Bidirectional skip-frame prediction for video anomaly detection with intra-domain disparity-driven attention

Abstract

Unsupervised video anomaly detection (VAD) trains only normal events to detect anomalies by large discrepancies during testing. Expanding the discriminative boundary between normal and abnormal events is the common goal and challenge of unsupervised VAD. To address this problem, we propose a Bidirectional Skip-frame Prediction (BiSP) method, leveraging intra-domain disparities between different features. Specifically, the BiSP skips frames in the training phase to achieve the forward and backward prediction respectively, and in the testing phase, it utilizes bidirectional consecutive frames to co-predict the same intermediate frames. Furthermore, we design the variance channel attention and context spatial attention from the perspectives of movement patterns and object scales, respectively. This design maximizes of the disparity between normal and abnormal in the feature extraction and delivery with different dimensions. Extensive experiments from four benchmark datasets demonstrate the effectiveness of the proposed BiSP, which substantially outperforms state-of-the-art competing methods. Ours code is available at https://github.com/jLooo/BiSP.