In-domain versus Out-of-domain transfer learning for document layout analysis

In-domain versus Out-of-domain transfer learning for document layout analysis | 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 In-domain versus Out-of-domain transfer learning for document layout analysis Axel De Nardin, Silvia Zottin, Claudio Piciarelli, Gian Luca Foresti, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4414436/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 19 Aug, 2024 Read the published version in International Journal on Document Analysis and Recognition (IJDAR) → Version 1 posted 10 You are reading this latest preprint version Abstract Data availability is a big concern in the field of document analysis, especially when working on tasks that require a high degree of precision when it comes to the definition of the ground truths on which to train deep learning models. A notable example is represented by the task of Document Layout Analysis in handwritten documents, which requires pixel-precise segmentation maps to highlight the different layout components of each document page. These segmentation maps are typically very time-consuming and require a high degree of domain knowledge to be defined, as they are intrinsically characterized by the content of the text. For this reason in the present work, we explore the effects of different initialization strategies for deep learning models employed for this type of task by relying on both in-domain and cross-domain datasets for their pre-training. To test the employed models we use 2 publicly available datasets with heterogeneous characteristics both regarding their structure as well as the languages of the contained documents. We show how a combination of cross-domain and in-domain transfer learning approaches leads to the best overall performance of the models, as well as speeding up their convergence process. Document analysis Layout segmentation Semantic segmentation Transfer learning Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 19 Aug, 2024 Read the published version in International Journal on Document Analysis and Recognition (IJDAR) → Version 1 posted Editorial decision: Revision requested 10 Jul, 2024 Reviews received at journal 07 Jul, 2024 Reviews received at journal 05 Jul, 2024 Reviewers agreed at journal 23 Jun, 2024 Reviewers agreed at journal 21 Jun, 2024 Reviewers agreed at journal 21 Jun, 2024 Reviewers invited by journal 21 Jun, 2024 Editor assigned by journal 14 May, 2024 Submission checks completed at journal 14 May, 2024 First submitted to journal 13 May, 2024 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-4414436","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":305699176,"identity":"8113710c-8767-409e-85c5-c59cbcaa6a17","order_by":0,"name":"Axel De Nardin","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA5ElEQVRIie3PrQrCUBTA8TOEpenqTfoK16z4LLtcmEVkIojBYJrFB1AUfQWHsHzgwFYuy0ZXzILF5sdQEcTNaLj/dMqPcw6ATvefGQhmNnCEIdxnDk6BeSPqSQrMi4DhP4a8NfZMIsKg2bUXFFBvRd0KkxtIz98JU5GDkLh9lrgezUPqm8z18g+LpxwNn8RYWZzKIQmfdfJ/qZF9vJGLWGdk+QPh8RRuBMUmI+MfSF1FHJ1EikCZHllRW/jWwUPH/U6qO5nuj4OWWKnS9mSNGmI9kUF6bua8f+/jCiwAOp1OpyvoCi+IWV381DcZAAAAAElFTkSuQmCC","orcid":"","institution":"University of Udine","correspondingAuthor":true,"prefix":"","firstName":"Axel","middleName":"","lastName":"De Nardin","suffix":""},{"id":305699177,"identity":"de0e1da6-0ec9-4ba1-be86-f50f9d0bb743","order_by":1,"name":"Silvia Zottin","email":"","orcid":"","institution":"University of Udine","correspondingAuthor":false,"prefix":"","firstName":"Silvia","middleName":"","lastName":"Zottin","suffix":""},{"id":305699178,"identity":"0bed29af-7233-4ef7-8612-4c37e65394c4","order_by":2,"name":"Claudio Piciarelli","email":"","orcid":"","institution":"University of Udine","correspondingAuthor":false,"prefix":"","firstName":"Claudio","middleName":"","lastName":"Piciarelli","suffix":""},{"id":305699179,"identity":"f19086a5-5b5c-4dd0-9501-f984e0a12b07","order_by":3,"name":"Gian Luca Foresti","email":"","orcid":"","institution":"University of Udine","correspondingAuthor":false,"prefix":"","firstName":"Gian","middleName":"Luca","lastName":"Foresti","suffix":""},{"id":305699180,"identity":"572f2e67-d172-46be-aad8-e0feeeef4e0d","order_by":4,"name":"Emanuela Colombi","email":"","orcid":"","institution":"University of Udine","correspondingAuthor":false,"prefix":"","firstName":"Emanuela","middleName":"","lastName":"Colombi","suffix":""}],"badges":[],"createdAt":"2024-05-13 16:17:09","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4414436/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4414436/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s10032-024-00497-4","type":"published","date":"2024-08-19T15:57:46+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":63300329,"identity":"8db87366-83f8-4bb5-b737-4d67ae09cb33","added_by":"auto","created_at":"2024-08-26 16:13:36","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4378429,"visible":true,"origin":"","legend":"","description":"","filename":"IJDARTransferLearningmanuscripts1.