VicTR: Video-conditioned Text Representations for Activity Recognition.

Kumara Kahatapitiya, Anurag Arnab, Arsha Nagrani, Michael S. Ryoo

Browse the full CVPR paper archive.