EgoSync: Self-Supervised Egocentric Video Synchronization through Mutual Temporal Embedding

EgoSync: Self-Supervised Egocentric Video Synchronization through Mutual Temporal Embedding | 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 EgoSync: Self-Supervised Egocentric Video Synchronization through Mutual Temporal Embedding David Whitfield, Mei-Ling Chen, Ayaan Verma, Karen Lindqvist, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8883832/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 Synchronizing egocentric videos captured by multiple wearers performing the same activity is a fundamental requirement for multi-view activity analysis, skill assessment, and collaborative AR guidance. Unlike third-person video synchronization, egocentric settings introduce severe viewpoint variations, rapid head motion, and object occlusions that make traditional alignment methods ineffective. We introduce EgoSync, a self-supervised framework for temporally synchronizing pairs of egocentric videos depicting the same procedural activity. Our approach learns a shared temporal embedding space through three complementary objectives: (1) a Mutual Temporal Embedding loss that aligns corresponding activity progress across viewpoints, (2) a Viewpoint-Invariant Contrastive loss that disentangles activity semantics from viewpoint-specific appearance, and (3) a Temporal Smoothness Regularizer that enforces locally consistent progression in the embedding space. EgoSync operates without manual synchronization annotations, using only the assumption that paired videos depict the same activity. Experiments on Ego4D, EgoExo4D, and CMU-MMAC demonstrate that EgoSync achieves 78.4% synchronization accuracy within a 2-second tolerance on Ego4D, outperforming adapted baselines by 6.1 percentage points. We further show that synchronized features improve downstream egocentric action recognition by 3.2% on Ego4D. Artificial Intelligence and Machine Learning egocentric video temporal synchronization self-supervised learning video alignment contrastive learning multi-view understanding 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-8883832","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":591537514,"identity":"dcfe791a-6425-4036-919c-5c1a8de546f3","order_by":0,"name":"David Whitfield","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABAUlEQVRIiWNgGAWjYDCCAzwMBxhsJBj4GJKPAbnMMmBRHkJaDqRJMLAxpKWBtPAQpYXhAFAxG0OOGXFa+G7kHjz8IcFCjo0959vjihprHn6JBMYHb9twa5G8kZdw4ECChDEbz9vthmeOpfNIzkhgNpyLR4vB7RyDAwd/SCS2SeRuk2xgO8xjcCOBTZqXkBagLfVtEjnPJBv+Heaxv5HA/psYLQlsEjlsko1tQFuAbGZ8WiTvvzE4cCZBwrCN55m5YWNfOo/EmYfNknPO4dbCd+aM8YeKhDp5fvbkZw8bvlnL8bcnH/zwpgy3FixAILGBJPVAwH+AVB2jYBSMglEwzAEANXxVh5hcweAAAAAASUVORK5CYII=","orcid":"","institution":"Bradley University","correspondingAuthor":true,"prefix":"","firstName":"David","middleName":"","lastName":"Whitfield","suffix":""},{"id":591538948,"identity":"092dc8ca-fd95-4679-894a-1b9834b0155a","order_by":1,"name":"Mei-Ling Chen","email":"","orcid":"","institution":"Illinois State University","correspondingAuthor":false,"prefix":"","firstName":"Mei-Ling","middleName":"","lastName":"Chen","suffix":""},{"id":591538949,"identity":"12d40c39-3c5a-4716-bfe6-ef856014ea92","order_by":2,"name":"Ayaan Verma","email":"","orcid":"","institution":"Northern Illinois University","correspondingAuthor":false,"prefix":"","firstName":"Ayaan","middleName":"","lastName":"Verma","suffix":""},{"id":591538950,"identity":"cfad8a0c-2ae2-4768-ab21-4b7b98086a11","order_by":3,"name":"Karen Lindqvist","email":"","orcid":"","institution":"Loyola University Chicago","correspondingAuthor":false,"prefix":"","firstName":"Karen","middleName":"","lastName":"Lindqvist","suffix":""},{"id":591538951,"identity":"e3908eca-de6a-4d96-9b02-20ae04880a68","order_by":4,"name":"Soo-Jin Park","email":"","orcid":"","institution":"Illinois State University","correspondingAuthor":false,"prefix":"","firstName":"Soo-Jin","middleName":"","lastName":"Park","suffix":""}],"badges":[],"createdAt":"2026-02-15 05:30:29","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-8883832/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8883832/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":102887450,"identity":"6286f2b4-2d0f-42d9-a74c-d9cd0992fa02","added_by":"auto","created_at":"2026-02-18 03:14:42","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":211887,"visible":true,"origin":"","legend":"","description":"","filename":"ergomotion.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8883832/v1_covered_4c9cce8f-7f54-4dc5-a243-c39e2070188e.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eEgoSync: Self-Supervised Egocentric Video Synchronization through Mutual Temporal Embedding\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[{"identity":"2cd08c4b-a8af-4639-ad29-f4525e05d46d","identifier":"10.13039/100000001","name":"National Science Foundation","awardNumber":"IIS-2289134","order_by":0}],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Bradley 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":"egocentric video, temporal synchronization, self-supervised learning, video alignment, contrastive learning, multi-view understanding","lastPublishedDoi":"10.21203/rs.3.rs-8883832/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8883832/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eSynchronizing egocentric videos captured by multiple wearers performing the same activity is a fundamental requirement for multi-view activity analysis, skill assessment, and collaborative AR guidance. Unlike third-person video synchronization, egocentric settings introduce severe viewpoint variations, rapid head motion, and object occlusions that make traditional alignment methods ineffective. We introduce EgoSync, a self-supervised framework for temporally synchronizing pairs of egocentric videos depicting the same procedural activity. Our approach learns a shared temporal embedding space through three complementary objectives: (1) a Mutual Temporal Embedding loss that aligns corresponding activity progress across viewpoints, (2) a Viewpoint-Invariant Contrastive loss that disentangles activity semantics from viewpoint-specific appearance, and (3) a Temporal Smoothness Regularizer that enforces locally consistent progression in the embedding space. EgoSync operates without manual synchronization annotations, using only the assumption that paired videos depict the same activity. Experiments on Ego4D, EgoExo4D, and CMU-MMAC demonstrate that EgoSync achieves 78.4% synchronization accuracy within a 2-second tolerance on Ego4D, outperforming adapted baselines by 6.1 percentage points. We further show that synchronized features improve downstream egocentric action recognition by 3.2% on Ego4D.\u003c/p\u003e","manuscriptTitle":"EgoSync: Self-Supervised Egocentric Video Synchronization through Mutual Temporal Embedding","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-18 03:14:36","doi":"10.21203/rs.3.rs-8883832/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":"fd39ff23-09f7-40c0-afe7-871b33e09fff","owner":[],"postedDate":"February 18th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":62948263,"name":"Artificial Intelligence and Machine Learning"}],"tags":[],"updatedAt":"2026-02-18T03:14:36+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-18 03:14:36","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8883832","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8883832","identity":"rs-8883832","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}