When Anomalies Depend on Context: Learning Conditional Compatibility for Anomaly Detection

2026-01-30

AI Summary

This work addresses the contextual nature of anomalies, which existing methods often overlook by treating abnormality as an intrinsic property of objects—ignoring that the same entity (e.g., a person running) may be normal in one context (a track) but anomalous in another (a highway). To this end, we present the first systematic study of context-dependent anomaly detection, introducing a conditional compatibility learning framework that models the compatibility between foreground entities and their surrounding contexts through vision-language representations, enabling weakly supervised detection. We further construct CAAD-3K, a novel benchmark dataset that controls for object identity while varying only the contextual background, along with a context-controllable data generation strategy. Experiments demonstrate that our approach significantly outperforms existing methods on CAAD-3K and achieves state-of-the-art performance on MVTec-AD and VisA, confirming the efficacy of explicit context modeling as a complementary paradigm for structural anomaly detection.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Machine Learning: Adversarial Learning & Robustness

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Anomaly detection is often formulated under the assumption that abnormality is an intrinsic property of an observation, independent of context. This assumption breaks down in many real-world settings, where the same object or action may be normal or anomalous depending on latent contextual factors (e.g., running on a track versus on a highway). We revisit \emph{contextual anomaly detection}, classically defined as context-dependent abnormality, and operationalize it in the visual domain, where anomaly labels depend on subject--context compatibility rather than intrinsic appearance. To enable systematic study of this setting, we introduce CAAD-3K, a benchmark that isolates contextual anomalies by controlling subject identity while varying context. We further propose a conditional compatibility learning framework that leverages vision--language representations to model subject--context relationships under limited supervision. Our method substantially outperforms existing approaches on CAAD-3K and achieves state-of-the-art performance on MVTec-AD and VisA, demonstrating that modeling context dependence complements traditional structural anomaly detection. Our code and dataset will be publicly released.

Problem

• anomaly detection
• contextual anomaly
• subject-context compatibility
• visual anomaly detection
• conditional compatibility

Innovation

• contextual anomaly detection
• conditional compatibility learning
• vision-language representation
• CAAD-3K benchmark
• subject-context compatibility