Abstract: Audio-visual event (AVE) localization aims to localize the temporal boundaries of events that contains visual and audio contents, to identify event categories in unconstrained videos.