WeldGen: Control-Guided Stable Diffusion for Diverse Weld Image Synthesis

WeldGen: Control-Guided Stable Diffusion for Diverse Weld Image Synthesis | 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 WeldGen: Control-Guided Stable Diffusion for Diverse Weld Image Synthesis Qin Zhang, Junhao Deng, Zhongyou Zhao, Zhelong Song, Zhenmin Wang, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9149901/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 4 You are reading this latest preprint version Abstract In this study, we propose a novel data augmentation framework for weld image classification based on the generativecapabilities of Stable Diffusion. First, we fine-tune a pre-trained diffusion model using DreamBooth to learn the conceptof weld seam appearance. Then, to promote structural diversity during image synthesis, we employ ControlNet to guidethe generation process with randomly selected mask images from the same weld category. During inference, a textualprompt, base image, and control mask are jointly used to steer the synthesis process toward realistic and structurallydiverse outputs. The generated images are then used to augment the original dataset. Experimental results on aweld classification task demonstrate that the inclusion of our generated data significantly improves model performance.Specifically, the classification accuracy increased from 93.3% to 99.2%, indicating the effectiveness of our approach inenhancing both data diversity and downstream task performance. Diffusion model Data augmentation Weld image Stable diffusion Deep learning Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 19 Mar, 2026 Editor assigned by journal 19 Mar, 2026 Submission checks completed at journal 19 Mar, 2026 First submitted to journal 17 Mar, 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-9149901","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":609069705,"identity":"f8b4d9bd-ab59-4d90-b568-65d69d5aef81","order_by":0,"name":"Qin Zhang","email":"","orcid":"","institution":"South China University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Qin","middleName":"","lastName":"Zhang","suffix":""},{"id":609069709,"identity":"3a5b87a6-90d7-4504-bbbd-3d06903080c3","order_by":1,"name":"Junhao Deng","email":"","orcid":"","institution":"South China University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Junhao","middleName":"","lastName":"Deng","suffix":""},{"id":609069711,"identity":"82cb4a9a-ee9a-42e9-9f70-5f56370937cf","order_by":2,"name":"Zhongyou Zhao","email":"","orcid":"","institution":"South China University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Zhongyou","middleName":"","lastName":"Zhao","suffix":""},{"id":609069713,"identity":"932fa39a-0107-4779-9919-c2f2187905f4","order_by":3,"name":"Zhelong Song","email":"","orcid":"","institution":"South China University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Zhelong","middleName":"","lastName":"Song","suffix":""},{"id":609069715,"identity":"e21622cb-0a45-44e0-bbb4-f7c6b87ff081","order_by":4,"name":"Zhenmin Wang","email":"","orcid":"","institution":"South China University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Zhenmin","middleName":"","lastName":"Wang","suffix":""},{"id":609069717,"identity":"02363af7-be01-4f2e-81ad-629c29f784e8","order_by":5,"name":"Hui-ping Wang","email":"","orcid":"","institution":"General Motors (United States)","correspondingAuthor":false,"prefix":"","firstName":"Hui-ping","middleName":"","lastName":"Wang","suffix":""},{"id":609069719,"identity":"187f9d3d-93c9-4fb9-91cb-58a062a196c9","order_by":6,"name":"Zixuan Wan","email":"","orcid":"","institution":"General Motors (United States)","correspondingAuthor":false,"prefix":"","firstName":"Zixuan","middleName":"","lastName":"Wan","suffix":""},{"id":609069721,"identity":"e7229149-779f-4f7d-b321-8cc452daec5a","order_by":7,"name":"Jorge Arinez","email":"","orcid":"","institution":"General Motors (United States)","correspondingAuthor":false,"prefix":"","firstName":"Jorge","middleName":"","lastName":"Arinez","suffix":""},{"id":609069723,"identity":"c4a1edc1-02e2-4bc2-a481-ac34daf630b0","order_by":8,"name":"Guangze Li","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAx0lEQVRIiWNgGAWjYDADPmbmAyRqYWNmSyBVCwOPAXEq5SOSnz388seGgY2d5/OrGwx2croNBLQY3kgzN5bhSQM6jHebdQ5DsrHZAUJaZiSYSUtIHAZrMc5hOJC4jbCW9G/SEgb/gVp4nhGnRV4ix0zyQ8IBkBbmx0RpMeB5UybNcCCZBxjIZsw5BkT4Rb49fZvkjz92cvz8hx9/zqmwkyOoxQCogJmHgQGIGNgkGIiJGvkGBgbGHxA28wciNIyCUTAKRsEIBAA16DcN9f+ongAAAABJRU5ErkJggg==","orcid":"","institution":"General Motors (United States)","correspondingAuthor":true,"prefix":"","firstName":"Guangze","middleName":"","lastName":"Li","suffix":""}],"badges":[],"createdAt":"2026-03-17 14:23:34","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9149901/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9149901/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":109233028,"identity":"48204a8c-b57a-4ac7-9ea7-01f922bd1853","added_by":"auto","created_at":"2026-05-14 04:03:51","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":5890194,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9149901/v1_covered_97642684-7e12-4967-8eac-36d020f77d1f.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"WeldGen: Control-Guided Stable Diffusion for Diverse Weld Image Synthesis","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":"the-visual-computer","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"tvcj","sideBox":"Learn more about [The Visual Computer](http://link.springer.com/journal/371)","snPcode":"371","submissionUrl":"https://submission.nature.com/new-submission/371/3","title":"The Visual Computer","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Diffusion model, Data augmentation, Weld image, Stable diffusion, Deep learning","lastPublishedDoi":"10.21203/rs.3.rs-9149901/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9149901/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"In this study, we propose a novel data augmentation framework for weld image classification based on the generativecapabilities of Stable Diffusion. First, we fine-tune a pre-trained diffusion model using DreamBooth to learn the conceptof weld seam appearance. Then, to promote structural diversity during image synthesis, we employ ControlNet to guidethe generation process with randomly selected mask images from the same weld category. During inference, a textualprompt, base image, and control mask are jointly used to steer the synthesis process toward realistic and structurallydiverse outputs. The generated images are then used to augment the original dataset. Experimental results on aweld classification task demonstrate that the inclusion of our generated data significantly improves model performance.Specifically, the classification accuracy increased from 93.3% to 99.2%, indicating the effectiveness of our approach inenhancing both data diversity and downstream task performance.","manuscriptTitle":"WeldGen: Control-Guided Stable Diffusion for Diverse Weld Image Synthesis","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-14 04:03:42","doi":"10.21203/rs.3.rs-9149901/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-03-19T18:09:46+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-03-19T17:15:52+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-03-19T17:04:10+00:00","index":"","fulltext":""},{"type":"submitted","content":"The Visual Computer","date":"2026-03-17T14:17:04+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"the-visual-computer","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"tvcj","sideBox":"Learn more about [The Visual Computer](http://link.springer.com/journal/371)","snPcode":"371","submissionUrl":"https://submission.nature.com/new-submission/371/3","title":"The Visual Computer","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"b8dadbb8-9612-40b0-8543-ca89dddf55c2","owner":[],"postedDate":"May 14th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-05-14T04:03:42+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-14 04:03:42","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9149901","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9149901","identity":"rs-9149901","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}