yamaama
PST MugBox-v2 Teleoperation Demonstrations
Teleoperation demonstrations of the MugBox-v2 task collected using a UR5e robot with Inspire Hand, with human hand retargeting via Leap Motion and wrist pose control via inverse kinematics. The dataset investigates how to collect high-quality demonstrations from non-expert operators under three different visual feedback conditions.
Downloads12
Episodes1474
Why This Matters for Physical AI
This dataset addresses the challenge of collecting scalable high-quality demonstrations from non-expert operators through optimized visual feedback, demonstrating practical teleoperation strategies for learning dexterous manipulation skills.
Technical Profile
- Modalities
- rgbdepthproprioceptionforce_torque
- Robot Embodiments
- UR5eInspire Hand
- Action Space
- end_effector_pose + finger_active
- Environment
- simulation
- Task Types
- dexterous_manipulationpick_and_place
- Episodes
- 1474
- Data Format
- HDF5
- Annotation Types
- operator_metadatasuccess_labelsparticle_visibility
Access
Need custom rgb data?
Claru builds purpose-built datasets for simulation applications with dense human annotations and quality assurance.
Request a Sample Pack