flex-pi
RoboTwin 2.0 — 3D (RGB + Depth)
Bimanual manipulation dataset from the RoboTwin 2.0 simulator with 27,500 episodes across 50 tasks, featuring synchronized RGB and ground-truth depth from three camera views in LeRobot v2.1 format.
Downloads0
Episodes27,500
Why This Matters for Physical AI
Provides large-scale, high-fidelity bimanual manipulation data with precise depth ground truth and language annotations, enabling development and evaluation of vision-based robot learning algorithms with multimodal inputs.
Technical Profile
- Modalities
- rgbdepthproprioceptionlanguage
- Robot Embodiments
- ALOHA-style bimanual
- Action Space
- joint_positions
- Environment
- simulation
- Task Types
- manipulationbimanual_manipulationpick_and_placegraspingobject_rearrangement
- Episodes
- 27,500
- Data Format
- LeRobot v2.1
- Annotation Types
- language_instructionscamera_intrinsicsdepth_ground_truth
Access
Need custom rgb data?
Claru builds purpose-built datasets for simulation applications with dense human annotations and quality assurance.
Request a Sample Pack