Uninformed Students: Student-Teacher Anomaly Detection with Discriminative Latent Embeddings

TL;DR

This paper presents a student-teacher framework for unsupervised anomaly detection and segmentation in high-resolution images, leveraging the discrepancy between student and teacher networks trained on natural image patches.

Contribution

It introduces a novel student-teacher approach that detects anomalies without prior annotations by comparing student and teacher outputs and utilizing student uncertainty as an anomaly score.

Findings

Outperforms existing unsupervised anomaly detection methods.

Effective on real-world datasets including MVTec AD.

Provides pixel-precise anomaly segmentation.

Abstract

We introduce a powerful student-teacher framework for the challenging problem of unsupervised anomaly detection and pixel-precise anomaly segmentation in high-resolution images. Student networks are trained to regress the output of a descriptive teacher network that was pretrained on a large dataset of patches from natural images. This circumvents the need for prior data annotation. Anomalies are detected when the outputs of the student networks differ from that of the teacher network. This happens when they fail to generalize outside the manifold of anomaly-free training data. The intrinsic uncertainty in the student networks is used as an additional scoring function that indicates anomalies. We compare our method to a large number of existing deep learning based methods for unsupervised anomaly detection. Our experiments demonstrate improvements over state-of-the-art methods on a number of real-world datasets, including the recently introduced MVTec Anomaly Detection dataset that was specifically designed to benchmark anomaly segmentation algorithms.