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4414436/v1_covered_5b8e8420-1d0b-4252-9b53-e3d5f3708bf6.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"In-domain versus Out-of-domain transfer learning for document layout analysis","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":"international-journal-on-document-analysis-and-recognition-ijdar","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ijda","sideBox":"Learn more about [International Journal on Document Analysis and Recognition (IJDAR)](http://link.springer.com/journal/10032)","snPcode":"10032","submissionUrl":"https://submission.nature.com/new-submission/10032/3","title":"International Journal on Document Analysis and Recognition (IJDAR)","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Document analysis, Layout segmentation, Semantic segmentation, Transfer learning","lastPublishedDoi":"10.21203/rs.3.rs-4414436/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4414436/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Data availability is a big concern in the field of document analysis, especially when working on tasks that require a high degree of precision when it comes to the definition of the ground truths on which to train deep learning models. A notable example is represented by the task of Document Layout Analysis in handwritten documents, which requires pixel-precise segmentation maps to highlight the different layout components of each document page. These segmentation maps are typically very time-consuming and require a high degree of domain knowledge to be defined, as they are intrinsically characterized by the content of the text. For this reason in the present work, we explore the effects of different initialization strategies for deep learning models employed for this type of task by relying on both in-domain and cross-domain datasets for their pre-training. To test the employed models we use 2 publicly available datasets with heterogeneous characteristics both regarding their structure as well as the languages of the contained documents. We show how a combination of cross-domain and in-domain transfer learning approaches leads to the best overall performance of the models, as well as speeding up their convergence process.","manuscriptTitle":"In-domain versus Out-of-domain transfer learning for document layout analysis","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-05-23 04:38:10","doi":"10.21203/rs.3.rs-4414436/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2024-07-10T13:33:15+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-07-07T14:55:17+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-07-05T19:19:53+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"211853408403083500486282031445904154716","date":"2024-06-23T13:39:54+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"113540601467491908690420701605780383080","date":"2024-06-22T01:02:42+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"100530615003168971253217385411617660658","date":"2024-06-21T18:23:16+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2024-06-21T11:52:44+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-05-15T01:02:15+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-05-14T12:19:35+00:00","index":"","fulltext":""},{"type":"submitted","content":"International Journal on Document Analysis and Recognition (IJDAR)","date":"2024-05-13T16:15:56+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"international-journal-on-document-analysis-and-recognition-ijdar","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ijda","sideBox":"Learn more about [International Journal on Document Analysis and Recognition (IJDAR)](http://link.springer.com/journal/10032)","snPcode":"10032","submissionUrl":"https://submission.nature.com/new-submission/10032/3","title":"International Journal on Document Analysis and Recognition (IJDAR)","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"399bacd1-c250-4400-bf3a-a64a47da5352","owner":[],"postedDate":"May 23rd, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2024-08-26T16:03:45+00:00","versionOfRecord":{"articleIdentity":"rs-4414436","link":"https://doi.org/10.1007/s10032-024-00497-4","journal":{"identity":"international-journal-on-document-analysis-and-recognition-ijdar","isVorOnly":false,"title":"International Journal on Document Analysis and Recognition (IJDAR)"},"publishedOn":"2024-08-19 15:57:46","publishedOnDateReadable":"August 19th, 2024"},"versionCreatedAt":"2024-05-23 04:38:10","video":"","vorDoi":"10.1007/s10032-024-00497-4","vorDoiUrl":"https://doi.org/10.1007/s10032-024-00497-4","workflowStages":[]},"version":"v1","identity":"rs-4414436","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4414436","identity":"rs-4414436","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}