glory-hyeok2026CC-BY-4.0
ActionNet subset100 with ground-truth depth
A 100-episode subset of Fourier Intelligence's ActionNet dataset with preserved ground-truth depth, corrected per-frame state alignment, and pre-contact filtering verdicts for evaluating action-video consistency.
Downloads68
Episodes100
Why This Matters for Physical AI
This dataset provides corrected, high-fidelity sensor data with validated action-observation consistency, enabling robust training and evaluation of embodied AI policies on manipulation tasks.
Technical Profile
- Modalities
- rgbdepthproprioception
- Robot Embodiments
- Fourier GR1-T1
- Action Space
- joint_positions
- Environment
- lab
- Task Types
- manipulationgrasping
- Episodes
- 100
- Data Format
- HDF5
- Annotation Types
- action_labelssegmentationdepth_ground_truthcontact_filtering
- License
- CC-BY-4.0
Access
Need custom rgb data?
Claru builds purpose-built datasets for lab applications with dense human annotations and quality assurance.
Request a Sample Pack