SRTube: Video-Language Pre-Training with Action-Centric Video Tube Features and Semantic Role Labeling.

Ju-Hee Lee, Je-Won Kang

Browse the full CVPR paper archive.