TransAnomaly: Video Anomaly Detection Using Video Vision Transformer

Video anomaly detection is challenging because abnormal events are unbounded, rare, equivocal, irregular in real scenes. In recent years, transformers have demonstrated powerful modelling abilities for sequence data. Thus, we attempt to apply video detection. this paper, propose a prediction-based approach named TransAnomaly. Our model combines the U-Net and Vision Transformer (ViViT) capture richer temporal information more global contexts. To make full use of ViViT prediction, modified it capable prediction. Experiments on benchmark datasets show that addition transformer module improves performance. addition, calculate regularity scores with sliding windows evaluate impact different window sizes strides. With proper settings, our outperforms other state-of-the-art approaches. Furthermore, can perform localization by tracking location patches lower scores.