DexGraspNet
PKU-EPIC
Synthetic dexterous grasp configurations generated for a Shadow Hand and validated in simulation.
What you can explore
Explore grasp synthesis, object coverage and simulated grasp evaluation.
Before you use it
Simulation validity is not proof of physical success or compatibility with another hand.
Original contributors
Ruicheng Wang and collaborators
License / access: CC BY-NC 4.0
Understand the data before downloading.
Synthetic static grasp configurations · NumPy grasp records + object meshes
SELECTED FIELD
qpos
Named configuration values
Hand translation, rotation parameters and named joints.
- Timing and units
- Not a time-series trajectory; frames per second and robot control frequency do not apply.
- Permissions
- Work and dataset: CC BY-NC 4.0.
First inspection
- Read the original access terms and download the needed assets.
- Open the official quick-example notebook in its documented environment.
- Inspect one grasp, mesh scale and hand configuration before generating or evaluating grasps.
A static grasp is not an action sequence. Use the notebook’s rotation conversion rather than treating all qpos values as joint angles.
Editorial reference · checked 2026-09-22