StructCore: Structure-Aware Image-Level Scoring for Training-Free Unsupervised Anomaly Detection

2026-02-18

AI Summary

This work addresses the limitation of conventional max-pooling–based image-level anomaly scoring methods, which often yield overlapping scores for normal and anomalous samples due to their neglect of spatial distribution and structural information in anomaly evidence. To overcome this, the authors propose StructCore, a training-free, structure-aware image-level scoring method that enhances discriminability without altering pixel-level anomaly localization. StructCore models the spatial layout of anomaly score maps using low-dimensional structural descriptors and incorporates diagonal Mahalanobis distance calibration derived from defect-free samples. As the first image-level scoring framework to integrate structural awareness, StructCore achieves state-of-the-art unsupervised performance, attaining image-level AUROC scores of 99.6% on MVTec AD and 98.4% on VisA.

Technology Category

• Machine Learning: Structured Learning
• Computer Vision: Adversarial Attacks & Robustness
• Data Mining & Knowledge Management: Anomaly/Outlier Detection

Application Category

• Web Mining and Content Analysis: Bridging structured and unstructured data
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Max pooling is the de facto standard for converting anomaly score maps into image-level decisions in memory-bank-based unsupervised anomaly detection (UAD). However, because it relies on a single extreme response, it discards most information about how anomaly evidence is distributed and structured across the image, often causing normal and anomalous scores to overlap. We propose StructCore, a training-free, structure-aware image-level scoring method that goes beyond max pooling. Given an anomaly score map, StructCore computes a low-dimensional structural descriptor phi(S) that captures distributional and spatial characteristics, and refines image-level scoring via a diagonal Mahalanobis calibration estimated from train-good samples, without modifying pixel-level localization. StructCore achieves image-level AUROC scores of 99.6% on MVTec AD and 98.4% on VisA, demonstrating robust image-level anomaly detection by exploiting structural signatures missed by max pooling.

Problem

• unsupervised anomaly detection
• image-level scoring
• max pooling
• anomaly score maps
• structural information

Innovation

• structure-aware
• training-free
• unsupervised anomaly detection
• Mahalanobis calibration
• anomaly score map