Textual-Visual Interaction for Enhanced Single Image Deraining using Adapter-Tuned VLMs

Textual-Visual Interaction for Enhanced Single Image Deraining using Adapter-Tuned VLMs | 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 Textual-Visual Interaction for Enhanced Single Image Deraining using Adapter-Tuned VLMs Qianfeng Yang, Pengpeng Li, Jiyu Jin, Guiyu Jin, Tianyu Song, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5715761/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 08 Mar, 2026 Read the published version in The Visual Computer → Version 1 posted 8 You are reading this latest preprint version Abstract This paper proposes a novel method called Textual-Visual Interaction for Enhanced Single Image Deraining using Adapter-Tuned VLMs (TVI-Derain). By leveraging the extensive textual knowledge from pretrained visual-language models (VLMs), we aim to improve the performance of single image deraining. To address the gap between VLMs and the restoration model, we introduce textual-aware intra-layer (TaIl) adapters that adapt the features of downstream data by capturing task-specific knowledge. Furthermore, a textual-visual feature interaction (TVI) module is designed to bridge the gap between textual and visual features, enabling reliable interaction. The proposed cross-attention feature interaction (CAFI) block within the TVI module effectively represents the interactive features. Semantic and degradation textual prompts are integrated as inputs to the text encoder to mitigate semantic disconnection arising from degraded samples. Extensive experimental results on benchmark datasets demonstrate that our method outperforms other competitive methods in terms of performance, showcasing its potential applications in automotive vision systems and surveillance and surveillance. The code will be released at github.com/ncfjd/TVI-Derain. Single image deraining Textual-visual feature interaction VLMs Adapter fine-tuning Prompt learning Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 08 Mar, 2026 Read the published version in The Visual Computer → Version 1 posted Editorial decision: Revision requested 04 Jun, 2025 Reviewers agreed at journal 19 May, 2025 Reviews received at journal 17 Feb, 2025 Reviewers agreed at journal 07 Feb, 2025 Reviewers invited by journal 30 Dec, 2024 Editor assigned by journal 27 Dec, 2024 Submission checks completed at journal 27 Dec, 2024 First submitted to journal 26 Dec, 2024 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-5715761","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":395095689,"identity":"db4a4757-6170-4b1c-b89f-4925d1f584a3","order_by":0,"name":"Qianfeng Yang","email":"","orcid":"","institution":"Dalian Polytechnic University","correspondingAuthor":false,"prefix":"","firstName":"Qianfeng","middleName":"","lastName":"Yang","suffix":""},{"id":395095690,"identity":"c0ee7ac7-3ab5-49a0-a162-d2e7b4304114","order_by":1,"name":"Pengpeng Li","email":"","orcid":"","institution":"Nanjing University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Pengpeng","middleName":"","lastName":"Li","suffix":""},{"id":395095691,"identity":"0d5a8336-b8b3-4468-9107-2887b690f85d","order_by":2,"name":"Jiyu Jin","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAz0lEQVRIiWNgGAWjYBACxmYQySPBwM+Q3AAWaCBai2RDIpFa4MDgALFamNt5zCQYZCzyjI8ntj3mYbCR3XCA+dkD/A4DaeGRKDY787DdmIchzXjDATZzA2K0JG67kdgmzcNwOHHDAR42CaK0bJ4B1vKfBC0bJMBaDhCjha3YIgGoZQbQL4ZzDJKNZx5mM8OrxbD/8MYbH3vqEvvbk489eFNhJ9t3vPkZfi0NHAYMiT1gNhswdoAUMz71QCDPwP6AgeEHTMsoGAWjYBSMAiwAAK1nQmyOri0FAAAAAElFTkSuQmCC","orcid":"","institution":"Dalian Polytechnic University","correspondingAuthor":true,"prefix":"","firstName":"Jiyu","middleName":"","lastName":"Jin","suffix":""},{"id":395095692,"identity":"b69f6ca3-feb3-44b8-bfbf-71d5894fbeb6","order_by":3,"name":"Guiyu Jin","email":"","orcid":"","institution":"Dalian Polytechnic University","correspondingAuthor":false,"prefix":"","firstName":"Guiyu","middleName":"","lastName":"Jin","suffix":""},{"id":395095693,"identity":"6391edbb-45da-4256-9bba-df0272f9a759","order_by":4,"name":"Tianyu Song","email":"","orcid":"","institution":"Dalian Maritime University","correspondingAuthor":false,"prefix":"","firstName":"Tianyu","middleName":"","lastName":"Song","suffix":""},{"id":395095694,"identity":"fe4b4aa6-3b42-45b6-91d1-9327c8b983c0","order_by":5,"name":"Shumin Fan","email":"","orcid":"","institution":"Dalian Maritime University","correspondingAuthor":false,"prefix":"","firstName":"Shumin","middleName":"","lastName":"Fan","suffix":""},{"id":395095695,"identity":"e4ba3ce4-e858-4bf2-a42d-c7dd3f485533","order_by":6,"name":"Hao