Efficient YOLOv12 for Multi-Scale Object Detection

Efficient YOLOv12 for Multi-Scale Object Detection | 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PDF Research Article Efficient YOLOv12 for Multi-Scale Object Detection Mahdi Zarkoosh This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7508677/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 This paper presents five optimized variants of YOLOv12, designed to improve computational efficiency while maintaining detection accuracy. The modifications involve pruning and reconfiguring the YOLOv12 model’s structure to enhance resource utilization. Each proposed model is tailored to specific object sizes: YOLOv12-Small (small objects), YOLOv12-Medium (medium objects), YOLOv12-Large (large objects), YOLOv12-SM (small and medium objects), and YOLOv12-ML (medium and large objects). To select the proper proposed model based on dataset characteristics, we developed a simple program that identifies the dataset’s object size distribution}. Experimental evaluations with the baseline YOLOv12 model demonstrate that the proposed models achieve superior computational efficiency while preserving detection accuracy. In addition, the proposed models have been compared with YOLOv8, YOLOv10, and YOLOv11, demonstrating improvements in model size and GFLOPs while achieving comparable accuracy. However, its inference time and power consumption are moderately greater than those of the earlier YOLO versions. Artificial Intelligence and Machine Learning YOLOv12 YOLOv11 YOLOv10 YOLOv8 Computer Vision Pruning Computational Efficiency Full Text Additional Declarations The authors declare no competing interests. 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-7508677","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":508552309,"identity":"e4d3b946-248c-427b-b11f-3e3d0c0fde2b","order_by":0,"name":"Mahdi Zarkoosh","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAz0lEQVRIiWNgGAWjYJACxgYgYcAOJi1I0cJzAERKkKJFIgHEJkKLbv/hZx9nVNjYm0s+v7rhR4EEA397dwJeLWYHjhnP3HAmLXHn7Jyymz1Ah0mcObsBv5aDDcaMD9sOJxjczkm7wQPUYiCRS0DLYfbPIC32BjfPpN38Q5SWYzzGjBvbDjNuuMF+7DZxtpzhKWacAfTLhjM5bLdlDCR4CPvl/PHNjD3AEDM4fvzZzTd/bOT423vxa0ECPAZgkljlIMD+gBTVo2AUjIJRMIIAAO4SSzifhYmfAAAAAElFTkSuQmCC","orcid":"","institution":"","correspondingAuthor":true,"prefix":"","firstName":"Mahdi","middleName":"","lastName":"Zarkoosh","suffix":""}],"badges":[],"createdAt":"2025-09-01 12:55:24","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-7508677/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7508677/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":90460309,"identity":"b5723bcb-af51-4439-b38f-63e266f97cca","added_by":"auto","created_at":"2025-09-03 03:43:08","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2512151,"visible":true,"origin":"","legend":"","description":"","filename":"EfficientYOLOv12forMultiScaleObjectDetection.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7508677/v1_covered_1aff808c-23c1-4b10-9b56-a3fe9871e254.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eEfficient YOLOv12 for Multi-Scale Object Detection\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Independent Researcher","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":"YOLOv12, YOLOv11, YOLOv10, YOLOv8, Computer Vision, Pruning, Computational Efficiency","lastPublishedDoi":"10.21203/rs.3.rs-7508677/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7508677/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis paper presents five optimized variants of YOLOv12, designed to improve computational efficiency while maintaining detection accuracy. The modifications involve pruning and reconfiguring the YOLOv12 model’s structure to enhance resource utilization. Each proposed model is tailored to specific object sizes: YOLOv12-Small (small objects), YOLOv12-Medium (medium objects), YOLOv12-Large (large objects), YOLOv12-SM (small and medium objects), and YOLOv12-ML (medium and large objects). To select the proper proposed model based on dataset characteristics, we developed a simple program that identifies the dataset’s object size distribution}. Experimental evaluations with the baseline YOLOv12 model demonstrate that the proposed models achieve superior computational efficiency while preserving detection accuracy. In addition, the proposed models have been compared with YOLOv8, YOLOv10, and YOLOv11, demonstrating improvements in model size and GFLOPs while achieving comparable accuracy. However, its inference time and power consumption are moderately greater than those of the earlier YOLO versions.\u003c/p\u003e","manuscriptTitle":"Efficient YOLOv12 for Multi-Scale Object Detection","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-03 03:34:57","doi":"10.21203/rs.3.rs-7508677/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":"216c3fbc-0222-4e68-ae30-79cb1ce41ea0","owner":[],"postedDate":"September 3rd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":54010721,"name":"Artificial Intelligence and Machine Learning"}],"tags":[],"updatedAt":"2025-09-03T03:34:57+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-03 03:34:57","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7508677","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7508677","identity":"rs-7508677","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}