Label-HyperYOLO: An Improved YOLOv13 Model with Label-Guided Hypergraph Learning and Edge-Aware Enhancement for Steel Surface Defect Detection

Label-HyperYOLO: An Improved YOLOv13 Model with Label-Guided Hypergraph Learning and Edge-Aware Enhancement for Steel Surface Defect Detection | 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 Label-HyperYOLO: An Improved YOLOv13 Model with Label-Guided Hypergraph Learning and Edge-Aware Enhancement for Steel Surface Defect Detection Hui Zhang, Xiaochun Qi This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8087727/v2 This work is licensed under a CC BY 4.0 License Status: Posted Version 2 posted You are reading this latest preprint version Show more versions Abstract To address the challenges of small targets being easily missed, inaccurate localization of weak-edge defects, and insufficient model generalization in steel surface defect detection, this paper proposes the Label-HyperYOLO model, which introduces three innovative modules based on YOLOv13: (1) Adaptive Weighted Loss Routing: An adaptive weighted gating mechanism is incorporated into the multi-scale feature fusion stage, dynamically allocating loss weights based on the importance of scale features to enhance the feature learning capability for small targets. (2) Edge-guided Auxiliary Enhancement Branch: An explicit edge gradient constraint is introduced during the training phase, leveraging edge consistency loss and smoothness loss to improve the model’s perception and localization ability for weak-edge defects. (3) Label-guided Hypergraph Contrastive Distillation Framework: Combining label-guided hypergraph contrastive learning with a teacher-student distillation mechanism, this framework utilizes the high-dimensional representation capability of the teacher model in hypergraph structures to guide the student model in learning more discriminative feature representations using only image inputs, thereby enhancing the robustness and cross-scene generalization capability of the detection model. Experimental results on two public datasets, NEU-DET and GC10-DET, demonstrate that the improved YOLOv13 model significantly outperforms mainstream detection models in metrics such as Recall, mAP50, mAP-small, and mAP-edge. Specifically, on the NEU-DET dataset, mAP50 increased by 3.4%, mAP-small by 2.9%, and mAP-edge by 8.7%. In the generalization experiment on the GC10-DET dataset, Precision improved by 11.6%, mAP-small by 4.5%, and mAP-edge by 10.4%. These results validate the effectiveness and superiority of the proposed method in detecting small targets and weak-edge defects. Steel plate defect detection Small target detection edge enhancement contrastive distillation hypergraph learning Full Text Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 2 posted You are reading this latest preprint version Show more versions 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-8087727","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":551346486,"identity":"f170b20e-c0b4-4a97-9df9-5fc057b9ec23","order_by":0,"name":"Hui Zhang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAy0lEQVRIiWNgGAWjYHACZoYEIGnAwNj4IKGihjQtzQYPzhwjUgsDWAsDm+TDFmbC6g2O9z42eLijlsFcIrmtIrGBjYG/vTsBv5Yzx40TEs8cZ7Cckdh2I3GHDIPEmbMb8Goxu5HGfCCx7RiDwQ2QljNsDAYSuSRoKUhsYyZOS0JiWw1YCwNRWuzPHGM2SGw7APTUw2aJhDPHeAj6RbK9jVnyZ1sdMOjSH378UVEjx9/ei18LFByub4CyeIhRDgJ1xCocBaNgFIyCkQgAYjJMWUYS0esAAAAASUVORK5CYII=","orcid":"","institution":"Yantai Vocational College","correspondingAuthor":true,"prefix":"","firstName":"Hui","middleName":"","lastName":"Zhang","suffix":""},{"id":619499670,"identity":"ea454a86-a90b-4c31-83b4-91fe543d2101","order_by":1,"name":"Xiaochun Qi","email":"","orcid":"","institution":"Yantai Vocational College","correspondingAuthor":false,"prefix":"","firstName":"Xiaochun","middleName":"","lastName":"Qi","suffix":""}],"badges":[],"createdAt":"2025-11-11 14:08:38","currentVersionCode":2,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-8087727/v2","doiUrl":"https://doi.org/10.21203/rs.3.rs-8087727/v2","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":106727451,"identity":"a26c1b92-86ee-4af0-9d9d-6ad8b6afb1da","added_by":"auto","created_at":"2026-04-12 18:39:09","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2072449,"visible":true,"origin":"","legend":"","description":"","filename":"SageLaTeXGuidelines.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8087727/v2_covered_4e576623-3096-42f0-a0e5-7c99f071a696.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"Label-HyperYOLO: An Improved YOLOv13 Model with Label-Guided Hypergraph Learning and Edge-Aware Enhancement for Steel Surface Defect Detection","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"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":"Steel plate defect detection, Small target detection, edge enhancement, contrastive distillation, hypergraph learning","lastPublishedDoi":"10.21203/rs.3.rs-8087727/v2","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8087727/v2","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eTo address the challenges of small targets being easily missed, inaccurate localization of weak-edge defects, and insufficient model generalization in steel surface defect detection, this paper proposes the Label-HyperYOLO model, which introduces three innovative modules based on YOLOv13: (1) Adaptive Weighted Loss Routing: An adaptive weighted gating mechanism is incorporated into the multi-scale feature fusion stage, dynamically allocating loss weights based on the importance of scale features to enhance the feature learning capability for small targets. (2) Edge-guided Auxiliary Enhancement Branch: An explicit edge gradient constraint is introduced during the training phase, leveraging edge consistency loss and smoothness loss to improve the model’s perception and localization ability for weak-edge defects. (3) Label-guided Hypergraph Contrastive Distillation Framework: Combining label-guided hypergraph contrastive learning with a teacher-student distillation mechanism, this framework utilizes the high-dimensional representation capability of the teacher model in hypergraph structures to guide the student model in learning more discriminative feature representations using only image inputs, thereby enhancing the robustness and cross-scene generalization capability of the detection model. Experimental results on two public datasets, NEU-DET and GC10-DET, demonstrate that the improved YOLOv13 model significantly outperforms mainstream detection models in metrics such as Recall, mAP50, mAP-small, and mAP-edge. Specifically, on the NEU-DET dataset, mAP50 increased by 3.4%, mAP-small by 2.9%, and mAP-edge by 8.7%. In the generalization experiment on the GC10-DET dataset, Precision improved by 11.6%, mAP-small by 4.5%, and mAP-edge by 10.4%. These results validate the effectiveness and superiority of the proposed method in detecting small targets and weak-edge defects.\u003c/p\u003e","manuscriptTitle":"Label-HyperYOLO: An Improved YOLOv13 Model with Label-Guided Hypergraph Learning and Edge-Aware Enhancement for Steel Surface Defect Detection","msid":"","msnumber":"","nonDraftVersions":[{"code":2,"date":"2026-04-10 18:43:55","doi":"10.21203/rs.3.rs-8087727/v2","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}},{"code":1,"date":"2025-11-26 17:52:39","doi":"10.21203/rs.3.rs-8087727/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":"fde9f85b-2919-4f57-92b3-ad5132f80eac","owner":[],"postedDate":"April 10th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-04-01T12:40:41+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-10 18:43:55","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v2","identity":"rs-8087727","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8087727","identity":"rs-8087727","version":["v2"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}