Structured Evidence Selection for Weakly Supervised Video Anomaly Detection

2026-07-11

AI Summary

This work addresses the challenge of weakly supervised video anomaly detection, where reliance solely on video-level labels hinders precise localization of anomalies in complex scenes and renders models susceptible to background distractions. To overcome these limitations, the authors propose SESAD, a Structured Evidence Selection framework that reframes anomaly detection as a structured reasoning process over clip-level visual evidence. SESAD employs semantic reorganization and contextual constraints to select discriminative evidence and introduces a lightweight geometric discrimination module to construct a dual-prototype embedding space, leveraging relative geometric relationships to mitigate semantic entanglement. The method achieves state-of-the-art AUC scores of 67.92, 97.99, and 88.46 on UBnormal, ShanghaiTech, and UCF-Crime benchmarks, respectively, demonstrating significantly enhanced robustness and accuracy while maintaining computational efficiency.

Abstract

Weakly supervised video anomaly detection relies solely on video-level labels for training, making it difficult to accurately localize anomalous events in complex scenes. In real-world videos, anomalous behaviors exhibit large variations in appearance and temporal duration, while scene appearance and action dynamics are often tightly entangled. Consequently, existing models tend to rely on scene-related statistical cues rather than true behavioral deviations, resulting in unstable detection performance. To address this challenge, we propose a Structured Evidence Selection framework (SESAD) that reformulates anomaly detection as a structured reasoning process over clip-level visual evidence. Instead of directly mapping aggregated features to anomaly scores, SESAD reorganizes clip representations into semantically structured candidate evidence and performs context-conditioned selection under scene and action constraints. This mechanism adaptively emphasizes anomaly-relevant semantics while suppressing scene interference, thereby alleviating semantic entanglement under weak supervision. Furthermore, we introduce a lightweight geometric discrimination module that constructs a dual-prototype structure in the embedding space, enabling anomaly decisions through relative geometric relations. Extensive experiments on UBnormal, ShanghaiTech, and UCF-Crime show that SESAD achieves 67.92, 97.99, and 88.46 AUC, respectively, while maintaining high computational efficiency and overall consistently stable anomaly discrimination.

Problem

• weakly supervised learning
• video anomaly detection
• anomaly localization
• semantic entanglement
• scene-action disentanglement

Innovation

• Structured Evidence Selection
• Weakly Supervised Anomaly Detection
• Semantic Disentanglement
• Geometric Discrimination
• Video Anomaly Localization