policy
VINN
NYU (Lerrel Pinto Lab) · PyTorch
or hover any field below to flag it
Overview
Name
VINN
Author
NYU (Lerrel Pinto Lab)
Framework
PyTorch
License
mit
Skill type
manipulation
Evidence level
verified
Task description
Non-parametric imitation learning via k-nearest-neighbor lookup in a BYOL-learned visual embedding space. No policy training needed beyond collecting demos -- at test time, finds visually similar demo frames and averages their actions. Simple but effective baseline.
Spaces
Action space
end-effector-pose · 3-dim · 1Hz
Observation space
- type: vision
- · primary_rgb
Links
HuggingFace repo
null
Paper (arXiv)
Compatible robots
0+1 mentioned but not in catalog yetNo robots list VINN as compatible yet. Know of one? Flag it above.
Compatible environments
1Datasets that reference this policy
0No datasets reference VINN yet.