kaifz2025cc-by-4.0
PGND: Particle-Grid Neural Dynamics Dataset
A dataset containing 2,180 trajectory windows across six deformable-object categories collected from RGB-D videos for learning deformable object models. Includes 160 evaluation windows with matching Gaussian splats.
Downloads0
Episodes2180
Why This Matters for Physical AI
This dataset enables training of neural dynamics models for deformable object manipulation using RGB-D observations, advancing the capability of robots to interact with and model complex non-rigid objects in physical environments.
Technical Profile
- Modalities
- rgbdepthsegmentation
- Environment
- lab
- Task Types
- deformable-object-manipulation
- Episodes
- 2180
- Data Format
- tar
- Annotation Types
- segmentationobject_maskscamera_calibration
- License
- cc-by-4.0
Access
Need custom rgb data?
Claru builds purpose-built datasets for lab applications with dense human annotations and quality assurance.
Request a Sample Pack