Towards Video Anomaly Detection from Event Streams: A Baseline and Benchmark Datasets
This work addresses the problem of video anomaly detection for surveillance and security applications by introducing event-based vision, but it is incremental as it builds on existing anomaly detection concepts with new data and hybrid methods.
The authors tackled the lack of datasets and methods for video anomaly detection using event streams by constructing synchronized event-RGB benchmarks and proposing the EWAD framework, which achieved significant improvements over existing approaches in experiments on three benchmarks.
Event-based vision, characterized by low redundancy, focus on dynamic motion, and inherent privacy-preserving properties, naturally fits the demands of video anomaly detection (VAD). However, the absence of dedicated event-stream anomaly detection datasets and effective modeling strategies has significantly hindered progress in this field. In this work, we take the first major step toward establishing event-based VAD as a unified research direction. We first construct multiple event-stream based benchmarks for video anomaly detection, featuring synchronized event and RGB recordings. Leveraging the unique properties of events, we then propose an EVent-centric spatiotemporal Video Anomaly Detection framework, namely EWAD, with three key innovations: an event density aware dynamic sampling strategy to select temporally informative segments; a density-modulated temporal modeling approach that captures contextual relations from sparse event streams; and an RGB-to-event knowledge distillation mechanism to enhance event-based representations under weak supervision. Extensive experiments on three benchmarks demonstrate that our EWAD achieves significant improvements over existing approaches, highlighting the potential and effectiveness of event-driven modeling for video anomaly detection. The benchmark datasets will be made publicly available.