Fetching the paper…

Track2Act: Predicting Point Tracks from Internet Videos enables Generalizable Robot Manipulation · Around