HP-FLDet: A High Performance Multi-task Joint Face and Landmark Detector

HP-FLDet: A High Performance Multi-task Joint Face and Landmark Detector | 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 HP-FLDet: A High Performance Multi-task Joint Face and Landmark Detector Xueli Liu, Feng Wang, Yinlong Liu, Kexue Fu, Xiao Chen, Xiaobo Yang, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-3926744/v2 This work is licensed under a CC BY 4.0 License Status: Posted Version 2 posted You are reading this latest preprint version Show more versions Abstract Face detection and landmark localization are usually the fundamental and important steps of facial analysis. The accuracy degradation of any task has impacts on the accuracy and robustness of the downstream analysis. In most facial analysis systems, face detection and landmark detection, as two independent tasks, are predicted sequentially with single-task detectors respectively, which suffer from the higher complexity and poorer consistency. In contrast, multi-task joint face and landmark detectors are more efficient and have more shared features. However, when two tasks are predicted in one model, it is difficult to improve the accuracy of all tasks simultaneously, due to the different requirements of the different targets to be detected. In our opinion, the multi-task detection model, integrating the perception ability of both scale variation and detailed texture information for face and landmark detection, is arduous but important in unconstrained settings. In this paper, a High-Performance joint Face and Landmark detector(HP-FLDet) is proposed, which can predict facial bounding boxes and landmarks in single stage and improve the accuracy of both face detection and landmark localization simultaneously. In this detector, the context inception block and scale-aware multi-task strategies are designed to resolve extreme scale variance and improve the perception ability of detailed texture information. It demonstrates superior performance on both face and landmark detection datasets including AFW, PASCAL Face, FDDB, Wider Face and AFLW. Especially, it achieves SOTA results on the most challenging WIDER FACE dataset with 0.970(Easy), 0.963(Medium), 0.921(Hard) mAP on the test subset, which outperforms all state-of-the-art rivals. Face detection Landmark detection Scale-aware Multi-task Single-stage mixture dense regression high resolution heatmap Full Text Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 2 posted You are reading this latest preprint version Show more versions 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-3926744","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":271270197,"identity":"c0e42848-bb46-44d8-89a9-d3c9ea0dd37a","order_by":0,"name":"Xueli Liu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA4ElEQVRIiWNgGAWjYBACPmYwZcHDz3yAQQLMPkBACxtEiwSPZFsCsVoglASDwTGitbDzmEnzVEjIGB/jMbzxcweDHN+NBMbPBXgdxmNszHNGgsfsGI+xZe8ZBmPJGwnM0jPwazF8zNsG1HK/x0yCt40hccONBKAgfi0Gh3n/SfAYt/GYSf5tY6gnRgvQlgYJHgM2oKeAtiQYENbCVmw455gEj8QxtmJr2TYJw5lnHjZL49PCz394m8SbGht7/jbmjTffttnI8x1PPvgZnxZ0AIoaxgYSNIyCUTAKRsEowAYAdUE6GKzU+jUAAAAASUVORK5CYII=","orcid":"","institution":"ENT Institute and Department of Otolaryngology, Eye \\\u0026 ENT Hospital of Fudan University","correspondingAuthor":true,"prefix":"","firstName":"Xueli","middleName":"","lastName":"Liu","suffix":""},{"id":271270198,"identity":"63c08bcd-a99b-4de8-8ddb-10523674c6de","order_by":1,"name":"Feng Wang","email":"","orcid":"","institution":"Department of Urology, General Hospital of Ningxia Medical University","correspondingAuthor":false,"prefix":"","firstName":"Feng","middleName":"","lastName":"Wang","suffix":""},{"id":271270199,"identity":"dc070a39-0e91-4783-9957-9ce3abb5aac1","order_by":2,"name":"Yinlong Liu","email":"","orcid":"","institution":"Faculty of Data Science, City University of Macau","correspondingAuthor":false,"prefix":"","firstName":"Yinlong","middleName":"","lastName":"Liu","suffix":""},{"id":271270200,"identity":"09dd4325-501d-406f-a674-80d4bbc04606","order_by":3,"name":"Kexue Fu","email":"","orcid":"","institution":"Academy for Engineering and Technology, Fudan University","correspondingAuthor":false,"prefix":"","firstName":"Kexue","middleName":"","lastName":"Fu","suffix":""},{"id":271270201,"identity":"71db31fb-a3cf-4696-818d-7df2d4a3693a","order_by":4,"name":"Xiao Chen","email":"","orcid":"","institution":"ENT Institute and Department of Otolaryngology, Eye \\\u0026 ENT Hospital of Fudan University","correspondingAuthor":false,"prefix":"","firstName":"Xiao","middleName":"","lastName":"Chen","suffix":""},{"id":271270202,"identity":"4110c39d-7512-4677-8686-8f9014e99a93","order_by":5,"name":"Xiaobo Yang","email":"","orcid":"","institution":"Department of Urology, General Hospital of Ningxia Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xiaobo","middleName":"","lastName":"Yang","suffix":""},{"id":433720214,"identity":"0c1efc80-7aa2-453d-b2ba-e0a7a1c4b339","order_by":6,"name":"Xinrong Chen","email":"","orcid":"","institution":"Academy for Engineering and Technology, Fudan University","correspondingAuthor":false,"prefix":"","firstName":"Xinrong","middleName":"","lastName":"Chen","suffix":""}],"badges":[],"createdAt":"2024-02-04 07:59:36","currentVersionCode":2,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-3926744/v2","doiUrl":"https://doi.org/10.21203/rs.3.rs-3926744/v2","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":79457508,"identity":"a55fcdac-86ee-4a18-a179-a8f86e2edb86","added_by":"auto","created_at":"2025-03-28 16:21:37","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3069704,"visible":true,"origin":"","legend":"","description":"","filename":"elsarticletemplatenum.pdf","url":"https://assets-eu.researchsquare.com/files/rs-3926744/v2_covered_f403de71-5a88-437c-808e-64899e34a0d6.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eHP-FLDet: A High Performance Multi-task Joint Face and Landmark Detector\u003c/p\u003e","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":"Face detection, Landmark detection, Scale-aware, Multi-task, Single-stage, mixture dense regression, high resolution heatmap","lastPublishedDoi":"10.21203/rs.3.rs-3926744/v2","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-3926744/v2","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Face detection and landmark localization are usually the fundamental and important steps of facial analysis. The accuracy degradation of any task has impacts on the accuracy and robustness of the downstream analysis. In most facial analysis systems, face detection and landmark detection, as two independent tasks, are predicted sequentially with single-task detectors respectively, which suffer from the higher complexity and poorer consistency. In contrast, multi-task joint face and landmark detectors are more efficient and have more shared features. However, when two tasks are predicted in one model, it is difficult to improve the accuracy of all tasks simultaneously, due to the different requirements of the different targets to be detected. In our opinion, the multi-task detection model, integrating the perception ability of both scale variation and detailed texture information for face and landmark detection, is arduous but important in unconstrained settings. In this paper, a High-Performance joint Face and Landmark detector(HP-FLDet) is proposed, which can predict facial bounding boxes and landmarks in single stage and improve the accuracy of both face detection and landmark localization simultaneously. In this detector, the context inception block and scale-aware multi-task strategies are designed to resolve extreme scale variance and improve the perception ability of detailed texture information. It demonstrates superior performance on both face and landmark detection datasets including AFW, PASCAL Face, FDDB, Wider Face and AFLW. Especially, it achieves SOTA results on the most challenging WIDER FACE dataset with 0.970(Easy), 0.963(Medium), 0.921(Hard) mAP on the test subset, which outperforms all state-of-the-art rivals.","manuscriptTitle":"HP-FLDet: A High Performance Multi-task Joint Face and Landmark Detector","msid":"","msnumber":"","nonDraftVersions":[{"code":2,"date":"2025-03-28 16:05:28","doi":"10.21203/rs.3.rs-3926744/v2","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}},{"code":1,"date":"2024-02-06 04:29:23","doi":"10.21203/rs.3.rs-3926744/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":"c9c23c7a-9050-4841-a1e5-596d648c5b7d","owner":[],"postedDate":"March 28th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2024-02-11T08:29:14+00:00","versionOfRecord":[],"versionCreatedAt":"2025-03-28 16:05:28","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v2","identity":"rs-3926744","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-3926744","identity":"rs-3926744","version":["v2"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}