CSAD: Unsupervised Component Segmentation for Logical Anomaly Detection

TL;DR

This paper introduces CSAD, an unsupervised component segmentation method that uses foundation models to improve logical anomaly detection, achieving state-of-the-art results with lower latency and no manual annotations.

Contribution

It presents a novel unsupervised segmentation technique leveraging foundation models, integrated with new modules for enhanced logical anomaly detection.

Findings

Achieves 95.3% AUROC on MVTec LOCO AD dataset.

Outperforms previous state-of-the-art methods.

Offers lower latency and higher throughput.

Abstract

To improve logical anomaly detection, some previous works have integrated segmentation techniques with conventional anomaly detection methods. Although these methods are effective, they frequently lead to unsatisfactory segmentation results and require manual annotations. To address these drawbacks, we develop an unsupervised component segmentation technique that leverages foundation models to autonomously generate training labels for a lightweight segmentation network without human labeling. Integrating this new segmentation technique with our proposed Patch Histogram module and the Local-Global Student-Teacher (LGST) module, we achieve a detection AUROC of 95.3% in the MVTec LOCO AD dataset, which surpasses previous SOTA methods. Furthermore, our proposed method provides lower latency and higher throughput than most existing approaches.