Dual-Modeling Decouple Distillation for Unsupervised Anomaly Detection

TL;DR

In DMDD, a Decouple Student-Teacher Network is proposed to decouple the initial student features into normality and abnormality features, and a Multi-perception Segmentation Network is proposed to achieve focused anomaly map fusion based on multiple attention.

Abstract

Knowledge distillation based on student-teacher network is one of the mainstream solution paradigms for the challenging unsupervised Anomaly Detection task, utilizing the difference in representation capabilities of the teacher and student networks to implement anomaly localization. However, over-generalization of the student network to the teacher network may lead to negligible differences in representation capabilities of anomaly, thus affecting the detection effectiveness. Existing methods address the possible over-generalization by using differentiated students and teachers from the structural perspective or explicitly expanding distilled information from the content perspective, which inevitably result in an increased likelihood of underfitting of the student network and poor anomaly detection capabilities in anomaly center or edge. In this paper, we propose Dual-Modeling Decouple Distillation (DMDD) for the unsupervised anomaly detection. In DMDD, a Decouple Student-Teacher Network is proposed to decouple the initial student features into normality and abnormality features. We further introduce Dual-Modeling Distillation based on normal-anomaly image pairs, fitting normality features of anomalous image and the teacher features of the corresponding normal image, widening the distance between abnormality features and the teacher features in anomalous regions. Synthesizing these two distillation ideas, we achieve anomaly detection which focuses on both edge and center of anomaly. Finally, a Multi-perception Segmentation Network is proposed to achieve focused anomaly map fusion based on multiple attention. Experimental results on MVTec AD show that DMDD surpasses SOTA localization performance of previous knowledge distillation-based methods, reaching 98.85% on pixel-level AUC and 96.13% on PRO.

Figures (8)

• Figure 1: (a) Structural Solutions. (b) Content Solutions. (c) Our proposed Dual-Modeling Decouple Distillation method with dual-branch carrying out distillation with different concerns.
• Figure 2: Overview of DMDD. Left: Our proposed Decoupled Student-Teacher Network is demonstrated. First, the student network uses a dual-branch design to decouple normality features and abnormality features. Then, the decoupled features are distilled through Normality Guidance Modeling (NGM) and Abnormality Inverse Mimicking (AIM) respectively. Right: The Segmentation Network is shown, where a Multi-perception Segmentation Head is trained by the ground-truth masks of synthetic anomalies. The anomaly maps and anomaly scores are obtained directly during inference.
• Figure 3: The proportions of normal and anomalous pixels in receptive fields located at anomaly's center and edge.
• Figure 4: Pyramid Modeling Network.
• Figure 5: Foreground-aware Anomaly Synthesis, where $\beta$ and $\bigodot$ represent opacity and element-wise multiplication.