Supervised Anomaly Detection for Complex Industrial Images

TL;DR

This paper introduces a new real-world industrial dataset with challenging defects and a novel segmentation-based anomaly detection method that outperforms existing approaches in industrial image inspection tasks.

Contribution

The paper presents the Valeo Anomaly Dataset (VAD) with real defects and a new Segmentation-based Anomaly Detector (SegAD) that combines segmentation maps and supervised classification for improved anomaly detection.

Findings

SegAD achieves +2.1% AUROC on VAD

SegAD outperforms previous methods on VisA dataset

The VAD dataset includes 5000 images with 2000 defect instances

Abstract

Automating visual inspection in industrial production lines is essential for increasing product quality across various industries. Anomaly detection (AD) methods serve as robust tools for this purpose. However, existing public datasets primarily consist of images without anomalies, limiting the practical application of AD methods in production settings. To address this challenge, we present (1) the Valeo Anomaly Dataset (VAD), a novel real-world industrial dataset comprising 5000 images, including 2000 instances of challenging real defects across more than 20 subclasses. Acknowledging that traditional AD methods struggle with this dataset, we introduce (2) Segmentation-based Anomaly Detector (SegAD). First, SegAD leverages anomaly maps as well as segmentation maps to compute local statistics. Next, SegAD uses these statistics and an optional supervised classifier score as input features for a Boosted Random Forest (BRF) classifier, yielding the final anomaly score. Our SegAD achieves state-of-the-art performance on both VAD (+2.1% AUROC) and the VisA dataset (+0.4% AUROC). The code and the models are publicly available.