huangxuan111MIT
MSU-Fold
Language-conditioned multi-step cloth folding demonstrations generated in SoftGym with RGB-D observations, pixel-level keypoints, and per-step natural language instructions for bimanual manipulation tasks.
Downloads7
Episodes100-1000 per task
Why This Matters for Physical AI
This dataset enables training of language-conditioned bimanual manipulation policies for complex cloth folding tasks, advancing multi-step reasoning and coordination for deformable object manipulation in robotics.
Technical Profile
- Modalities
- rgbdepthlanguage
- Robot Embodiments
- bimanual_manipulator
- Action Space
- end_effector_delta
- Environment
- simulation
- Task Types
- cloth_foldingmanipulation
- Episodes
- 100-1000 per task
- Data Format
- pickle
- Annotation Types
- language_instructionsaction_labelskeypoint_annotations
- License
- MIT
Access
Need custom rgb data?
Claru builds purpose-built datasets for simulation applications with dense human annotations and quality assurance.
Request a Sample Pack