MSFViT: A Lightweight MobileViT basedmodel for Steel Surface Defect Detection onEdge Constrained Devices

MSFViT: A Lightweight MobileViT basedmodel for Steel Surface Defect Detection onEdge Constrained Devices | 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 Research Article MSFViT: A Lightweight MobileViT basedmodel for Steel Surface Defect Detection onEdge Constrained Devices Neethu Radha Gopan, Nithin Das, P. Sarin Mohan, Rahul Cyriac, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9463530/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 10 You are reading this latest preprint version Abstract Steel surface integrity is a critical concern in modern manufacturing, whereeven minor defects can significantly impact product reliability and safety. Traditionalmanual inspection methods are time-consuming, subjective, and unsuitable for high-speedproduction environments, creating a strong need for automated, edge-deployable inspection systems. Our work investigates lightweight hybrid Vision Transformer architectures for real-timesteel surface defect classification, focusing on models that effectively capture both localtexture and global contextual information while maintaining low computational complexity. We propose MSFViT-XXS, a stripe-free variant of MobileViT that replaces conventional patch-based tokenisation with global average pooling and bilinear interpolation,thereby eliminating spatial artifacts introduced by rigid grid partitioning. To further enhance efficiency and performance, a three-component knowledge distillation framework is employed to transfer representations from a larger MSFViT-XS teachermodel to a compact student model. Additionally, we evaluate MobileViT-XXS with Exponential Moving Average (EMA) training for improved generalisation. All models are evaluated on a multi-class steel surface defect dataset and comparedagainst standard Vision Transformer baselines. The results demonstrate that the proposed approaches achieve an effective balance between accuracy, computational efficiency,and model compactness, making them well-suited for real-time deployment on resourceconstrained edge devices in industrial inspection systems. steel surface defect detection industrial inspection hybrid vision transformer lightweight deep learning MobileViT stripe-free attention global context modeling knowledge distillation exponential moving average edge AI real-time inference model compression efficient neural networks Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Reviews received at journal 19 May, 2026 Reviewers agreed at journal 16 May, 2026 Reviewers agreed at journal 14 May, 2026 Reviews received at journal 13 May, 2026 Reviewers agreed at journal 12 May, 2026 Reviewers agreed at journal 27 Apr, 2026 Reviewers invited by journal 27 Apr, 2026 Editor assigned by journal 26 Apr, 2026 Submission checks completed at journal 26 Apr, 2026 First submitted to journal 19 Apr, 2026 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-9463530","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":631476129,"identity":"64258cef-9b08-4aff-9516-5218f5e248cb","order_by":0,"name":"Neethu Radha Gopan","email":"data:image/png;base64,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","orcid":"","institution":"Rajagiri School of Engineering \u0026 Technology","correspondingAuthor":true,"prefix":"","firstName":"Neethu","middleName":"Radha","lastName":"Gopan","suffix":""},{"id":631476131,"identity":"d48bca50-d63d-4005-b9d2-dd0079d9cd64","order_by":1,"name":"Nithin Das","email":"","orcid":"","institution":"Rajagiri School of Engineering \u0026 Technology","correspondingAuthor":false,"prefix":"","firstName":"Nithin","middleName":"","lastName":"Das","suffix":""},{"id":631476134,"identity":"396eac88-4bc6-4720-9462-b2602a44f9a7","order_by":2,"name":"P. Sarin Mohan","email":"","orcid":"","institution":"Rajagiri School of Engineering \u0026 Technology","correspondingAuthor":false,"prefix":"","firstName":"P.","middleName":"Sarin","lastName":"Mohan","suffix":""},{"id":631476143,"identity":"ae5498ae-cb60-47c5-96ad-1a50ffad29f3","order_by":3,"name":"Rahul Cyriac","email":"","orcid":"","institution":"Rajagiri School of Engineering \u0026 Technology","correspondingAuthor":false,"prefix":"","firstName":"Rahul","middleName":"","lastName":"Cyriac","suffix":""},{"id":631476146,"identity":"89fc24fa-737c-4336-a8b3-4e5704cbe03b","order_by":4,"name":"Sreekiran M","email":"","orcid":"","institution":"Rajagiri School of Engineering \u0026 Technology","correspondingAuthor":false,"prefix":"","firstName":"Sreekiran","middleName":"","lastName":"M","suffix":""}],"badges":[],"createdAt":"2026-04-19 16:10:12","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9463530/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9463530/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108181843,"identity":"89fb80ce-2c08-4f2e-abc5-f229a85dedd1","added_by":"auto","created_at":"2026-04-30 08:58:58","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2485602,"visible":true,"origin":"","legend":"","description":"","filename":"MSFViTpaper1.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9463530/v1_covered_bc297e1d-a7fe-42c2-be1a-431104e4a4a6.