Context-structured Video Anomaly Detection with Large Vision-Language Models

2026-07-21

AI Summary

This work addresses the challenges in video anomaly detection posed by the scarcity of anomalous samples, high annotation costs, and the tendency of existing training-free methods to overlook context specificity. To this end, the authors propose CSI-VAD, a training-free approach that innovatively decomposes video content into three contextual dimensions—scene, object, and temporal—and leverages large vision-language models to perform independent reasoning along each branch. By relying solely on visual cues, CSI-VAD achieves context-aware anomaly detection without requiring predefined textual prompts or dataset-specific fine-tuning. Experimental results on UCF-Crime and UBnormal demonstrate that CSI-VAD significantly outperforms holistic baselines and attains performance competitive with state-of-the-art methods.

Abstract

Training video anomaly detectors is challenging due to the difficulty and cost of annotating diverse and rare abnormal events. Although recent large vision-language models enable training-free inference, existing approaches mostly rely on holistic inference over sampled video and may miss context-specific anomaly cues. In this paper, we present CSI-VAD, a training-free video anomaly detector that identifies abnormal events across diverse contexts. The key idea is to decompose each video into three distinct contexts (environment, objects, time) and perform context-specific inference in separate branches. Because we ground anomaly judgments solely in context-specific visual cues, we do not require predefined text prompts describing abnormal events or dataset-specific tuning. Experiments on UCF-Crime and UBnormal show that CSI-VAD consistently improves over the direct holistic baseline and achieves competitive performance against existing methods, showing the advantage of structured context decomposition for training-free video anomaly detection.

Problem

• video anomaly detection
• context decomposition
• training-free
• large vision-language models
• abnormal event detection

Innovation

• context decomposition
• training-free anomaly detection
• vision-language models
• video anomaly detection
• context-specific inference