Self-Supervised Learning for Industrial Image Anomaly Detection by Simulating Anomalous Samples

IF 2.9 4区计算机科学

International Journal of Computational Intelligence Systems Pub Date : 2023-09-20 DOI:10.1007/s44196-023-00328-0

Mingjing Pei, Ningzhong Liu, Bing Zhao, Han Sun

{"title":"Self-Supervised Learning for Industrial Image Anomaly Detection by Simulating Anomalous Samples","authors":"Mingjing Pei, Ningzhong Liu, Bing Zhao, Han Sun","doi":"10.1007/s44196-023-00328-0","DOIUrl":null,"url":null,"abstract":"Abstract Industrial image anomaly detection (AD) is a critical issue that has been investigated in different research areas. Many works have attempted to detect anomalies by simulating anomalous samples. However, how to simulate abnormal samples remains a significant challenge. In this study, a method for simulating anomalous samples is designed. First, for the object category, patch extraction and patch paste are designed to ensure that the extracted image patches come from the objects and are pasted to the objects in the image. Second, based on the statistical analysis of various anomalies’ presence, a combination of data augmentation is proposed to cover various anomalies as much as possible. The method is evaluated on MVTec AD and BTAD datasets; the experimental results demonstrate that our method achieves an overall detection AUC of 97.6% in MVTec AD datasets, outperforming the baseline by 1.5%, and the improvement over VT-ADL method is 4.3% on the BTAD datasets, demonstrating our method’s effectiveness and generalization.","PeriodicalId":54967,"journal":{"name":"International Journal of Computational Intelligence Systems","volume":"16 1","pages":"0"},"PeriodicalIF":2.9000,"publicationDate":"2023-09-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"International Journal of Computational Intelligence Systems","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1007/s44196-023-00328-0","RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}

引用次数: 0

Abstract

Abstract Industrial image anomaly detection (AD) is a critical issue that has been investigated in different research areas. Many works have attempted to detect anomalies by simulating anomalous samples. However, how to simulate abnormal samples remains a significant challenge. In this study, a method for simulating anomalous samples is designed. First, for the object category, patch extraction and patch paste are designed to ensure that the extracted image patches come from the objects and are pasted to the objects in the image. Second, based on the statistical analysis of various anomalies’ presence, a combination of data augmentation is proposed to cover various anomalies as much as possible. The method is evaluated on MVTec AD and BTAD datasets; the experimental results demonstrate that our method achieves an overall detection AUC of 97.6% in MVTec AD datasets, outperforming the baseline by 1.5%, and the improvement over VT-ADL method is 4.3% on the BTAD datasets, demonstrating our method’s effectiveness and generalization.

查看原文本刊更多论文

基于模拟异常样本的工业图像异常检测自监督学习

摘要工业图像异常检测是目前国内外研究的热点问题之一。许多工作都试图通过模拟异常样本来检测异常。然而，如何模拟异常样品仍然是一个重大的挑战。本文设计了一种模拟异常样本的方法。首先，对于目标类别，设计了补丁提取和补丁粘贴，确保提取的图像补丁来自于目标，并粘贴到图像中的目标上。其次，在对各种异常存在情况进行统计分析的基础上，提出结合数据增强的方法，尽可能覆盖各种异常。在MVTec AD和BTAD数据集上对该方法进行了评价;实验结果表明，该方法在MVTec AD数据集上的总体检测AUC达到97.6%，比基线提高1.5%，在BTAD数据集上比VT-ADL方法提高4.3%，证明了该方法的有效性和泛化性。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

International Journal of Computational Intelligence Systems 工程技术-计算机：跨学科应用

自引率

3.40%

发文量

期刊介绍： The International Journal of Computational Intelligence Systems publishes original research on all aspects of applied computational intelligence, especially targeting papers demonstrating the use of techniques and methods originating from computational intelligence theory. The core theories of computational intelligence are fuzzy logic, neural networks, evolutionary computation and probabilistic reasoning. The journal publishes only articles related to the use of computational intelligence and broadly covers the following topics: -Autonomous reasoning- Bio-informatics- Cloud computing- Condition monitoring- Data science- Data mining- Data visualization- Decision support systems- Fault diagnosis- Intelligent information retrieval- Human-machine interaction and interfaces- Image processing- Internet and networks- Noise analysis- Pattern recognition- Prediction systems- Power (nuclear) safety systems- Process and system control- Real-time systems- Risk analysis and safety-related issues- Robotics- Signal and image processing- IoT and smart environments- Systems integration- System control- System modelling and optimization- Telecommunications- Time series prediction- Warning systems- Virtual reality- Web intelligence- Deep learning