Enhanced Industrial Anomaly Detection via CutMask Data Augmentation: A Self-Supervised Approach

Enhanced Industrial Anomaly Detection via CutMask Data Augmentation: A Self-Supervised Approach | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Method Article Enhanced Industrial Anomaly Detection via CutMask Data Augmentation: A Self-Supervised Approach Le Yang, Dingjian Yao, Wenhan Yang This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8803927/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Deep learning techniques have revolutionized industrial anomaly detection; how-ever, their reliance on substantial labeled data poses challenges in scenarios where anomalous samples are scarce. This paper introduces a novel self-supervised anomaly detection framework, CutMask-based Anomaly Detection (CMAD), designed to detect anomalies using only normal samples. CMAD incorporates an improved data augmentation method, CutMask, which leverages prior knowledge of defect shapes to generate realistic simulated defect samples. Furthermore, it employs an enhanced Cyclical Adversarial Focal Loss function to improve sample discrimination and utilizes a lightweight ResNet-18 model for efficient defect detection. A Self-Supervised Predictive Convolutional Attentive Block (SSP-CAB) module is integrated to enhance feature modeling. Experimental results on the MVTecAD dataset and a practical presswork dataset demonstrate CMAD’s superior performance, achieving an AUC-ROC score of 97.8% on MVTecAD, out-performing existing methods. This framework offers a practical solution to the challenge of limited anomalous data in industrial settings.The code and dataset are available at: https://github.com/Jared-Yao/CutMask. Artificial Intelligence and Machine Learning Anomaly detection CutMask Data augmentation Deep learning MVTecAD dataset Presswork production Full Text Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8803927","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Method Article","associatedPublications":[],"authors":[{"id":586790261,"identity":"6dce3332-6219-4f9b-93a8-5af81dbf2b9c","order_by":0,"name":"Le Yang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA1ElEQVRIiWNgGAWjYLCCBww2UBYbsVoSGNJI13KYBC0Gx88efpFQcT6PXyJ5A8OHssMM/LMbCGg5k5dmkXDmdrHkjLQCxhnnDjNI3DlAQMuBHDODxLbbiRtu5Bgw87YdZjCQSCCg5fwboJZ/5xL3g7T8JUrLjRzjB4kNBxI3SAC1MBKjRfLGGzOGhGPJiTPOPCs42HMunUfiBgEtfOdzjD98qLFL7G9P3vjgR5m1HP8MAloUDjCwScAceQBI8OBXDwTyDQzMH2BaCKoeBaNgFIyCkQkA+l1JnyHKKPcAAAAASUVORK5CYII=","orcid":"","institution":"College of Information Science and Engineering, Northeastern University, Shenyang 110819, China","correspondingAuthor":true,"prefix":"","firstName":"Le","middleName":"","lastName":"Yang","suffix":""},{"id":586790262,"identity":"0a9ecca8-aeab-4726-b312-bc8c6229f525","order_by":1,"name":"Dingjian Yao","email":"","orcid":"","institution":"School of Control Engineering, Northeastern University at Qinhuangdao, Qinhuangdao 066004, China","correspondingAuthor":false,"prefix":"","firstName":"Dingjian","middleName":"","lastName":"Yao","suffix":""},{"id":586790263,"identity":"e395457a-89cb-49d9-90d3-a45359f02031","order_by":2,"name":"Wenhan Yang","email":"","orcid":"","institution":"School of Control Engineering, Northeastern University at Qinhuangdao, Qinhuangdao 066004, China","correspondingAuthor":false,"prefix":"","firstName":"Wenhan","middleName":"","lastName":"Yang","suffix":""}],"badges":[],"createdAt":"2026-02-06 07:33:48","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-8803927/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8803927/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":102172579,"identity":"7a6875b8-9c75-48bb-a10b-07593b42e75e","added_by":"auto","created_at":"2026-02-09 05:04:56","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":6547329,"visible":true,"origin":"","legend":"","description":"","filename":"cutmaskfile.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8803927/v1_covered_3f30a063-9f21-49ea-bab5-60601eb3d4e7.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eEnhanced Industrial Anomaly Detection via CutMask Data Augmentation: A Self-Supervised Approach\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Anomaly detection, CutMask, Data augmentation, Deep learning, MVTecAD dataset, Presswork production","lastPublishedDoi":"10.21203/rs.3.rs-8803927/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8803927/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eDeep learning techniques have revolutionized industrial anomaly detection; how-ever, their reliance on substantial labeled data poses challenges in scenarios where anomalous samples are scarce. This paper introduces a novel self-supervised anomaly detection framework, CutMask-based Anomaly Detection (CMAD), designed to detect anomalies using only normal samples. CMAD incorporates an improved data augmentation method, CutMask, which leverages prior knowledge of defect shapes to generate realistic simulated defect samples. Furthermore, it employs an enhanced Cyclical Adversarial Focal Loss function to improve sample discrimination and utilizes a lightweight ResNet-18 model for efficient defect detection. A Self-Supervised Predictive Convolutional Attentive Block (SSP-CAB) module is integrated to enhance feature modeling. Experimental results on the MVTecAD dataset and a practical presswork dataset demonstrate CMAD’s superior performance, achieving an AUC-ROC score of 97.8% on MVTecAD, out-performing existing methods. This framework offers a practical solution to the challenge of limited anomalous data in industrial settings.The code and dataset are available at: https://github.com/Jared-Yao/CutMask.\u003c/p\u003e","manuscriptTitle":"Enhanced Industrial Anomaly Detection via CutMask Data Augmentation: A Self-Supervised Approach","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-09 05:04:32","doi":"10.21203/rs.3.rs-8803927/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"62de666c-ed1b-4cb4-a60a-f420080c1d60","owner":[],"postedDate":"February 9th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":62434168,"name":"Artificial Intelligence and Machine Learning"}],"tags":[],"updatedAt":"2026-02-16T13:24:12+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-09 05:04:32","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8803927","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8803927","identity":"rs-8803927","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}