Learning local and global prototypes with optimal transport for unsupervised anomaly detection and localization

2025-08-18

AI Summary

This paper addresses unsupervised image anomaly detection and localization using only normal samples. Methodologically, it leverages a pre-trained encoder to extract features and jointly learns local and global prototypes; structured embedding representations are constructed via optimal transport, jointly encoding feature similarity and spatial layout. Prototype optimization explicitly models the intrinsic geometric structure of normal data, enhancing sensitivity to local inconsistencies. The key contribution is the integration of optimal transport with prototype learning, enabling structural-aware anomaly scoring through a combined feature-spatial cost. Evaluated on the MVTec AD and VisA industrial benchmarks, the method achieves state-of-the-art performance in both image-level detection and pixel-level localization, demonstrating the effectiveness and generalizability of structured representation learning.

Technology Category

• Computer Vision: Representation Learning for Vision
• Machine Learning: Structured Learning
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Bridging structured and unstructured data

Abstract

Unsupervised anomaly detection aims to detect defective parts of a sample by having access, during training, to a set of normal, i.e. defect-free, data. It has many applications in fields, such as industrial inspection or medical imaging, where acquiring labels is costly or when we want to avoid introducing biases in the type of anomalies that can be spotted. In this work, we propose a novel UAD method based on prototype learning and introduce a metric to compare a structured set of embeddings that balances a feature-based cost and a spatial-based cost. We leverage this metric to learn local and global prototypes with optimal transport from latent representations extracted with a pre-trained image encoder. We demonstrate that our approach can enforce a structural constraint when learning the prototypes, allowing to capture the underlying organization of the normal samples, thus improving the detection of incoherencies in images. Our model achieves performance that is on par with strong baselines on two reference benchmarks for anomaly detection on industrial images. The code is available at https://github.com/robintrmbtt/pradot.

Problem

• Detect defects in samples using only normal training data
• Balance feature and spatial costs for anomaly detection
• Learn structured prototypes to improve anomaly localization

Innovation

• Learning local and global prototypes
• Optimal transport for anomaly detection
• Balancing feature and spatial costs