armin10012026Multiple (BSD 3-Clause for specific assets, see provenance section)
RoboCasa Transfer Experiment Inputs
Input snapshot for OpenCabinet, CloseCabinet, and PickPlaceCounterToCabinet transfer experiments, including cached VLA features, demonstration data, and simulator assets for imitation learning research.
Downloads27
Episodes1,263 episodes (reference tasks) with 108 demonstrations for PickPlace target
Why This Matters for Physical AI
This dataset provides cached VLA features and demonstration data essential for studying transfer learning and imitation learning in complex manipulation tasks across multiple kitchen environments and robot embodiments.
Technical Profile
- Modalities
- rgblanguage
- Robot Embodiments
- ALOHAUnitree H1
- Environment
- simulationkitchen
- Task Types
- manipulationpick_and_placecabinet_openingcabinet_closingnavigation
- Episodes
- 1,263 episodes (reference tasks) with 108 demonstrations for PickPlace target
- Data Format
- Parquet, Zstandard-compressed Tar
- Annotation Types
- language_instructionsaction_labelstask_metadata
- License
- Multiple (BSD 3-Clause for specific assets, see provenance section)
Access
Need custom rgb data?
Claru builds purpose-built datasets for simulation applications with dense human annotations and quality assurance.
Request a Sample Pack