LUNet: An enhanced upsampling fusion network with efficient self-attention for semantic segmentation

LUNet: An enhanced upsampling fusion network with efficient self-attention for semantic segmentation | 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 LUNet: An enhanced upsampling fusion network with efficient self-attention for semantic segmentation Yan Zhou, Haibin Zhou, Yin Yang, Jianxun Li, Richard Irampaye, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4255035/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 16 Sep, 2024 Read the published version in The Visual Computer → Version 1 posted 9 You are reading this latest preprint version Abstract Semantic segmentation is an essential aspect of many computer vision tasks. Self-attention (SA) based deep learning methods have shown impressive results in semantic segmentation by capturing long-range dependencies and contextual information. However, the standard SA module has high computational complexity, which limits its use in resource-constrained scenarios. This paper proposes a novel LUNet to improve semantic segmentation performance while addressing the computational challenges of SA. The lightweight self-attention plus (LSA++) module is introduced as a lightweight and efficient variant of the SA module. LSA++ uses compact feature representation and local position embedding to significantly reduce computational complexity while surpassing the accuracy of the standard SA module. Furthermore, to address the loss of edge details during decoding, we propose the Enhanced Upsampling Fusion Module (EUP-FM). This module comprises an enhanced upsampling module and a semantic vector-guided fusion mechanism. EUP-FM effectively recovers edge information and improves the precision of the segmentation map. Comprehensive experiments on PASCAL VOC 2012, Cityscapes, COCO, and SegPC 2021 demonstrate that LUNet outperforms all compared methods. It achieves superior runtime performance and accurate segmentation with excellent model generalization ability. The code is available at https://github.com/hbzhou530/LUNet . Semantic segmentation Lightweight model Upsampling fusion Self-attention Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 16 Sep, 2024 Read the published version in The Visual Computer → Version 1 posted Editorial decision: Revision requested 09 May, 2024 Reviews received at journal 04 May, 2024 Reviews received at journal 26 Apr, 2024 Reviewers agreed at journal 14 Apr, 2024 Reviewers agreed at journal 14 Apr, 2024 Reviewers invited by journal 14 Apr, 2024 Editor assigned by journal 13 Apr, 2024 Submission checks completed at journal 13 Apr, 2024 First submitted to journal 11 Apr, 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-4255035","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":291038728,"identity":"0596cba5-91e3-41ed-acb3-e4f18e89b499","order_by":0,"name":"Yan Zhou","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAyElEQVRIiWNgGAWjYJACZiCWgzDZSNBiDFFNipbEBqK1GBw/e/hzQcXh9A33ewwYPpQdZuCf3UBAy5m8BOMZZw7nzmzjMWCcce4wg8SdAwS0HMgxSOZtO5zbz8ZjwAxkMBhIJBDQcv6NwWGgynQ2kJa/RGm5kWPYDNSSwA/SwkiMFskbb4yZec6kG85sSys42HMunUfiBgEtfOdzjD/zVFjLGxw+vPHBjzJrOf4ZBLQoHEDigNg8+NUDgXwDQSWjYBSMglEw4gEA364/zaW6zMcAAAAASUVORK5CYII=","orcid":"","institution":"Xiangtan University","correspondingAuthor":true,"prefix":"","firstName":"Yan","middleName":"","lastName":"Zhou","suffix":""},{"id":291038729,"identity":"e7972193-9a71-4721-979e-103c46b044be","order_by":1,"name":"Haibin Zhou","email":"","orcid":"","institution":"Xiangtan University","correspondingAuthor":false,"prefix":"","firstName":"Haibin","middleName":"","lastName":"Zhou","suffix":""},{"id":291038730,"identity":"7cee3895-c5bd-47cc-a355-8131573d9776","order_by":2,"name":"Yin Yang","email":"","orcid":"","institution":"Xiangtan University","correspondingAuthor":false,"prefix":"","firstName":"Yin","middleName":"","lastName":"Yang","suffix":""},{"id":291038731,"identity":"b054a719-0af7-4a77-bc13-f0ec71e4af49","order_by":3,"name":"Jianxun Li","email":"","orcid":"","institution":"Shanghai Jiao Tong University","correspondingAuthor":false,"prefix":"","firstName":"Jianxun","middleName":"","lastName":"Li","suffix":""},{"id":291038733,"identity":"93f84e36-488d-401b-a035-944ec754609b","order_by":4,"name":"Richard Irampaye","email":"","orcid":"","institution":"Xiangtan University","correspondingAuthor":false,"prefix":"","firstName":"Richard","middleName":"","lastName":"Irampaye","suffix":""},{"id":291038734,"identity":"2cc20950-eff8-4ed1-bf49-d7e6613d0832","order_by":5,"name":"Dongli Wang","email":"","orcid":"","institution":"Xiangtan