Abstract: In video understanding, action spotting consists in temporally localizing human-induced events annotated with single timestamps. In this paper, we propose a novel loss function that ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback