YOLOv6+: Simple and Optimized Object Detection Model for INT8 quantized inference on mobile devices

YOLOv6+: Simple and Optimized Object Detection Model for INT8 quantized inference on mobile devices | 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 YOLOv6+: Simple and Optimized Object Detection Model for INT8 quantized inference on mobile devices Hyeon-Cheol Moon, Seungho Lee, Jinwoo Jeong, Sungjei Kim This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5738660/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 28 May, 2025 Read the published version in Signal, Image and Video Processing → Version 1 posted 10 You are reading this latest preprint version Abstract The You Only Look Once (YOLO) series stands out for its exceptional scalability, enabling seamless deployment on a variety of diverse software and hardware platforms. This scalability has driven its utilization in numerous industrial sites. Recently, there has been an increasing focus on developing quantization-friendly architectures, especially for INT8 inference, to support real-time processing on low-power devices such as mobile platforms. In this paper, we propose the simple and novel approach to enhance the performance of the YOLOv6 model, a widely used object detector in industrial applications, by incorporating skip connections in selected re-parameterization blocks to achieve a quantization-friendly architecture. In addition, we introduce an regression normalization method to address the performance degradation in the head part that often occurs during the TFLite INT8 conversion for mobile environments. The proposed YOLOv6+ architecture outperforms the original YOLOv6 and its successor YOLOv8 by achieving comparable speed in FP/INT8 precision inference while improving mAP performance and enhancing quantization-friendliness. Furthermore, the regression normalization method effectively mitigates performance degradation during TFLite INT8 conversion and is verified to be applicable to other recently developed YOLO series models. object detection YOLOv6 On-device AI Mobile AI TFLite Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 28 May, 2025 Read the published version in Signal, Image and Video Processing → Version 1 posted Editorial decision: Revision requested 30 Mar, 2025 Reviews received at journal 29 Mar, 2025 Reviews received at journal 12 Mar, 2025 Reviewers agreed at journal 12 Mar, 2025 Reviewers agreed at journal 11 Mar, 2025 Reviewers agreed at journal 10 Mar, 2025 Reviewers agreed at journal 10 Mar, 2025 Reviewers invited by journal 10 Mar, 2025 Submission checks completed at journal 10 Mar, 2025 First submitted to journal 07 Mar, 2025 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-5738660","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":436064891,"identity":"686af269-ec96-454c-9551-ad16f7ce2922","order_by":0,"name":"Hyeon-Cheol Moon","email":"","orcid":"","institution":"Korea Electronics Technology Institutes","correspondingAuthor":false,"prefix":"","firstName":"Hyeon-Cheol","middleName":"","lastName":"Moon","suffix":""},{"id":436064892,"identity":"c0666191-ddd9-4f05-acf8-b62ffa65bc72","order_by":1,"name":"Seungho Lee","email":"","orcid":"","institution":"COCONSOFT","correspondingAuthor":false,"prefix":"","firstName":"Seungho","middleName":"","lastName":"Lee","suffix":""},{"id":436064893,"identity":"753665bd-95d0-49b8-a386-a6a70e9345c1","order_by":2,"name":"Jinwoo Jeong","email":"","orcid":"","institution":"Korea Electronics Technology Institutes","correspondingAuthor":false,"prefix":"","firstName":"Jinwoo","middleName":"","lastName":"Jeong","suffix":""},{"id":436064894,"identity":"a6ddff67-d145-4c01-9eb1-6a6d905105dc","order_by":3,"name":"Sungjei Kim","email":"data:image/png;base64,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","orcid":"","institution":"Korea University of Technology and Education","correspondingAuthor":true,"prefix":"","firstName":"Sungjei","middleName":"","lastName":"Kim","suffix":""}],"badges":[],"createdAt":"2024-12-31 02:53:10","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5738660/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5738660/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s11760-025-04234-0","type":"published","date":"2025-05-28T15:57:33+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":83782887,"identity":"e7bcc2e5-e521-4b69-b1d9-36cc14ccbe4b","added_by":"auto","created_at":"2025-06-02 16:08:04","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1560089,"visible":true,"origin":"","legend":"","description":"","filename":"yolov6plusr3.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5738660/v1_covered_e9d1b55d-3a2e-4217-af0c-d925589ca6b7.