Norimichi Ukita

2D motion inputs can rival 3D data in performance, unlocking new possibilities for real-world applications of Motion Language Models without the need for extensive retraining.

DINO can be repurposed to learn group activity features without group activity annotations by predicting person flow and group-relevant object locations.

Detecting who is paying attention is now more accurate, thanks to a new method that jointly infers group membership and shared attention.