lance-format2024apache-2.0
LeRobot PushT (Lance Format)
Lance-formatted version of the PushT benchmark from the Diffusion Policy paper, organized as three tables (frames, videos, episodes) for flexible training and analysis of robotic manipulation tasks.
Downloads11
Why This Matters for Physical AI
PushT is a canonical benchmark for visuomotor policy learning that enables researchers to develop and evaluate diffusion-based and other approaches for robotic manipulation through a standardized, well-structured dataset.
Technical Profile
- Modalities
- rgbvideo
- Environment
- lab
- Task Types
- manipulationpushing
- Data Format
- lance
- License
- apache-2.0
Access
Need custom rgb data?
Claru builds purpose-built datasets for lab applications with dense human annotations and quality assurance.
Request a Sample Pack