Convex Combination Consistency between Neighbors for Weakly-supervised Action Localization

arXiv:2205.00400v3 Announce Type: replace
Abstract: Weakly-supervised temporal action localization (WTAL) intends to detect action instances with only weak supervision, e.g., video-level labels. The current~\textit{de facto} pipeline locates action instances by thresholding and grouping continuous high-score regions on temporal class activation sequences. In this route, the capacity of the model to recognize the relationships between adjacent snippets is of vital importance which determines the quality of the action boundaries. However, it is error-prone since the variations between adjacent snippets are typically …

arxiv combination cs.cv localization neighbors type weakly-supervised