Industrial Image Anomaly Detection via Synthetic-Anomaly Contrastive Distillation

Industrial image anomaly detection plays a critical role in intelligent manufacturing by automatically identifying defective products through visual inspection. While unsupervised approaches eliminate dependency on annotated anomaly samples, current teacher–student framework-based methods still face...

Full description

Saved in:
Bibliographic Details
Main Authors: Junxian Li, Mingxing Li, Shucheng Huang, Gang Wang, Xinjing Zhao
Format: Article
Language:English
Published: MDPI AG 2025-06-01
Series:Sensors
Subjects:
Online Access:https://www.mdpi.com/1424-8220/25/12/3721
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849431553753481216
author Junxian Li
Mingxing Li
Shucheng Huang
Gang Wang
Xinjing Zhao
author_facet Junxian Li
Mingxing Li
Shucheng Huang
Gang Wang
Xinjing Zhao
author_sort Junxian Li
collection DOAJ
description Industrial image anomaly detection plays a critical role in intelligent manufacturing by automatically identifying defective products through visual inspection. While unsupervised approaches eliminate dependency on annotated anomaly samples, current teacher–student framework-based methods still face two fundamental limitations: insufficient discriminative capability for structural anomalies and suboptimal anomaly feature decoupling efficiency. To address these challenges, we propose a Synthetic-Anomaly Contrastive Distillation (<i>SACD</i>) framework for industrial anomaly detection. <i>SACD</i> comprises two pivotal components: (1) a reverse distillation (RD) paradigm whereby a pre-trained teacher network extracts hierarchically structured representations, subsequently guiding the student network with inverse architectural configuration to establish hierarchical feature alignment; (2) a group of feature calibration (<i>FeaCali</i>) modules designed to refine the student’s outputs by eliminating anomalous feature responses. During training, <i>SACD</i> adopts a dual-branch strategy, where one branch encodes multi-scale features from defect-free images, while a Siamese anomaly branch processes synthetically corrupted counterparts. <i>FeaCali</i> modules are trained to strip out a student’s anomalous patterns in anomaly branches, enhancing the student network’s exclusive modeling of normal patterns. We construct a dual-objective optimization integrating cross-model distillation loss and intra-model contrastive loss to train <i>SACD</i> for feature alignment and discrepancy amplification. At the inference stage, pixel-wise anomaly scores are computed through multi-layer feature discrepancies between the teacher’s representations and the student’s refined outputs. Comprehensive evaluations on the MVTec AD and BTAD benchmark demonstrate that our method is effective and superior to current knowledge distillation-based approaches.
format Article
id doaj-art-a12a164f06cc4cbd8700a043b50f02c9
institution Kabale University
issn 1424-8220
language English
publishDate 2025-06-01
publisher MDPI AG
record_format Article
series Sensors
spelling doaj-art-a12a164f06cc4cbd8700a043b50f02c92025-08-20T03:27:36ZengMDPI AGSensors1424-82202025-06-012512372110.3390/s25123721Industrial Image Anomaly Detection via Synthetic-Anomaly Contrastive DistillationJunxian Li0Mingxing Li1Shucheng Huang2Gang Wang3Xinjing Zhao4School of Information Engineering, Yangzhou Polytechnic College, Yangzhou 225009, ChinaSchool of Electrical and Information Engineering, Jiangsu University JingJiang College, Zhenjiang 212013, ChinaSchool of Computer, Jiangsu University of Science and Technology, Zhenjiang 212003, ChinaSchool of Computing and Artificial Intelligence, Southwest Jiaotong University, Chengdu 611756, ChinaSuzhou Zhanchi Tongyang Talent Technology Company, Suzhou 215558, ChinaIndustrial image anomaly detection plays a critical role in intelligent manufacturing by automatically identifying defective products through visual inspection. While unsupervised approaches eliminate dependency on annotated anomaly samples, current teacher–student framework-based methods still face two fundamental limitations: insufficient discriminative capability for structural anomalies and suboptimal anomaly feature decoupling efficiency. To address these challenges, we propose a Synthetic-Anomaly Contrastive Distillation (<i>SACD</i>) framework for industrial anomaly detection. <i>SACD</i> comprises two pivotal components: (1) a reverse distillation (RD) paradigm whereby a pre-trained teacher network extracts hierarchically structured representations, subsequently guiding the student network with inverse architectural configuration to establish hierarchical feature alignment; (2) a group of feature calibration (<i>FeaCali</i>) modules designed to refine the student’s outputs by eliminating anomalous feature responses. During training, <i>SACD</i> adopts a dual-branch strategy, where one branch encodes multi-scale features from defect-free images, while a Siamese anomaly branch processes synthetically corrupted counterparts. <i>FeaCali</i> modules are trained to strip out a student’s anomalous patterns in anomaly branches, enhancing the student network’s exclusive modeling of normal patterns. We construct a dual-objective optimization integrating cross-model distillation loss and intra-model contrastive loss to train <i>SACD</i> for feature alignment and discrepancy amplification. At the inference stage, pixel-wise anomaly scores are computed through multi-layer feature discrepancies between the teacher’s representations and the student’s refined outputs. Comprehensive evaluations on the MVTec AD and BTAD benchmark demonstrate that our method is effective and superior to current knowledge distillation-based approaches.https://www.mdpi.com/1424-8220/25/12/3721anomaly detectionanomaly localizationknowledge distillationfeature refinementabnormal synthesis
spellingShingle Junxian Li
Mingxing Li
Shucheng Huang
Gang Wang
Xinjing Zhao
Industrial Image Anomaly Detection via Synthetic-Anomaly Contrastive Distillation
Sensors
anomaly detection
anomaly localization
knowledge distillation
feature refinement
abnormal synthesis
title Industrial Image Anomaly Detection via Synthetic-Anomaly Contrastive Distillation
title_full Industrial Image Anomaly Detection via Synthetic-Anomaly Contrastive Distillation
title_fullStr Industrial Image Anomaly Detection via Synthetic-Anomaly Contrastive Distillation
title_full_unstemmed Industrial Image Anomaly Detection via Synthetic-Anomaly Contrastive Distillation
title_short Industrial Image Anomaly Detection via Synthetic-Anomaly Contrastive Distillation
title_sort industrial image anomaly detection via synthetic anomaly contrastive distillation
topic anomaly detection
anomaly localization
knowledge distillation
feature refinement
abnormal synthesis
url https://www.mdpi.com/1424-8220/25/12/3721
work_keys_str_mv AT junxianli industrialimageanomalydetectionviasyntheticanomalycontrastivedistillation
AT mingxingli industrialimageanomalydetectionviasyntheticanomalycontrastivedistillation
AT shuchenghuang industrialimageanomalydetectionviasyntheticanomalycontrastivedistillation
AT gangwang industrialimageanomalydetectionviasyntheticanomalycontrastivedistillation
AT xinjingzhao industrialimageanomalydetectionviasyntheticanomalycontrastivedistillation