Enhancing Tiny Object Detection without Fine Tuning: Dynamic Adaptive Guided Object Inference Slicing Framework with Latest YOLO Models and RT-DETR Transformer

Enhancing Tiny Object Detection without Fine Tuning: Dynamic Adaptive Guided Object Inference Slicing Framework with Latest YOLO Models and RT-DETR Transformer | 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 Enhancing Tiny Object Detection without Fine Tuning: Dynamic Adaptive Guided Object Inference Slicing Framework with Latest YOLO Models and RT-DETR Transformer MUHAMMAD MUZAMMUL, Xuewei LI, Xi Li This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5780163/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 Tiny Object Detection (TOD) in high-resolution imagery presents persistent challenges in computer vision, including low resolution, occlusion, and cluttered backgrounds. This paper introduces the Dynamic Adaptive Guided Object Inference Slicing (GOIS) framework, a novel two-stage adaptive slicing approach that dynamically reallocates computational resources to Regions of Interest (ROIs). This methodology significantly enhances detection precision and efficiency, achieving 3–4× improvements in Average Precision (AP) and Average Recall (AR) metrics for small objects. Additionally, the framework demonstrates substantial gains of 50–60% across other metrics, ensuring robust performance across various object scales. While slight declines in large-object detection were noted in specific scenarios, GOIS consistently excels in detecting small and medium-sized objects, effectively addressing critical challenges inherent to TOD. The GOIS framework integrates adaptive slicing, multi-scale representation, and context-aware modeling, surpassing the limitations of static slicing methods by mitigating boundary artifacts and optimizing computational efficiency. Its architecture-agnostic design allows seamless integration with diverse state-of-the-art detection models, including YOLO11, RT-DETR-L, and YOLOv8n, without requiring extensive retraining. Rigorous validation on the VisDrone2019-DET dataset, supplemented by evaluations on low-resolution images, video streams, and live camera feeds, highlights GOIS’s transformative potential. These findings establish its applicability to critical domains such as UAV-based surveillance, autonomous navigation, and precision diagnostics. The code and results are publicly available at https: // github. com/ MMUZAMMUL/ GOIS with a live demonstration accessible at https: // youtu. be/ T5t5eb_ w0S4 . Computational Neuroscience Tiny Object Detection Guided Object Inference Slicing dynamic slicing methodology UAV-based surveillance high-resolution imagery analysis computational efficiency in object detection 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-5780163","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":398789141,"identity":"74744da3-2422-4d5f-94fe-54662d3190a6","order_by":0,"name":"MUHAMMAD MUZAMMUL","email":"","orcid":"","institution":"Zhejiang University,China","correspondingAuthor":false,"prefix":"","firstName":"MUHAMMAD","middleName":"","lastName":"MUZAMMUL","suffix":""},{"id":398789142,"identity":"3283aec2-7b39-45e5-8af1-4d08aa3f5ff4","order_by":1,"name":"Xuewei LI","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA3klEQVRIiWNgGAWjYNACAwsefiBpAGQyNhCpRYJHsoE0LQwSDAYHGEA6iNBicCP5mMSHAgkZ4/OHNxTzMNjIbjjA/OwBPi2SM9LSJGcAHWZ2I63AmIchzXjDATZzA3xa+CVyzG7zgLXwGAC1HE7ccICHTQKfFjaJ/G+3/wC1GPefAWn5T1gL0Ba226AQM2DIAWk5QFiLZM8z8589QC0SQL8YzjFINp55mM0MrxaD48mPDX78sbHn7z+8zeBNhZ1s3/HmZ3i1oPjLABw1zMSqB6l9QILiUTAKRsEoGEEAAOgTQJDNrcCSAAAAAElFTkSuQmCC","orcid":"","institution":"Shanghai Dianji University,China","correspondingAuthor":true,"prefix":"","firstName":"Xuewei","middleName":"","lastName":"LI","suffix":""},{"id":398791852,"identity":"264ee4f2-f248-4e9c-85d7-c3b1bf019a5d","order_by":2,"name":"Xi Li","email":"","orcid":"","institution":"Zhejiang University,China","correspondingAuthor":false,"prefix":"","firstName":"Xi","middleName":"","lastName":"Li","suffix":""}],"badges":[],"createdAt":"2025-01-07 10:02:37","currentVersionCode":1,"declarations":{"humanSubjects":true,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":true,"humanSubjectConsent":true,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-5780163/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5780163/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":73240896,"identity":"c3c02b50-868d-44d6-9dd7-f0219e57eb4c","added_by":"auto","created_at":"2025-01-08 06:04:23","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":5869907,"visible":true,"origin":"","legend":"","description":"","filename":"NEUCOMD24102152.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5780163/v1_covered_d6e01965-1497-4967-884c-db891ef679c1.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eEnhancing Tiny Object Detection without Fine Tuning: Dynamic Adaptive Guided Object Inference Slicing Framework with Latest YOLO Models and RT-DETR Transformer\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Zhejiang University","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":"Tiny Object Detection, Guided Object Inference Slicing, dynamic slicing methodology, UAV-based surveillance, high-resolution imagery analysis, computational efficiency in object detection","lastPublishedDoi":"10.21203/rs.3.rs-5780163/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5780163/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eTiny Object Detection (TOD) in high-resolution imagery presents persistent challenges in computer vision, including low resolution, occlusion, and cluttered backgrounds. This paper introduces the Dynamic Adaptive Guided Object Inference Slicing (GOIS) framework, a novel two-stage adaptive slicing approach that dynamically reallocates computational resources to Regions of Interest (ROIs). This methodology significantly enhances detection precision and efficiency, achieving 3–4× improvements in Average Precision (AP) and Average Recall (AR) metrics for small objects. Additionally, the framework demonstrates substantial gains of 50–60% across other metrics, ensuring robust performance across various object scales. While slight declines in large-object detection were noted in specific scenarios, GOIS consistently excels in detecting small and medium-sized objects, effectively addressing critical challenges inherent to TOD. The GOIS framework integrates adaptive slicing, multi-scale representation, and context-aware modeling, surpassing the limitations of static slicing methods by mitigating boundary artifacts and optimizing computational efficiency. Its architecture-agnostic design allows seamless integration with diverse state-of-the-art detection models, including YOLO11, RT-DETR-L, and YOLOv8n, without requiring extensive retraining. Rigorous validation on the VisDrone2019-DET dataset, supplemented by evaluations on low-resolution images, video streams, and live camera feeds, highlights GOIS’s transformative potential. These findings establish its applicability to critical domains such as UAV-based surveillance, autonomous navigation, and precision diagnostics. The code and results are publicly available at https: // github. com/ MMUZAMMUL/ GOIS with a live demonstration accessible at https: // youtu. be/ T5t5eb_ w0S4 .\u003c/p\u003e","manuscriptTitle":"Enhancing Tiny Object Detection without Fine Tuning: Dynamic Adaptive Guided Object Inference Slicing Framework with Latest YOLO Models and RT-DETR Transformer","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-01-08 05:40:14","doi":"10.21203/rs.3.rs-5780163/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":"bc27f298-fb47-4b27-95f2-62a7ded85f05","owner":[],"postedDate":"January 8th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":42484761,"name":"Computational Neuroscience"}],"tags":[],"updatedAt":"2025-01-08T05:40:14+00:00","versionOfRecord":[],"versionCreatedAt":"2025-01-08 05:40:14","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5780163","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5780163","identity":"rs-5780163","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}