HS-Kempten
AGCO Korb entladen — Single-Cam Sim+Real Merge v2
Combined training dataset of simulated and real G1 teleoperation recordings for the basket unloading task, unified to a common joint-space schema with a single head camera. Designed to merge both data sources in a GR00T finetune despite their original differences in action spaces, state spaces, and camera setups.
Downloads0
Episodes71
Why This Matters for Physical AI
This dataset demonstrates sim-to-real transfer for bimanual humanoid manipulation by merging simulated and real-world teleoperation data in a unified schema, enabling training of embodied foundation models like GR00T on diverse data sources.
Technical Profile
- Modalities
- rgbproprioception
- Robot Embodiments
- Unitree G1
- Action Space
- mixed (waist: absolute, left_arm: relative, right_arm: relative, left_hand: absolute, right_hand: absolute, navigate_command: absolute, base_height_command: absolute)
- Environment
- labsimulation
- Task Types
- pick_and_placemanipulation
- Episodes
- 71
- Data Format
- RLDS
- Annotation Types
- action_labels
Access
Need custom rgb data?
Claru builds purpose-built datasets for lab applications with dense human annotations and quality assurance.
Request a Sample Pack