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"MSFViT: A Lightweight MobileViT basedmodel for Steel Surface Defect Detection onEdge Constrained Devices","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"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":"international-journal-of-machine-learning-and-cybernetics","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"jmlc","sideBox":"Learn more about [International Journal of Machine Learning and Cybernetics](http://actavetscand.biomedcentral.com/)","snPcode":"13042","submissionUrl":"https://submission.nature.com/new-submission/13042/3","title":"International Journal of Machine Learning and Cybernetics","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"steel surface defect detection, industrial inspection, hybrid vision transformer, lightweight deep learning, MobileViT, stripe-free attention, global context modeling, knowledge distillation, exponential moving average, edge AI, real-time inference, model compression, efficient neural networks ","lastPublishedDoi":"10.21203/rs.3.rs-9463530/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9463530/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eSteel surface integrity is a critical concern in modern manufacturing, whereeven minor defects can significantly impact product reliability and safety. Traditionalmanual inspection methods are time-consuming, subjective, and unsuitable for high-speedproduction environments, creating a strong need for automated, edge-deployable inspection systems.\u003c/p\u003e\n\u003cp\u003eOur work investigates lightweight hybrid Vision Transformer architectures for real-timesteel surface defect classification, focusing on models that effectively capture both localtexture and global contextual information while maintaining low computational complexity. We propose MSFViT-XXS, a stripe-free variant of MobileViT that replaces conventional patch-based tokenisation with global average pooling and bilinear interpolation,thereby eliminating spatial artifacts introduced by rigid grid partitioning.\u003c/p\u003e\n\u003cp\u003eTo further enhance efficiency and performance, a three-component knowledge distillation framework is employed to transfer representations from a larger MSFViT-XS teachermodel to a compact student model. Additionally, we evaluate MobileViT-XXS with Exponential Moving Average (EMA) training for improved generalisation.\u003c/p\u003e\n\u003cp\u003eAll models are evaluated on a multi-class steel surface defect dataset and comparedagainst standard Vision Transformer baselines. The results demonstrate that the proposed approaches achieve an effective balance between accuracy, computational efficiency,and model compactness, making them well-suited for real-time deployment on resourceconstrained edge devices in industrial inspection systems.\u003c/p\u003e","manuscriptTitle":"MSFViT: A Lightweight MobileViT basedmodel for Steel Surface Defect Detection onEdge Constrained Devices","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-29 12:06:37","doi":"10.21203/rs.3.rs-9463530/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"editorInvitedReview","content":"","date":"2026-05-19T07:23:23+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"156604846074456409155034781207753968047","date":"2026-05-16T05:10:59+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"102218784372704843650156653724581624064","date":"2026-05-14T06:03:33+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-13T17:06:44+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"173418353499096600302785638691989871922","date":"2026-05-12T12:13:09+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"273887375945440272717098605743350917865","date":"2026-04-27T15:00:52+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-27T09:26:55+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-04-27T02:34:50+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-04-26T07:12:25+00:00","index":"","fulltext":""},{"type":"submitted","content":"International Journal of Machine Learning and Cybernetics","date":"2026-04-19T16:05:06+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"international-journal-of-machine-learning-and-cybernetics","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"jmlc","sideBox":"Learn more about [International Journal of Machine Learning and Cybernetics](http://actavetscand.biomedcentral.com/)","snPcode":"13042","submissionUrl":"https://submission.nature.com/new-submission/13042/3","title":"International Journal of Machine Learning and Cybernetics","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"e8b12832-abb3-499b-9ca1-b38ee86cb6a7","owner":[],"postedDate":"April 29th, 2026","published":true,"recentEditorialEvents":[{"type":"editorInvitedReview","content":"","date":"2026-05-19T07:23:23+00:00","index":54,"fulltext":""},{"type":"reviewerAgreed","content":"156604846074456409155034781207753968047","date":"2026-05-16T05:10:59+00:00","index":53,"fulltext":""},{"type":"reviewerAgreed","content":"102218784372704843650156653724581624064","date":"2026-05-14T06:03:33+00:00","index":51,"fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-13T17:06:44+00:00","index":50,"fulltext":""},{"type":"reviewerAgreed","content":"173418353499096600302785638691989871922","date":"2026-05-12T12:13:09+00:00","index":49,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-04-29T12:06:37+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-29 12:06:37","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9463530","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9463530","identity":"rs-9463530","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}