REB: Reducing Biases in Representation for Industrial Anomaly Detection

TL;DR

REB introduces a novel approach to industrial anomaly detection by reducing domain bias and local density bias in feature representations, significantly improving detection accuracy across multiple datasets.

Contribution

The paper proposes REB, a method that combines domain bias reduction with a local-density KNN, and introduces DefectMaker for synthetic defect generation, enhancing anomaly detection performance.

Findings

Achieves 99.5% Im.AUROC on MVTec AD with small backbones

Attains 88.8% Im.AUROC on MVTec LOCO AD

Reaches 96.0% Im.AUROC on BTAD dataset

Abstract

Existing representation-based methods usually conduct industrial anomaly detection in two stages: obtain feature representations with a pre-trained model and perform distance measures for anomaly detection. Among them, K-nearest neighbor (KNN) retrieval-based anomaly detection methods show promising results. However, the features are not fully exploited as these methods ignore domain bias of pre-trained models and the difference of local density in feature space, which limits the detection performance. In this paper, we propose Reducing Biases (REB) in representation by considering the domain bias and building a self-supervised learning task for better domain adaption with a defect generation strategy (DefectMaker) that ensures a strong diversity in the synthetic defects. Additionally, we propose a local-density KNN (LDKNN) to reduce the local density bias in the feature space and obtain effective anomaly detection. The proposed REB method achieves a promising result of 99.5\% Im.AUROC on the widely used MVTec AD, with smaller backbone networks such as Vgg11 and Resnet18. The method also achieves an impressive 88.8\% Im.AUROC on the MVTec LOCO AD dataset and a remarkable 96.0\% on the BTAD dataset, outperforming other representation-based approaches. These results indicate the effectiveness and efficiency of REB for practical industrial applications. Code:https://github.com/ShuaiLYU/REB.