Hou","email":"","orcid":"","institution":"Nanjing University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Hao","middleName":"","lastName":"Hou","suffix":""}],"badges":[],"createdAt":"2024-12-26 11:23:11","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5715761/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5715761/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s00371-026-04362-9","type":"published","date":"2026-03-08T15:59:23+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":104251630,"identity":"fadbefcb-324b-40f1-91cd-c4bf09726ccc","added_by":"auto","created_at":"2026-03-09 16:14:31","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3692262,"visible":true,"origin":"","legend":"","description":"","filename":"TVCmanuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5715761/v1_covered_ec82aa94-cc60-4331-8319-b32dbc86ff9f.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Textual-Visual Interaction for Enhanced Single Image Deraining using Adapter-Tuned VLMs","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":"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":"Single image deraining, Textual-visual feature interaction, VLMs, Adapter fine-tuning, Prompt learning","lastPublishedDoi":"10.21203/rs.3.rs-5715761/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5715761/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis paper proposes a novel method called Textual-Visual Interaction for Enhanced Single Image Deraining using Adapter-Tuned VLMs (TVI-Derain). By leveraging the extensive textual knowledge from pretrained visual-language models (VLMs), we aim to improve the performance of single image deraining. To address the gap between VLMs and the restoration model, we introduce textual-aware intra-layer (TaIl) adapters that adapt the features of downstream data by capturing task-specific knowledge. Furthermore, a textual-visual feature interaction (TVI) module is designed to bridge the gap between textual and visual features, enabling reliable interaction. The proposed cross-attention feature interaction (CAFI) block within the TVI module effectively represents the interactive features. Semantic and degradation textual prompts are integrated as inputs to the text encoder to mitigate semantic disconnection arising from degraded samples. Extensive experimental results on benchmark datasets demonstrate that our method outperforms other competitive methods in terms of performance, showcasing its potential applications in automotive vision systems and surveillance and\u003c/p\u003e\n\u003cp\u003esurveillance. The code will be released at github.com/ncfjd/TVI-Derain.\u003c/p\u003e","manuscriptTitle":"Textual-Visual Interaction for Enhanced Single Image Deraining using Adapter-Tuned VLMs","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-12-30 05:10:08","doi":"10.21203/rs.3.rs-5715761/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-06-04T15:30:13+00:00","index":"","fulltext":""},{"type":"reviewerAgreed","content":"214220465326743594273986232370297425484","date":"2025-05-19T08:07:02+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-02-18T03:47:37+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"74625988339838583930975740577784741089","date":"2025-02-07T08:31:36+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2024-12-30T15:12:55+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-12-27T18:44:09+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-12-27T15:00:55+00:00","index":"","fulltext":""},{"type":"submitted","content":"The Visual Computer","date":"2024-12-26T11:12:33+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":"397049ac-55d4-4119-b549-9b72fd3c1bbc","owner":[],"postedDate":"December 30th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2026-03-09T16:10:50+00:00","versionOfRecord":{"articleIdentity":"rs-5715761","link":"https://doi.org/10.1007/s00371-026-04362-9","journal":{"identity":"the-visual-computer","isVorOnly":false,"title":"The Visual Computer"},"publishedOn":"2026-03-08 15:59:23","publishedOnDateReadable":"March 8th, 2026"},"versionCreatedAt":"2024-12-30 05:10:08","video":"","vorDoi":"10.1007/s00371-026-04362-9","vorDoiUrl":"https://doi.org/10.1007/s00371-026-04362-9","workflowStages":[]},"version":"v1","identity":"rs-5715761","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5715761","identity":"rs-5715761","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}