EFE-SDG: Efficient Feature Extraction of Finetuning-Free Model in Subject-Driven Generation

EFE-SDG: Efficient Feature Extraction of Finetuning-Free Model in Subject-Driven Generation | 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 EFE-SDG: Efficient Feature Extraction of Finetuning-Free Model in Subject-Driven Generation Hao Li, Yongzhen Ke, Shuai Yang, Kai Wang, Yemeng Wu This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7499053/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 10 Mar, 2026 Read the published version in Multimedia Systems → Version 1 posted 11 You are reading this latest preprint version Abstract Text-to-image models mark a major advance in AI, and subject-driven generation techniques hold promise. The subject-driven models that are currently avail- able still necessitate a compromise between the fidelity of the subject and the extent of text-to-image generation. Furthermore, they cannot provide a compre- hensive understanding of the reference image based on a limited amount of data. To address these limitations, we propose the EFE-SDG model, which is divided into three blocks: (i) the MOCI block employs the Gpt-4o and other large mod- els for the processing of the reference image to obtain more information about the reference image. (ii) The ASCA module, based on the decoupling strategy of the cross-attention module, performs additional processing on the reference image to obtain more detailed and richer high-level features of the subject. (iii) The subject Feature Adaptive Attention Rules is used to fuse the low-level fea- tures extracted by the ref-diffusion model and the high-level features extracted by ASCA at the Attention layer of the main-diffusion model. In addition, we use the ref-diffusion model to extract low-level feature inputs from the reference features to the main-diffusion model, which circumvents the different training distributions of the other encoders and the main-diffusion model. In our compar- ative experiments, with respect to graphic alignment, our approach demonstrates performance similar to that of the other methods, but our approach uses sub- stantially smaller datasets and fewer computational resources during the training phase.Our code will be available at:https://gitee.com/yongzhenke/efe-sdg Diffusion Model Text-to-image Generation Subject-driven Generation Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 10 Mar, 2026 Read the published version in Multimedia Systems → Version 1 posted Editorial decision: Revision requested 08 Dec, 2025 Reviews received at journal 30 Nov, 2025 Reviewers agreed at journal 25 Nov, 2025 Reviews received at journal 23 Oct, 2025 Reviewers agreed at journal 19 Oct, 2025 Reviewers agreed at journal 16 Oct, 2025 Reviewers agreed at journal 14 Oct, 2025 Reviewers invited by journal 14 Oct, 2025 Editor assigned by journal 11 Oct, 2025 Submission checks completed at journal 11 Sep, 2025 First submitted to journal 31 Aug, 2025 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-7499053","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":533960321,"identity":"0df9add0-f78b-443f-a7b2-1f0afa495a71","order_by":0,"name":"Hao Li","email":"","orcid":"","institution":"Tianjin Polytechnic University","correspondingAuthor":false,"prefix":"","firstName":"Hao","middleName":"","lastName":"Li","suffix":""},{"id":533960322,"identity":"64fd335d-5f1d-463a-afdb-12d266ce7b6b","order_by":1,"name":"Yongzhen Ke","email":"data:image/png;base64,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","orcid":"","institution":"Tianjin Polytechnic University","correspondingAuthor":true,"prefix":"","firstName":"Yongzhen","middleName":"","lastName":"Ke","suffix":""},{"id":533960323,"identity":"53add72b-40d9-4eba-9b68-4c902c7e5ba5","order_by":2,"name":"Shuai Yang","email":"","orcid":"","institution":"Tianjin Polytechnic University","correspondingAuthor":false,"prefix":"","firstName":"Shuai","middleName":"","lastName":"Yang","suffix":""},{"id":533960325,"identity":"cb3562d8-4926-4453-92fd-b1e6b71ecb49","order_by":3,"name":"Kai Wang","email":"","orcid":"","institution":"Tianjin Polytechnic University","correspondingAuthor":false,"prefix":"","firstName":"Kai","middleName":"","lastName":"Wang","suffix":""},{"id":533960327,"identity":"efee4995-56d5-41b5-bf70-fadcc2d7dbe3","order_by":4,"name":"Yemeng Wu","email":"","orcid":"","institution":"Beijing Jiaotong University","correspondingAuthor":false,"prefix":"","firstName":"Yemeng","middleName":"","lastName":"Wu","suffix":""}],"badges":[],"createdAt":"2025-08-31 07:38:17","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7499053/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7499053/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s00530-026-02248-3","type":"published","date":"2026-03-10T15:57:46+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":94584720,"identity":"ed9adafd-e911-41cc-851f-c12434b30574","added_by":"auto","created_at":"2025-10-28 18:15:32","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":53337181,"visible":true,"origin":"","legend":"","description":"","filename":"news.