Blien2002
ShakeBench
A collection of seven Franka Panda manipulation datasets with 2,400 episodes and 830,175 frames covering diverse tasks including pick-and-place, pushing, stacking, and panel operation.
Downloads144
Episodes2400
Why This Matters for Physical AI
ShakeBench provides a large-scale, task-diverse manipulation dataset in simulation with consistent action and state representations that enables training and evaluation of robotic manipulation policies across multiple manipulation skills.
Technical Profile
- Modalities
- rgbproprioception
- Robot Embodiments
- Franka Panda
- Action Space
- end_effector_delta
- Environment
- simulation
- Task Types
- pick_and_placepushingstackingmanipulationpanel_operationgrasping
- Episodes
- 2400
- Data Format
- LeRobot
- Annotation Types
- reward_labelsaction_labelstask_labelssuccess_labels
Access
Need custom rgb data?
Claru builds purpose-built datasets for simulation applications with dense human annotations and quality assurance.
Request a Sample Pack