Exploring Diffusion Models for Unsupervised Video Anomaly Detection

TL;DR

This paper explores the use of diffusion models for unsupervised video anomaly detection, leveraging their reconstruction capabilities to identify abnormal events without annotations, and demonstrates superior performance over existing generative models.

Contribution

First application of diffusion models to video anomaly detection, analyzing parameter effects and achieving improved results in unsupervised surveillance scenarios.

Findings

Diffusion models outperform state-of-the-art generative models in VAD.

The method achieves better scores than more complex models in some cases.

Parameter analysis provides guidance for VAD using diffusion models.

Abstract

This paper investigates the performance of diffusion models for video anomaly detection (VAD) within the most challenging but also the most operational scenario in which the data annotations are not used. As being sparse, diverse, contextual, and often ambiguous, detecting abnormal events precisely is a very ambitious task. To this end, we rely only on the information-rich spatio-temporal data, and the reconstruction power of the diffusion models such that a high reconstruction error is utilized to decide the abnormality. Experiments performed on two large-scale video anomaly detection datasets demonstrate the consistent improvement of the proposed method over the state-of-the-art generative models while in some cases our method achieves better scores than the more complex models. This is the first study using a diffusion model and examining its parameters' influence to present guidance for VAD in surveillance scenarios.