ACT

Learning fine-grained bimanual manipulation with low-cost hardware. · By Stanford

ACT trains a transformer-based CVAE to predict action chunks from images, enabling precise tasks from 50 demonstrations on the ALOHA platform.