BayPrAnoMeta: Bayesian Proto-MAML for Few-Shot Industrial Image Anomaly Detection

2026-01-27

AI Summary

This work addresses the challenge of few-shot learning in industrial image anomaly detection, where defect samples are scarce and class imbalance is extreme. The authors propose BayPrAnoMeta, a Bayesian meta-learning framework that extends Proto-MAML by replacing deterministic prototypes with task-specific probabilistic prototypes. By incorporating a Normal-Inverse-Wishart prior, the method derives a Student-t posterior predictive distribution, enabling uncertainty-aware anomaly scoring with heavy-tailed characteristics. Furthermore, the approach integrates supervised contrastive regularization and is embedded within a federated meta-learning architecture to support distributed deployment. Evaluated under few-shot settings on the MVTec AD benchmark, BayPrAnoMeta achieves significantly higher AUROC than baseline methods including MAML, Proto-MAML, and PatchCore, demonstrating its effectiveness and novelty.

Technology Category

• Machine Learning: Multi-instance/Multi-view Learning
• Computer Vision: Adversarial Attacks & Robustness
• Data Mining & Knowledge Management: Anomaly/Outlier Detection

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Industrial image anomaly detection is a challenging problem owing to extreme class imbalance and the scarcity of labeled defective samples, particularly in few-shot settings. We propose BayPrAnoMeta, a Bayesian generalization of Proto-MAML for few-shot industrial image anomaly detection. Unlike existing Proto-MAML approaches that rely on deterministic class prototypes and distance-based adaptation, BayPrAnoMeta replaces prototypes with task-specific probabilistic normality models and performs inner-loop adaptation via a Bayesian posterior predictive likelihood. We model normal support embeddings with a Normal-Inverse-Wishart (NIW) prior, producing a Student-$t$ predictive distribution that enables uncertainty-aware, heavy-tailed anomaly scoring and is essential for robustness in extreme few-shot settings. We further extend BayPrAnoMeta to a federated meta-learning framework with supervised contrastive regularization for heterogeneous industrial clients and prove convergence to stationary points of the resulting nonconvex objective. Experiments on the MVTec AD benchmark demonstrate consistent and significant AUROC improvements over MAML, Proto-MAML, and PatchCore-based methods in few-shot anomaly detection settings.

Problem

• industrial image anomaly detection
• few-shot learning
• class imbalance
• labeled data scarcity
• anomaly detection

Innovation

• Bayesian meta-learning
• few-shot anomaly detection
• probabilistic prototypes
• federated learning
• uncertainty-aware scoring