The nature of cognition and action

Realistic videos of human actions exhibit rich spatiotemporal structures at multiple levels of granularity: an action can always be decomposed into multiple finer-grained elements in both space and time. To capture this intuition, we propose to ...

Weakly-supervised temporal action localization (W-TAL) is to locate the boundaries of action instances and classify them in an untrimmed video, which is a challenging task due to only video-level labels during training. Existing methods mainly ...