Content-oriented video anomaly detection using a self-attention–based deep learning model

Video anomaly detection, which differs from traditional video analysis, is a research hotspot in the field of computer vision, attracting many researchers. Usually, abnormal events occur only in a small percentage of the video pixels and therefore, it is unnecessary to focus on all the video pixels as most of them are harmless—called "the background."

analysis anomaly anomaly detection attention computer computer sciences computer vision deep learning detection events focus pixels research researchers self-attention small video video analysis vision