Image Classification on Small Datasets using Query Attention Module

Image Classification on Small Datasets using Query Attention Module | 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 Image Classification on Small Datasets using Query Attention Module Chunyu Jiang, Renwei Li, Hanxiang Zhang This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4447366/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 Vision Transformers (VTs) are increasingly popular in computer vision due to their robust global modeling. However, they do not have the same learning advantages as Convolutional Neural Networks (CNNs), which can be trained effectively with less data. This paper proposes a plug-and-play module, query attention. Without altering the backbone structure, it can be integrated with existing VTs and CNNs. Furthermore, to reduce the training cost of the VT backbone, we integrate query attention with a downsized backbone to construct a shrunk structure. We selected three classical VTs, ViT, Swin, and T2T-ViT, and tested them on four small datasets(CIFAR10, CIFAR100, Tiny-ImageNet, CINIC10). The results show that query attention can improve model performance, especially with the shrunk structure, where we can maintain competitive performance while reducing computational and memory complexity. For example, on the Tiny-ImageNet, after adding query attention to ViT, the classification accuracy increases by 5.77%. Additionally, based on the shrunk structure of the ViT, we achieved a 20% reduction in parameters and a 32.76% reduction in computation cost, while improving classification accuracy by 4.26%. Image classification Small datasets Attention mechanism Vision Transformers 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-4447366","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":311461121,"identity":"36e7e6cf-a7a7-43a9-899b-5d9c19f44119","order_by":0,"name":"Chunyu Jiang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA/ElEQVRIiWNgGAWjYBACxmYwdQDMfiBRIAHmShCrhdlAwkCCgYeQFigAa2GTYDBgIKyFuZ352cMvNXcSt0vkmFVYGFjI2zMwH7zNw2CXh9thbObGMseeJe6ckWN2A+gwwx4GtmRrHobkYjx+MZOWbDicuOEGREsCDwOPmTQPw4HEBpxa2L/BtRRAtPB/I6CFx0zyI1QLA9QWNkJayqQZjh023nDmWbEE2C+H2Ywt5xgk49Ri2H98m+SPmsOyG44nb/wsUVEnz97e/PDGmwo73FqAEsygqGAQSGBgBscHM4gwwKEeCORBjvsBYvEfYGD8gFvhKBgFo2AUjGAAAA7RUDOAF0P4AAAAAElFTkSuQmCC","orcid":"","institution":"Jilin University of Chemical Technology","correspondingAuthor":true,"prefix":"","firstName":"Chunyu","middleName":"","lastName":"Jiang","suffix":""},{"id":311461124,"identity":"17ddf14b-8baf-4ee5-8deb-1c26abdcb8f0","order_by":1,"name":"Renwei Li","email":"","orcid":"","institution":"Jilin University of Chemical Technology","correspondingAuthor":false,"prefix":"","firstName":"Renwei","middleName":"","lastName":"Li","suffix":""},{"id":311461123,"identity":"2dac1669-4a91-4f0c-986a-1a9ace816e32","order_by":2,"name":"Hanxiang Zhang","email":"","orcid":"","institution":"Zhejiang Business College","correspondingAuthor":false,"prefix":"","firstName":"Hanxiang","middleName":"","lastName":"Zhang","suffix":""}],"badges":[],"createdAt":"2024-05-20 07:33:50","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-4447366/v2","doiUrl":"https://doi.org/10.21203/rs.3.rs-4447366/v2","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":64190315,"identity":"2a0d4189-1530-4fc2-ac99-abbd4a08cb35","added_by":"auto","created_at":"2024-09-09 18:10:43","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2114104,"visible":true,"origin":"","legend":"","description":"","filename":"ImageClassificationonSmallDatasetswithoutPreTrainingusingQueryAttention.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4447366/v2_covered_dc960199-6c77-4641-8afb-3464ccc83053.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"Image Classification on Small Datasets using Query Attention Module","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":"Image classification, Small datasets, Attention mechanism, Vision Transformers","lastPublishedDoi":"10.21203/rs.3.rs-4447366/v2","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4447366/v2","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eVision Transformers (VTs) are increasingly popular in computer vision due to their robust global modeling. However, they do not have the same learning advantages as Convolutional Neural Networks (CNNs), which can be trained effectively with less data. This paper proposes a plug-and-play module, query attention. Without altering the backbone structure, it can be integrated with existing VTs and CNNs. Furthermore, to reduce the training cost of the VT backbone, we integrate query attention with a downsized backbone to construct a shrunk structure. We selected three classical VTs, ViT, Swin, and T2T-ViT, and tested them on four small datasets(CIFAR10, CIFAR100, Tiny-ImageNet, CINIC10). The results show that query attention can improve model performance, especially with the shrunk structure, where we can maintain competitive performance while reducing computational and memory complexity. For example, on the Tiny-ImageNet, after adding query attention to ViT, the classification accuracy increases by 5.77%. Additionally, based on the shrunk structure of the ViT, we achieved a 20% reduction in parameters and a 32.76% reduction in computation cost, while improving classification accuracy by 4.26%.\u003c/p\u003e","manuscriptTitle":"Image Classification on Small Datasets using Query Attention Module","msid":"","msnumber":"","nonDraftVersions":[{"code":2,"date":"2024-09-09 18:02:38","doi":"10.21203/rs.3.rs-4447366/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":"2024-06-06 16:52:22","doi":"10.21203/rs.3.rs-4447366/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":"61f974fc-f226-4551-a147-8d2f1fe1b95a","owner":[],"postedDate":"September 9th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2024-06-23T04:53:24+00:00","versionOfRecord":[],"versionCreatedAt":"2024-09-09 18:02:38","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v2","identity":"rs-4447366","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4447366","identity":"rs-4447366","version":["v2"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}