liorbenhorin-nv2026apache-2.0
SO-101 Vials Domain-Randomized Grasp Dataset
A dataset containing 260 episodes of robotic grasping tasks on vials, combining 250 domain-randomized simulation episodes and 10 real robot episodes collected with the SO-101 follower robot. The dataset includes wrist and external RGB camera observations at 640×480 resolution at 30 FPS.
Downloads20
Episodes260
Why This Matters for Physical AI
This sim-to-real dataset demonstrates domain randomization techniques for closing the reality gap in robotic grasping, providing a benchmark for evaluating transfer learning from simulation to real robot control.
Technical Profile
- Modalities
- rgbproprioception
- Robot Embodiments
- SO-101 follower
- Action Space
- joint_positions
- Environment
- simulationlab
- Task Types
- graspingmanipulation
- Episodes
- 260
- Data Format
- LeRobot
- License
- apache-2.0
Access
Need custom rgb data?
Claru builds purpose-built datasets for simulation applications with dense human annotations and quality assurance.
Request a Sample Pack