Adapting the Hypersphere Loss Function from Anomaly Detection to Anomaly Segmentation

TL;DR

This paper improves the FCDD method for image anomaly segmentation by redesigning its loss function to better align with the Hypersphere Classifier, resulting in enhanced defect localization performance.

Contribution

It introduces a new loss function for FCDD that enhances anomaly segmentation accuracy by better mimicking the Hypersphere Classifier's behavior.

Findings

Improved segmentation accuracy on MVTec-AD dataset

Better pixel-wise supervision design enhances defect localization

Consistent performance gains over the original FCDD

Abstract

We propose an incremental improvement to Fully Convolutional Data Description (FCDD), an adaptation of the one-class classification approach from anomaly detection to image anomaly segmentation (a.k.a. anomaly localization). We analyze its original loss function and propose a substitute that better resembles its predecessor, the Hypersphere Classifier (HSC). Both are compared on the MVTec Anomaly Detection Dataset (MVTec-AD) -- training images are flawless objects/textures and the goal is to segment unseen defects -- showing that consistent improvement is achieved by better designing the pixel-wise supervision.