Diffusion Policy
real-stanford
A visuomotor policy-learning framework that generates action sequences through diffusion.
What you can explore
Compare documented training, evaluation and dataset workflows as a baseline for manipulation learning.
Before you use it
Tasks and robot action spaces vary. Adapting the framework to a new dexterous hand requires suitable demonstrations and a validated control interface.
Original contributors
Cheng Chi and Diffusion Policy collaborators
License / access: Code: MIT · check external data, models and hardware terms separately
Keep exploring
ACT / Action Chunking with Transformers
An imitation-learning implementation that predicts chunks of robot actions from demonstrations.
robomimic
A framework for training and evaluating manipulation policies from demonstrations.
DexUMI
A wearable interface that records dexterous demonstrations and adapts their visual appearance for robot learning.
Editorial reference · checked 2026-09-22