EID-DETR: Efficient real-time Transformer with enhanced defect detection for UAV inspections

EID-DETR: Efficient real-time Transformer with enhanced defect detection for UAV inspections | 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 EID-DETR: Efficient real-time Transformer with enhanced defect detection for UAV inspections Qingan Yao, Hongxin Wang, Hongmei Wang, Yuncong Feng, Hongtao Yang, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9432286/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract In unmanned aerial vehicle power line inspections, minor insulator defects are easily obscured by complex high-frequency background noise. Meanwhile, existing high-precision models incur substantial computational overhead, creating a single-frame processing bottleneck that depletes the algorithmic headroom essential for multi-UAV collaborative deployments. To resolve this, we propose EID-DETR, an efficient real-time object detection architecture. Its core CSP-LGLB backbone integrates cross-stage partial gradient flows and local-global linear attention to minimize computational redundancy while capturing long-range dependencies. Building upon this, an intra-scale feature interaction module based on an efficient prompt-guided operator explicitly filters background noise, dynamically directing the computational budget toward critical defect features. Subsequently, an enhanced cross-scale feature fusion module incorporating an SGAF mechanism and GSConvE module is formulated to bridge semantic gaps and preserve high-frequency details. Requiring merely 13.9M parameters and 38.2 GFLOPs, EID-DETR achieves an [email protected] of 93.6% while operating at a blistering 238.6 FPS on a hybrid benchmark dataset. Ultimately, EID-DETR establishes a dominant Pareto frontier between inference speed and accuracy, delivering a robust, hardware-friendly solution for real-time aerial monitoring. Object detection Lightweight model DETR Defect detection Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted 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-9432286","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":629600682,"identity":"b2abdaf5-f3dd-4c13-ad2c-78ea56ed5794","order_by":0,"name":"Qingan Yao","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAArUlEQVRIiWNgGAWjYHACxgdQhgHRWphBSiVI0sImQZoWg+OHj1X8qDlcx8DevE2CoeYOEVrOpKXd7Dl2WIKB51iZBMOxZ0RoOZBjdpuxAahFIsdMAsggQsv5N2bFYC3yb4jVciPHjBliCw+RWiRvPEuW7DmWLtnGk1ZskXCMCC1855MPfvhRY83Pz354440PNURoUTgAZbCBiATCGhgY5BuIUTUKRsEoGAUjGwAAyYk2KzkV1/EAAAAASUVORK5CYII=","orcid":"","institution":"Changchun University of Technology","correspondingAuthor":true,"prefix":"","firstName":"Qingan","middleName":"","lastName":"Yao","suffix":""},{"id":629600699,"identity":"78410017-786f-4c96-9bbd-a1362a979ec8","order_by":1,"name":"Hongxin Wang","email":"","orcid":"","institution":"Changchun University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Hongxin","middleName":"","lastName":"Wang","suffix":""},{"id":629600713,"identity":"ce196acd-f903-49f3-acef-4a47348e1bc0","order_by":2,"name":"Hongmei Wang","email":"","orcid":"","institution":"Changchun University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Hongmei","middleName":"","lastName":"Wang","suffix":""},{"id":629600721,"identity":"9429c5e5-2621-490f-aa97-75c9849ad9e3","order_by":3,"name":"Yuncong Feng","email":"","orcid":"","institution":"Changchun University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Yuncong","middleName":"","lastName":"Feng","suffix":""},{"id":629600725,"identity":"6e7a491f-7d28-40b8-814c-5f1fd94aae40","order_by":4,"name":"Hongtao Yang","email":"","orcid":"","institution":"Changchun University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Hongtao","middleName":"","lastName":"Yang","suffix":""},{"id":629600727,"identity":"f23397e6-6a9e-48eb-8b9c-f2c35147bf29","order_by":5,"name":"Miao Han","email":"","orcid":"","institution":"Changchun University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Miao","middleName":"","lastName":"Han","suffix":""},{"id":629600731,"identity":"f12ce2c8-b1b6-4f2e-b88f-9d02561f232a","order_by":6,"name":"Lili Liu","email":"","orcid":"","institution":"Changchun University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Lili","middleName":"","lastName":"Liu","suffix":""}],"badges":[],"createdAt":"2026-04-16 02:53:16","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9432286/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9432286/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108181183,"identity":"c7a1a5a7-2141-42ea-8b86-4d20e4fc5d8b","added_by":"auto","created_at":"2026-04-30 08:58:18","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":7447661,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9432286/v1_covered_feea9d43-16a8-49f8-899b-f73aced06c6e.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"EID-DETR: Efficient real-time Transformer with enhanced defect detection for UAV inspections","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":"Object detection, Lightweight model, DETR, Defect detection","lastPublishedDoi":"10.21203/rs.3.rs-9432286/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9432286/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"In unmanned aerial vehicle power line inspections, minor insulator defects are easily obscured by complex high-frequency background noise. Meanwhile, existing high-precision models incur substantial computational overhead, creating a single-frame processing bottleneck that depletes the algorithmic headroom essential for multi-UAV collaborative deployments. To resolve this, we propose EID-DETR, an efficient real-time object detection architecture. Its core CSP-LGLB backbone integrates cross-stage partial gradient flows and local-global linear attention to minimize computational redundancy while capturing long-range dependencies. Building upon this, an intra-scale feature interaction module based on an efficient prompt-guided operator explicitly filters background noise, dynamically directing the computational budget toward critical defect features. Subsequently, an enhanced cross-scale feature fusion module incorporating an SGAF mechanism and GSConvE module is formulated to bridge semantic gaps and preserve high-frequency details. Requiring merely 13.9M parameters and 38.2 GFLOPs, EID-DETR achieves an [email protected] of 93.6\\% while operating at a blistering 238.6 FPS on a hybrid benchmark dataset. Ultimately, EID-DETR establishes a dominant Pareto frontier between inference speed and accuracy, delivering a robust, hardware-friendly solution for real-time aerial monitoring.","manuscriptTitle":"EID-DETR: Efficient real-time Transformer with enhanced defect detection for UAV inspections","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-27 07:53:12","doi":"10.21203/rs.3.rs-9432286/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":"d99bddca-cf0c-4c51-8c2b-458d25c00fc5","owner":[],"postedDate":"April 27th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-04-28T15:25:09+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-27 07:53:12","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9432286","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9432286","identity":"rs-9432286","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}