SwinAD: Multi-stage feature reconstruction for unsupervised industrial anomaly detection

2026-07-15

AI Summary

This work addresses the challenge of over-generalization and reduced discriminability in unsupervised anomaly detection within multi-class industrial settings, where diverse normal patterns complicate accurate modeling. To preserve reconstruction diversity while avoiding trivial mappings, the authors propose a multi-stage reconstruction-based approach leveraging a frozen pretrained Swin Transformer V2 to extract multi-scale semantic features. A dual-branch decoder with a Dropout bottleneck is designed to enable diverse reconstructions, and high-precision anomaly maps are generated by aggregating multi-scale feature discrepancies. The method achieves state-of-the-art performance across multiple benchmarks—MVTec AD, VisA, and Real-IAD—with particularly notable improvements in pixel-level average precision and F1 score on MVTec AD.

Abstract

Industrial anomaly detection aims to identify and localize defective regions without relying on exhaustive annotations of all possible defect types. Although recent unsupervised methods have achieved strong performance, most are primarily designed for single-class settings and often struggle in multi-class scenarios, where diverse normal patterns may lead to over-generalization and reduce the discriminative capability between normal and anomalous regions. In this paper, we propose SwinAD, a reconstruction-based framework for multi-class unsupervised anomaly detection that leverages a frozen pretrained Swin Transformer V2 encoder and a feature diversity-preserving reconstruction decoder. The hierarchical encoder provides semantically rich multi-scale features, while stage-wise bottleneck modules with dropout prevent trivial identity mapping and encourage robust reconstruction of normal patterns. To further improve localization, we introduce a feature diversity-preserving reconstruction framework that maintains complementary reconstruction hypotheses instead of relying on a single decoding branch. The discrepancies between encoder features and the two reconstructed features are then aggregated across multiple scales to produce the final anomaly map. Experiments conducted on three industrial anomaly detection benchmarks, including MVTec AD, VisA, and Real-IAD, demonstrate that SwinAD achieves competitive image-level performance and strong pixel-level localization accuracy, with particularly notable improvements in pixel-level AP and 1 on MVTec AD. These results indicate that combining hierarchical Swin features with diverse multi-scale reconstruction substantially improve pixel-level localization in multi-class unsupervised anomaly setting.

Problem

• unsupervised anomaly detection
• multi-class industrial inspection
• feature over-generalization
• anomaly localization
• normal pattern diversity

Innovation

• Swin Transformer
• multi-stage reconstruction
• feature diversity
• unsupervised anomaly detection
• multi-class industrial inspection