huzheyuanapache-2.0

3lego WHOLE-EPISODE sessions: round1

Dual-arm YAM robot block-sorting dataset with 12 whole-episode sessions (216,000 frames @ 60 Hz) reconverted from raw recordings for in-context learning, featuring multi-camera observations and both cartesian and joint-space action representations.

Downloads0
Episodes12
Hours1.00

Why This Matters for Physical AI

This dataset enables in-context learning research by providing whole-episode dual-arm manipulation trajectories with latent task structure (block permutations), multi-view camera observations, and fine-grained subtask language annotations for studying how robots learn from in-episode context.

Technical Profile

Modalities
rgbproprioceptionlanguage
Robot Embodiments
dual-arm manipulator
Action Space
end_effector_delta
Environment
lab
Task Types
manipulationpick_and_placesorting
Episodes
12
Total Hours
1.00
Data Format
LeRobot
Annotation Types
language_instructionsaction_labelssubtask_segmentation
License
apache-2.0
Part of the 3lego family

Access

Need custom rgb data?

Claru builds purpose-built datasets for lab applications with dense human annotations and quality assurance.

Request a Sample Pack

Related Datasets