ResAD: A Simple Framework for Class Generalizable Anomaly Detection

TL;DR

This work proposes a simple but effective framework (called ResAD) that can be directly applied to detect anomalies in new classes by learning the residual feature distribution rather than the initial feature distribution, which can significantly reduce feature variations.

Abstract

This paper explores the problem of class-generalizable anomaly detection, where the objective is to train one unified AD model that can generalize to detect anomalies in diverse classes from different domains without any retraining or fine-tuning on the target data. Because normal feature representations vary significantly across classes, this will cause the widely studied one-for-one AD models to be poorly classgeneralizable (i.e., performance drops dramatically when used for new classes). In this work, we propose a simple but effective framework (called ResAD) that can be directly applied to detect anomalies in new classes. Our main insight is to learn the residual feature distribution rather than the initial feature distribution. In this way, we can significantly reduce feature variations. Even in new classes, the distribution of normal residual features would not remarkably shift from the learned distribution. Therefore, the learned model can be directly adapted to new classes. ResAD consists of three components: (1) a Feature Converter that converts initial features into residual features; (2) a simple and shallow Feature Constraintor that constrains normal residual features into a spatial hypersphere for further reducing feature variations and maintaining consistency in feature scales among different classes; (3) a Feature Distribution Estimator that estimates the normal residual feature distribution, anomalies can be recognized as out-of-distribution. Despite the simplicity, ResAD can achieve remarkable anomaly detection results when directly used in new classes. The code is available at https://github.com/xcyao00/ResAD.

Figures (5)

• Figure 1: (a): Intuitive illustration of class-generalizable anomaly detection (b): Conceptual illustration of residual features. The residual feature space has fewer variations compared to the initial feature space. The decision boundary of the residual feature distribution can more effectively distinguish anomalies in new classes, rather than treating features of new classes as anomalies.
• Figure 2: Framework overview. Note that the training samples belong to different classes. First, few-shot normal reference samples are fed into a pre-trained Feature Extractor to obtain normal reference features. Each initial feature will match the nearest normal reference feature and subtract it to form the residual feature. Then, a Feature Constraintor is utilized to transform the normal residual features into a constrained spatial hypersphere. Finally, we employ a normalizing flow model as the Feature Distribution Estimator to learn and estimate the residual feature distribution.
• Figure 3: Feature t-SNE visualization. (a) In the initial feature space, the features from different classes are significantly different. (b) In the residual feature space, even the residual feature distribution of unknown classes would not remarkably shift from the known distribution. Note that in (a) and (b), we only show normal residual features and use different colors to represent different classes. (c) The initial residual features. (d) The residual features after the Feature Constraintor.
• Figure 4: Qualitative results. The anomaly score maps are generated under the "VisA to MVTecAD" case.
• Figure 5: Additional qualitative results on MVTecAD.