Towards Total Recall in Industrial Anomaly Detection

TLDR

Detecting defective parts is essential in large‑scale manufacturing, and state‑of‑the‑art methods combine ImageNet embeddings with outlier detection to handle many tasks automatically. The study aims to solve the cold‑start problem of anomaly detection by training a model solely on nominal images. PatchCore builds a memory bank of representative nominal patch features and applies outlier detection to identify anomalies. PatchCore achieves state‑of‑the‑art performance, attaining 99.6 % AUROC on MVTec AD and competitive results on other datasets, even with few samples. Code is available at github.com/amazon-research/patchcore-inspection.

Abstract

Being able to spot defective parts is a critical component in large-scale industrial manufacturing. A particular challenge that we address in this work is the cold-start problem: fit a model using nominal (non-defective) example images only. While handcrafted solutions per class are possible, the goal is to build systems that work well simultaneously on many different tasks automatically. The best peforming approaches combine embeddings from ImageNet models with an outlier detection model. In this paper, we extend on this line of work and propose PatchCore, which uses a maximally representative memory bank of nominal patch-features. PatchCore offers competitive inference times while achieving state-of-the-art performance for both detection and localization. On the challenging, widely used MVTec AD benchmark PatchCore achieves an image-level anomaly detection AUROC score of up to 99.6%, more than halving the error compared to the next best competitor. We further report competitive results on two additional datasets and also find competitive results in the few samples regime. Code: github.com/amazon-research/patchcore-inspection.