrally_can8_redopoka_bright_sanitized_collision_runtime_v2, stuntrally_des11_persiancity_sanitized, stuntrally_des15_citywall_sanitized, stuntrally_des17_nilecity_sanitized, stuntrally_fin1_lakes_sanitized, stuntrally_for13_castle_sanitized, stuntrally_for18_mountcaro_sanitized, stuntrally_for19_crolkarstmntn_sanitized, stuntrally_grc10_beachcity_sanitized, stuntrally_grc12_kapaba_rivers_bright_open_bridge_collision_runtime_v3, stuntrally_grc14_squazelo_lakes_sanitized_collision_runtime_v2, stuntrally_grc8_slopecity_sanitized, stuntrally_grc9_oasis_sanitized, stuntrally_isl13_beauty_sanitized, stuntrally_isl15_islandvillages_sanitized, stuntrally_isl17_adapterislands_sanitized, stuntrally_isl2_sandy_sanitized, stuntrally_jng11_temple_bright_sanitized_collision_runtime_v1, stuntrally_jng21_paradise_bright_sanitized_collision_runtime_v1, stuntrally_jng26_milkyway_sanitized, stuntrally_wnt11_glacier_day_sanitized, stuntrally_wnt20_firstsnow_sanitized, stuntrally_wnt6_snowyeights_day_sanitized_collision_runtime_v2, unvanquished_forlorn, xonotic_go, xonotic_runningman, xonotic_runningmanctf, xonotic_solarium, xonotic_vorix
validation: carla_town15, emerald_square, namaqualand, stk_candela_city, xonotic_afterslime
test: pawn_shop, unvanquished_vega, xonotic_trident

Citation

@misc{loci_ue_revisit_2026,
  title  = {LOCI: <title placeholder>},
  author = {<authors>},
  year   = {2026},
  note   = {UE-Revisit dataset v1}
}

Downloads last month: 42

Paper for sum0214/LOCI-revisit-data

LOCI: Spatial Linear Memory for Streaming World Models

Paper • 2609.40222 • Published 3 days ago • 6