Abnormal event detection by a weakly supervised temporal attention network

Abnormal event detection aims to automatically identify unusual events that do not comply with expectation. Recently, many methods have been proposed obtain the temporal locations of abnormal under various determined thresholds. However, specific categories are mostly neglect, which important help in monitoring agents make decisions. In this study, a Temporal Attention Network (TANet) is capture both and weakly supervised manner. The TANet learns anomaly score category for each video segment only video-level labels. An recognition module exploited predict scores while attention learn value. Finally, categories, three constraints considered: constraint, separation constraint smoothness constraint. Experiments on University Central Florida Crime dataset demonstrate effectiveness method.