Jogima-cyber2026cc-by-4.0
Coupled Local and Global World Models: G1 dataset
Real-robot data from G1 humanoid performing ego-centric grasping and lifting tasks, collected for training world models in model-based reinforcement learning. Contains 752,861 frames (20.9 hours) of RGB video, proprioception, commands, and reward labels.
Downloads32
Hours20.9
Why This Matters for Physical AI
This dataset demonstrates world model training on real humanoid robot manipulation data, enabling efficient model-based reinforcement learning for complex embodied tasks.
Technical Profile
- Modalities
- rgbproprioception
- Robot Embodiments
- G1 humanoid
- Action Space
- joint_positions
- Environment
- lab
- Task Types
- manipulationgraspinglift
- Total Hours
- 20.9
- Data Format
- HDF5
- Annotation Types
- reward_labels
- 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