Enhancing Anomaly Detection with Local-Global Temporal Dependence Network for Video Sequences

As both video sequences and anomalous events are composed of consecutive segments, there exists a temporal dependence between these segments. Therefore, simultaneously learning the local temporal dependence between adjacent segments in the video and the global temporal dependence between adjacent segments can improve the detection performance of the model. However, previous weakly supervised methods ignored the long-term and short-term temporal dependence between segments. To mitigate this issue, this paper proposes an end-to-end local-global temporal dependence network by introducing the temporal dependence between segments, which enables the model to focus more on the correlations between segments.