geonmin-kimapache-2.0
SO101-lv4-3color-cube-mat-to-mat-no-human-reset-alldir-v1-target-id-leadin-pos
A derived dataset of 426 robot manipulation episodes with per-episode annotations for target cube identity, start position, lead-in pose, and reach-complete pose for a pick-and-place task with colored cubes.
Downloads0
Episodes426
Why This Matters for Physical AI
This annotated dataset enables training and evaluation of robotic manipulation policies with precise grasp detection, reach-complete pose identification, and target object localization for real-world pick-and-place tasks.
Technical Profile
- Modalities
- rgbproprioception
- Robot Embodiments
- SO101
- Action Space
- joint_positions
- Environment
- lab
- Task Types
- pick_and_placemanipulationgrasping
- Episodes
- 426
- Data Format
- parquet
- Annotation Types
- language_instructionsaction_labelssegmentationbounding_boxes
- 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