Hugging Face's logo

Datasets:
B111ue
/
RoboFactory-5Task-RGBD-Decentralized

Modalities:
Text
Formats:
json
Size:
< 1K
Tags:
robotics
imitation-learning
decentralized-control
multi-robot
rgb-d
Libraries:
Datasets
Dask
License:
Dataset card Data Studio Files
xet
 Community
1
Dataset Viewer
Auto-converted to Parquet Duplicate
version
stringclasses
1 value
task
stringclasses
5 values
seeds
listlengths
100
100
selection_method
stringclasses
1 value
training_seed_range
listlengths
2
2
training_seed_overlap
bool
1 class
strict640x480-v2
camera_alignment
[ 1905860028, 1707521464, 1282484962, 1116633513, 1791641211, 1312103664, 1576923859, 1579168412, 1425043585, 1198934783, 1071275813, 1791357909, 1371389916, 1719794939, 1041308800, 1894003198, 1730183932, 1614753899, 1388088053, 1258102374, 1903844521, 1039627509, 12338383...
frozen random sample without replacement; Python Random(20260728)
[ 400000, 800999 ]
false
strict640x480-v2
lift_barrier
[ 1940938499, 1151242481, 1511303784, 1484628033, 1111870947, 1905246391, 1684440623, 1913249789, 1166892126, 1627598591, 1707967699, 1501024243, 1797636676, 1609044297, 1515889304, 1633435252, 1837474316, 1364032793, 1820311071, 1945955218, 1121330154, 1418884065, 17842835...
frozen random sample without replacement; Python Random(20260728)
[ 400000, 800999 ]
false
strict640x480-v2
long_pipeline_delivery
[ 1377975857, 1929227305, 1996105486, 1784630386, 1744462179, 1989278646, 1115697836, 1118794547, 1561536534, 1328134359, 1213395708, 1216691922, 1653629554, 1771709052, 1085334846, 1934002399, 1673447263, 1044836506, 1755162436, 1017039386, 1363743670, 1610515961, 17052789...
frozen random sample without replacement; Python Random(20260728)
[ 400000, 800999 ]
false
strict640x480-v2
take_photo
[ 1723817216, 1510543437, 1557902170, 1922948762, 1414812038, 1726106858, 1175975658, 1451910270, 1979870721, 1528457461, 1041889680, 1370486243, 1556098999, 1543729260, 1735061721, 1571025471, 1615058059, 1941538617, 1316812231, 1866836215, 1180413028, 1091497730, 12222220...
frozen random sample without replacement; Python Random(20260728)
[ 400000, 800999 ]
false
strict640x480-v2
three_robots_stack_cube
[ 1254439012, 1961465680, 1972434534, 1309337370, 1065629742, 1282584547, 1936019905, 1756957599, 1352940387, 1307874592, 1350557094, 1793260453, 1457453868, 1968131735, 1578961835, 1220947502, 1387131595, 1606967287, 1323827755, 1528042882, 1583240340, 1006977529, 13379436...
frozen random sample without replacement; Python Random(20260728)
[ 400000, 800999 ]
false

RoboFactory Five-Task Decentralized Wrist RGB-D

Public research corpus for reproducible Stereo-CoRE shared-policy experiments.

Contract

  • 500 successful synchronized demonstrations: 100 each of LiftBarrier (2 robots), CameraAlignment (3), ThreeRobotsStackCube (3), LongPipelineDelivery (4), and TakePhoto (4).
  • A policy stream contains only one panda_hand wrist RGB-D observation and that robot's qpos.
  • RGB is 640x480; depth is native metric depth stored in millimetres.
  • No task ID, agent ID, peer/global/right-camera image, language, or communication is a policy input.
  • HDF5 convention: action has T records, while RGB/depth/qpos have T+1 records. The final observation is the post-action state.

Reproducible training protocol

All methods use the same corpus, held-out episode split, action chunk 100, ACT posterior encoder with 4 layers, and policy decoder with 7 layers. The main Stereo-CoRE run uses hierarchical weighted item sampling: equal probability for task, then demonstration, local arm, and time index. The sampling label is never supplied to the policy.

Companion releases

The repository-level corpus_manifest.json and per-task audit/manifest files describe the public payload. Users remain responsible for complying with RoboFactory and simulator asset licenses.
Downloads last month
160
Number of rows:
5
Total file size:
180 GB

Models trained or fine-tuned on B111ue/RoboFactory-5Task-RGBD-Decentralized