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"YOLOv6+: Simple and Optimized Object Detection Model for INT8 quantized inference on mobile devices","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":"signal-image-and-video-processing","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"sivp","sideBox":"Learn more about [Signal, Image and Video Processing](http://link.springer.com/journal/11760)","snPcode":"11760","submissionUrl":"https://submission.nature.com/new-submission/11760/3","title":"Signal, Image and Video Processing","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"object detection, YOLOv6, On-device AI, Mobile AI, TFLite","lastPublishedDoi":"10.21203/rs.3.rs-5738660/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5738660/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"The You Only Look Once (YOLO) series stands out for its exceptional scalability, enabling seamless deployment on a variety of diverse software and hardware platforms. This scalability has driven its utilization in numerous industrial sites. Recently, there has been an increasing focus on developing quantization-friendly architectures, especially for INT8 inference, to support real-time processing on low-power devices such as mobile platforms. In this paper, we propose the simple and novel approach to enhance the performance of the YOLOv6 model, a widely used object detector in industrial applications, by incorporating skip connections in selected re-parameterization blocks to achieve a quantization-friendly architecture. In addition, we introduce an regression normalization method to address the performance degradation in the head part that often occurs during the TFLite INT8 conversion for mobile environments. The proposed YOLOv6+ architecture outperforms the original YOLOv6 and its successor YOLOv8 by achieving comparable speed in FP/INT8 precision inference while improving mAP performance and enhancing quantization-friendliness. Furthermore, the regression normalization method effectively mitigates performance degradation during TFLite INT8 conversion and is verified to be applicable to other recently developed YOLO series models.","manuscriptTitle":"YOLOv6+: Simple and Optimized Object Detection Model for INT8 quantized inference on mobile devices","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-04-25 17:54:48","doi":"10.21203/rs.3.rs-5738660/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-03-31T02:28:04+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-03-29T21:43:11+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-03-12T17:25:45+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"317585635168984848210347733171674783185","date":"2025-03-12T17:22:34+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"318509124462558450096311281837674241238","date":"2025-03-11T17:20:34+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"334597671950310707642483379033280253114","date":"2025-03-11T02:31:34+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"222473354954690258139449865443507716939","date":"2025-03-10T18:46:32+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-03-10T18:43:34+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-03-10T08:19:17+00:00","index":"","fulltext":""},{"type":"submitted","content":"Signal, Image and Video Processing","date":"2025-03-07T16:02:31+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"signal-image-and-video-processing","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"sivp","sideBox":"Learn more about [Signal, Image and Video Processing](http://link.springer.com/journal/11760)","snPcode":"11760","submissionUrl":"https://submission.nature.com/new-submission/11760/3","title":"Signal, Image and Video Processing","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"94128f69-b9fe-4829-b648-bcecb222c0c4","owner":[],"postedDate":"April 25th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-06-02T16:01:38+00:00","versionOfRecord":{"articleIdentity":"rs-5738660","link":"https://doi.org/10.1007/s11760-025-04234-0","journal":{"identity":"signal-image-and-video-processing","isVorOnly":false,"title":"Signal, Image and Video Processing"},"publishedOn":"2025-05-28 15:57:33","publishedOnDateReadable":"May 28th, 2025"},"versionCreatedAt":"2025-04-25 17:54:48","video":"","vorDoi":"10.1007/s11760-025-04234-0","vorDoiUrl":"https://doi.org/10.1007/s11760-025-04234-0","workflowStages":[]},"version":"v1","identity":"rs-5738660","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5738660","identity":"rs-5738660","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}