Attention-based residual autoencoder for video anomaly detection

Abstract Automatic anomaly detection is a crucial task in video surveillance system intensively used for public safety and others. The present adopts spatial branch temporal unified network that exploits both information effectively. has residual autoencoder architecture, consisting of deep convolutional neural network-based encoder multi-stage channel attention-based decoder, trained an unsupervised manner. shift method exploiting the feature, whereas contextual dependency extracted by attention modules. System performance evaluated using three standard benchmark datasets. Result suggests our outperforms state-of-the-art methods, achieving 97.4% UCSD Ped2, 86.7% CUHK Avenue, 73.6% ShanghaiTech dataset term Area Under Curve, respectively.