Toward Long-Tailed Online Anomaly Detection through Class-Agnostic Concepts

TL;DR

This paper introduces a class-agnostic framework for long-tailed online anomaly detection, outperforming existing methods in both offline and online settings across industrial and medical domains.

Contribution

It proposes a novel class-agnostic approach for long-tailed anomaly detection and adapts it to online learning, addressing the limitations of class-aware offline methods.

Findings

Outperforms SOTA in offline LTAD with +4.63% image-AUROC on MVTec

Achieves +0.53% image-AUROC in challenging online LTAD setting

Provides a new benchmark for long-tailed online anomaly detection

Abstract

Anomaly detection (AD) identifies the defect regions of a given image. Recent works have studied AD, focusing on learning AD without abnormal images, with long-tailed distributed training data, and using a unified model for all classes. In addition, online AD learning has also been explored. In this work, we expand in both directions to a realistic setting by considering the novel task of long-tailed online AD (LTOAD). We first identified that the offline state-of-the-art LTAD methods cannot be directly applied to the online setting. Specifically, LTAD is class-aware, requiring class labels that are not available in the online setting. To address this challenge, we propose a class-agnostic framework for LTAD and then adapt it to our online learning setting. Our method outperforms the SOTA baselines in most offline LTAD settings, including both the industrial manufacturing and the medical domain. In particular, we observe +4.63% image-AUROC on MVTec even compared to methods that have access to class labels and the number of classes. In the most challenging long-tailed online setting, we achieve +0.53% image-AUROC compared to baselines. Our LTOAD benchmark is released here: https://doi.org/10.5281/zenodo.16283852 .