An optimization method for out-of-distribution anomaly detection models

TL;DR

This paper introduces an optimization approach combining an SVM classifier and sample synthesis to reduce false alarms in out-of-distribution anomaly detection, significantly improving industrial anomaly detection performance.

Contribution

It proposes a novel post-processing and data synthesis method to enhance the accuracy of unsupervised anomaly detection models in industrial settings.

Findings

Improved detection accuracy at image and pixel levels.

Effective reduction of false alarms in industrial applications.

Enhanced performance of segmentation models with the proposed method.

Abstract

Frequent false alarms impede the promotion of unsupervised anomaly detection algorithms in industrial applications. Potential characteristics of false alarms depending on the trained detector are revealed by investigating density probability distributions of prediction scores in the out-of-distribution anomaly detection tasks. An SVM-based classifier is exploited as a post-processing module to identify false alarms from the anomaly map at the object level. Besides, a sample synthesis strategy is devised to incorporate fuzzy prior knowledge on the specific application in the anomaly-free training dataset. Experimental results illustrate that the proposed method comprehensively improves the performances of two segmentation models at both image and pixel levels on two industrial applications.