Pseudo Anomalies Are All You Need: Diffusion-Based Generation for Weakly-Supervised Video Anomaly Detection

2025-12-07

AI Summary

Video anomaly detection (VAD) faces significant challenges due to the scarcity and high acquisition cost of authentic abnormal video samples. To address this, we propose PA-VAD, the first weakly supervised VAD framework that operates entirely without real abnormal videos. Leveraging only a small set of normal images, PA-VAD employs CLIP to select initial images and jointly optimizes textual prompts with a vision-language model to guide a video diffusion model in synthesizing high-fidelity, scene-consistent pseudo-anomalous videos. Additionally, a domain-alignment regularization module is introduced to suppress excessive spatiotemporal perturbations during generation. Evaluated on ShanghaiTech and UCF-Crime, PA-VAD achieves AUC scores of 98.2% and 82.5%, respectively—substantially outperforming all existing methods requiring real anomalies as well as state-of-the-art unsupervised approaches (e.g., UVAD).

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Adversarial Learning & Robustness
• Multiagent Systems: Adversarial Agents

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Deploying video anomaly detection in practice is hampered by the scarcity and collection cost of real abnormal footage. We address this by training without any real abnormal videos while evaluating under the standard weakly supervised split, and we introduce PA-VAD, a generation-driven approach that learns a detector from synthesized pseudo-abnormal videos paired with real normal videos, using only a small set of real normal images to drive synthesis. For synthesis, we select class-relevant initial images with CLIP and refine textual prompts with a vision-language model to improve fidelity and scene consistency before invoking a video diffusion model. For training, we mitigate excessive spatiotemporal magnitude in synthesized anomalies by an domain-aligned regularized module that combines domain alignment and memory usage-aware updates. Extensive experiments show that our approach reaches 98.2% on ShanghaiTech and 82.5% on UCF-Crime, surpassing the strongest real-abnormal method on ShanghaiTech by +0.6% and outperforming the UVAD state-of-the-art on UCF-Crime by +1.9%. The results demonstrate that high-accuracy anomaly detection can be obtained without collecting real anomalies, providing a practical path toward scalable deployment.

Problem

• Detects video anomalies without real abnormal footage
• Generates pseudo-abnormal videos using diffusion models
• Improves detection accuracy with domain-aligned regularization

Innovation

• Generates pseudo-abnormal videos using diffusion models
• Refines prompts with vision-language models for fidelity
• Mitigates excessive anomalies via domain-aligned regularization