Novel Anomaly Detection Scenarios and Evaluation Metrics to Address the Ambiguity in the Definition of Normal Samples

TL;DR

This paper introduces new anomaly detection scenarios and metrics to handle ambiguity in defining normal samples, proposing RePaste for improved learning, which achieves state-of-the-art results on benchmarks.

Contribution

It presents novel scenarios and evaluation metrics for anomaly detection under ambiguous normal definitions, and introduces RePaste, a method that re-pastes high anomaly regions to enhance learning.

Findings

RePaste achieved state-of-the-art performance on MVTec AD benchmark.

RePaste maintains high AUROC and PRO scores.

The proposed scenarios address real-world ambiguity in normal sample definitions.

Abstract

In conventional anomaly detection, training data consist of only normal samples. However, in real-world scenarios, the definition of a normal sample is often ambiguous. For example, there are cases where a sample has small scratches or stains but is still acceptable for practical usage. On the other hand, higher precision is required when manufacturing equipment is upgraded. In such cases, normal samples may include small scratches, tiny dust particles, or a foreign object that we would prefer to classify as an anomaly. Such cases frequently occur in industrial settings, yet they have not been discussed until now. Thus, we propose novel scenarios and an evaluation metric to accommodate specification changes in real-world applications. Furthermore, to address the ambiguity of normal samples, we propose the RePaste, which enhances learning by re-pasting regions with high anomaly scores from the previous step into the input for the next step. On our scenarios using the MVTec AD benchmark, RePaste achieved the state-of-the-art performance with respect to the proposed evaluation metric, while maintaining high AUROC and PRO scores. Code: https://github.com/ReijiSoftmaxSaito/Scenario