Mixture Prototype Flow Matching for Open-Set Supervised Anomaly Detection

Open-set supervised anomaly detection (OSAD) aims to identify unseen anomalies using limited anomalous supervision. However, existing prototype-based methods typically model normal data via a unimodal Gaussian prior, failing to capture inherent multi-modality and resulting in blurred decision boundaries. To address this, we propose Mixture Prototype Flow Matching (MPFM), a framework that learns a continuous transformation from normal feature distributions to a structured Gaussian mixture prototype space. Departing from traditional flow-based approaches that rely on a single velocity vector, MPFM explicitly models the velocity field as a Gaussian mixture prior where each component corresponds to a distinct normal class. This design facilitates mode-aware and semantically coherent distribution transport. Furthermore, we introduce a Mutual Information Maximization Regularizer (MIMR) to prevent prototype collapse and maximize normal-anomaly separability. Extensive experiments demonstrate that MPFM achieves state-of-the-art performance across diverse benchmarks under both single- and multi-anomaly settings.

Automated anomaly detection systems are widely used to identify irregularities such as manufacturing defects or medical abnormalities. To function effectively, these models must accurately learn what constitutes "normal" data. Currently, most approaches assume that all normal examples share a single, uniform pattern. However, real-world normal data is highly diverse and contains multiple distinct variations. When models force these diverse patterns into one broad category, their detection boundaries become blurred. Consequently, they often misclassify rare but normal instances as anomalies, while failing to detect true irregularities that fall into the ambiguous gaps between patterns. Our paper addresses this by explicitly modeling normal data as a collection of distinct sub-patterns. Instead of drawing one loose boundary, our framework learns precise, separate representations for each specific variation of normal data. This creates sharper decision boundaries, enabling the system to reliably identify both familiar anomalies and entirely unseen deviations.