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7499053/v1/0a76471fda550f08947138d1.pdf"},{"id":94585062,"identity":"32d56432-a7dd-441b-8b1d-a73fbf3a96c0","added_by":"auto","created_at":"2025-10-28 18:15:51","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6560,"visible":true,"origin":"","legend":"","description":"","filename":"65316c8fe29b40899712c4640e661304.json","url":"https://assets-eu.researchsquare.com/files/rs-7499053/v1/a4249b04719688861d994fc1.json"},{"id":104739481,"identity":"4ed9849c-9ce9-4e80-8b0c-ef8eee1f4f0f","added_by":"auto","created_at":"2026-03-16 16:07:31","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":7421246,"visible":true,"origin":"","legend":"","description":"","filename":"news.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7499053/v1_covered_650cb603-2775-4e32-a612-fed64e7624e2.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"EFE-SDG: Efficient Feature Extraction of Finetuning-Free Model in Subject-Driven Generation","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"multimedia-systems","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"mmsj","sideBox":"Learn more about [Multimedia Systems](http://link.springer.com/journal/530)","snPcode":"530","submissionUrl":"https://submission.nature.com/new-submission/530/3","title":"Multimedia Systems","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Diffusion Model, Text-to-image Generation, Subject-driven Generation","lastPublishedDoi":"10.21203/rs.3.rs-7499053/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7499053/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eText-to-image models mark a major advance in AI, and subject-driven generation techniques hold promise. The subject-driven models that are currently avail- able still necessitate a compromise between the fidelity of the subject and the extent of text-to-image generation. Furthermore, they cannot provide a compre- hensive understanding of the reference image based on a limited amount of data. To address these limitations, we propose the EFE-SDG model, which is divided into three blocks: (i) the MOCI block employs the Gpt-4o and other large mod- els for the processing of the reference image to obtain more information about the reference image. (ii) The ASCA module, based on the decoupling strategy of the cross-attention module, performs additional processing on the reference image to obtain more detailed and richer high-level features of the subject. (iii) The subject Feature Adaptive Attention Rules is used to fuse the low-level fea- tures extracted by the ref-diffusion model and the high-level features extracted by ASCA at the Attention layer of the main-diffusion model. In addition, we use the ref-diffusion model to extract low-level feature inputs from the reference features to the main-diffusion model, which circumvents the different training\u0026nbsp; distributions of the other encoders and the main-diffusion model. In our compar- ative experiments, with respect to graphic alignment, our approach demonstrates performance similar to that of the other methods, but our approach uses sub- stantially smaller datasets and fewer computational resources during the training phase.Our code will be available at:https://gitee.com/yongzhenke/efe-sdg\u0026nbsp;\u003c/p\u003e","manuscriptTitle":"EFE-SDG: Efficient Feature Extraction of Finetuning-Free Model in Subject-Driven Generation","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-28 16:31:44","doi":"10.21203/rs.3.rs-7499053/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-12-08T06:55:07+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-30T15:07:13+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"167772229989324384885554631279083131902","date":"2025-11-26T02:54:29+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-23T12:32:14+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"44590290867540517008734204462754079865","date":"2025-10-20T03:22:28+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"117710500222959431522093365211018154083","date":"2025-10-17T03:03:40+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"324282953271666330320478638633033408390","date":"2025-10-15T01:06:53+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-10-15T00:45:16+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-10-11T14:29:11+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-09-11T10:04:33+00:00","index":"","fulltext":""},{"type":"submitted","content":"Multimedia Systems","date":"2025-08-31T07:32:28+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"multimedia-systems","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"mmsj","sideBox":"Learn more about [Multimedia Systems](http://link.springer.com/journal/530)","snPcode":"530","submissionUrl":"https://submission.nature.com/new-submission/530/3","title":"Multimedia Systems","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"147496e0-16e9-44cd-8feb-7d6337cf01fd","owner":[],"postedDate":"October 28th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2026-03-16T16:03:21+00:00","versionOfRecord":{"articleIdentity":"rs-7499053","link":"https://doi.org/10.1007/s00530-026-02248-3","journal":{"identity":"multimedia-systems","isVorOnly":false,"title":"Multimedia Systems"},"publishedOn":"2026-03-10 15:57:46","publishedOnDateReadable":"March 10th, 2026"},"versionCreatedAt":"2025-10-28 16:31:44","video":"","vorDoi":"10.1007/s00530-026-02248-3","vorDoiUrl":"https://doi.org/10.1007/s00530-026-02248-3","workflowStages":[]},"version":"v1","identity":"rs-7499053","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7499053","identity":"rs-7499053","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}