本文提出了一种新的基于端到端局部-全局时间依赖网络的弱监督方法，通过引入片段之间的时间依赖关系来提高视频异常事件检测的性能。这种方法结合了局部时间依赖关系和全局时间依赖关系，以更好地捕捉片段之间的关联。通过这种方式，可以有效地缓解以前方法中忽略片段之间长期和短期时间依赖关系的问题，提高检测性能。 翻译成英文

This article proposes a new weakly supervised method based on an end-to-end local-global temporal dependency network to improve the performance of video anomaly detection by introducing temporal dependency relationships between segments. This method combines local and global temporal dependency relationships to better capture the correlation between segments. Through this approach, the problem of ignoring long-term and short-term temporal dependency relationships between segments in previous methods can be effectively alleviated, and detection performance can be improved.