University","correspondingAuthor":false,"prefix":"","firstName":"Dongli","middleName":"","lastName":"Wang","suffix":""},{"id":291038736,"identity":"8f939be7-77f9-4fc5-9be2-26e68e17c78b","order_by":6,"name":"Zhengpeng Zhang","email":"","orcid":"","institution":"Xiangtan University","correspondingAuthor":false,"prefix":"","firstName":"Zhengpeng","middleName":"","lastName":"Zhang","suffix":""}],"badges":[],"createdAt":"2024-04-12 03:12:22","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4255035/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4255035/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s00371-024-03590-1","type":"published","date":"2024-09-16T15:57:51+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":65104707,"identity":"2cbdd3a4-4f41-418c-bb06-59f64826fe12","added_by":"auto","created_at":"2024-09-23 16:14:18","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3929242,"visible":true,"origin":"","legend":"","description":"","filename":"main.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4255035/v1_covered_90f6caf0-b55f-4612-9661-f598e7331acd.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"LUNet: An enhanced upsampling fusion network with efficient self-attention for semantic segmentation","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":"Semantic segmentation, Lightweight model, Upsampling fusion, Self-attention","lastPublishedDoi":"10.21203/rs.3.rs-4255035/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4255035/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Semantic segmentation is an essential aspect of many computer vision tasks. Self-attention (SA) based deep learning methods have shown impressive results in semantic segmentation by capturing long-range dependencies and contextual information. However, the standard SA module has high computational complexity, which limits its use in resource-constrained scenarios. This paper proposes a novel LUNet to improve semantic segmentation performance while addressing the computational challenges of SA. The lightweight self-attention plus (LSA++) module is introduced as a lightweight and efficient variant of the SA module. LSA++ uses compact feature representation and local position embedding to significantly reduce computational complexity while surpassing the accuracy of the standard SA module. Furthermore, to address the loss of edge details during decoding, we propose the Enhanced Upsampling Fusion Module (EUP-FM). This module comprises an enhanced upsampling module and a semantic vector-guided fusion mechanism. EUP-FM effectively recovers edge information and improves the precision of the segmentation map. Comprehensive experiments on PASCAL VOC 2012, Cityscapes, COCO, and SegPC 2021 demonstrate that LUNet outperforms all compared methods. It achieves superior runtime performance and accurate segmentation with excellent model generalization ability. The code is available at https://github.com/hbzhou530/LUNet.","manuscriptTitle":"LUNet: An enhanced upsampling fusion network with efficient self-attention for semantic segmentation","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-04-17 18:42:49","doi":"10.21203/rs.3.rs-4255035/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2024-05-09T21:27:44+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-05-05T01:16:41+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-04-26T15:03:50+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"4f7ad51f-91fc-4c1f-b26b-6d2aaac066e5","date":"2024-04-14T16:19:38+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"a3ba2793-5ea2-4b1f-a3e8-0fc2b7829c20","date":"2024-04-14T11:23:03+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2024-04-14T10:24:51+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-04-13T13:20:24+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-04-13T05:59:20+00:00","index":"","fulltext":""},{"type":"submitted","content":"The Visual Computer","date":"2024-04-12T03:09:42+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":"da92499d-a760-44c6-b465-f26a145c4472","owner":[],"postedDate":"April 17th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2024-09-23T16:11:30+00:00","versionOfRecord":{"articleIdentity":"rs-4255035","link":"https://doi.org/10.1007/s00371-024-03590-1","journal":{"identity":"the-visual-computer","isVorOnly":false,"title":"The Visual Computer"},"publishedOn":"2024-09-16 15:57:51","publishedOnDateReadable":"September 16th, 2024"},"versionCreatedAt":"2024-04-17 18:42:49","video":"","vorDoi":"10.1007/s00371-024-03590-1","vorDoiUrl":"https://doi.org/10.1007/s00371-024-03590-1","workflowStages":[]},"version":"v1","identity":"rs-4255035","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4255035","identity":"rs-4255035","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}