Multi-scale Feature Imitation for Unsupervised Anomaly Localization

TL;DR

This paper introduces a multi-scale feature imitation approach with a teacher-student network for unsupervised anomaly localization, effectively handling multiple anomaly types and improving detection accuracy on industrial datasets.

Contribution

It proposes a novel multi-scale feature imitation network with a gradient-based importance search to enhance unsupervised anomaly localization performance.

Findings

Outperforms existing feature modeling methods on industrial datasets

Multi-scale strategy significantly improves detection accuracy

Network simplification via gradient descent reduces complexity

Abstract

The unsupervised anomaly localization task faces the challenge of missing anomaly sample training, detecting multiple types of anomalies, and dealing with the proportion of the area of multiple anomalies. A separate teacher-student feature imitation network structure and a multi-scale processing strategy combining an image and feature pyramid are proposed to solve these problems. A network module importance search method based on gradient descent optimization is proposed to simplify the network structure. The experimental results show that the proposed algorithm performs better than the feature modeling anomaly localization method on the real industrial product detection dataset in the same period. The multi-scale strategy can effectively improve the effect compared with the benchmark method.