ThermalCLIP: Hardware-Agnostic Industrial Anomaly Detection via Physics-Informed Thermal Proxies from RGB

ThermalCLIP: Hardware-Agnostic Industrial Anomaly Detection via Physics-Informed Thermal Proxies from RGB | 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 ThermalCLIP: Hardware-Agnostic Industrial Anomaly Detection via Physics-Informed Thermal Proxies from RGB Hridika Chanda, Nusrat Sultana This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9385007/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 Thermal infrared imaging is physically well-suited to industrial anomaly detection. However, calibrated thermal cameras cost thousands of dollars per device, making it inaccessible for most industrial environments. In this paper, we directly address this barrier. We prove that Kirchhoff's law of thermal radiation, establishes thermal emissivity and optical reflectance are complementary quantities at thermal equilibrium. Since luminance approximates reflectance for Lambertian surfaces, the synthetic thermal proxy ε = 1 – L follows directly from standard RGB input. This proxy is validated against physics-based IR references using r = 0.785 average Pearson correlation across a range of industrial surfaces. We propose ThermalCLIP, which combines this proxy with frozen CLIP ViT-B/16 patch features and PDE-based multiscale residuals. The core finding is that on surface categories where thermal diffusion structure is the primary discriminative signal, highly repetitive textures where all appearance-based patch features are near-identical, the physics-informed Laplacian residual provides information that appearance-based methods are structurally unable to recover. On the grid category of MVTec AD, ThermalCLIP achieves an image-level AUROC of 0.9942 compared to 0.6817 for PatchCore under identical evaluation conditions, a gap that no appearance-based method is positioned to close. Across all 15 MVTec AD categories, ThermalCLIP achieves a mean image level AUROC of 0.9441, and 0.8724 in zero-shot transfer to the VisA dataset. The complete framework runs at 7.5 FPS on a consumer-grade GPU with no infrared sensor required. Anomaly detection Thermal proxy Physics-informed CLIP features Laplacian residuals RGB inspection 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-9385007","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":621295807,"identity":"2a0361c0-634e-46ac-847b-c40aa1134b78","order_by":0,"name":"Hridika Chanda","email":"","orcid":"","institution":"Chittagong University of Engineering and Technology","correspondingAuthor":false,"prefix":"","firstName":"Hridika","middleName":"","lastName":"Chanda","suffix":""},{"id":621295808,"identity":"08621783-073f-4872-84fb-b2e1c85f6bbd","order_by":1,"name":"Nusrat Sultana","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA+ElEQVRIiWNgGAWjYDADPjBZAecfYGzAqz6BgYGNDcQ4A8RsJGlhbCNCi7lE+rMPP3/UJbbJNz978HOeXWK/fAPjhx8Md2RxabGckWM8syfhcGIbG5u5Ye+25MSZbQzMkj0Mz4xxaTG4kcPMwJNwAKiFwUyCd9uBxA3HGBikGRgOJ+LWkv6Y8U8C0GFs7N8k/845kLj/GAPzb/xaEoyZeRKYgVp4zKR5G4C2sDGw4bflzBtjZpm0w8ZtbDll0jLHko1nHEtss+wxwOOX40CHvbGpk+1nPr5N8k2NnWx/8+HDN35U4A4xbAAUIwYkqB8Fo2AUjIJRgAEAGGlX90djPUkAAAAASUVORK5CYII=","orcid":"","institution":"Chittagong University of Engineering and Technology","correspondingAuthor":true,"prefix":"","firstName":"Nusrat","middleName":"","lastName":"Sultana","suffix":""}],"badges":[],"createdAt":"2026-04-11 06:14:04","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-9385007/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9385007/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":106852743,"identity":"35ed1555-f1a8-40f7-a433-25d5a137b6aa","added_by":"auto","created_at":"2026-04-14 06:50:45","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":946804,"visible":true,"origin":"","legend":"","description":"","filename":"ThermalCLIP.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9385007/v1_covered_22c00db6-7253-4bfb-afc2-d6318db054f0.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003e\u003cstrong\u003eThermalCLIP: Hardware-Agnostic Industrial Anomaly Detection via Physics-Informed Thermal Proxies from RGB\u003c/strong\u003e\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Chittagong University of Engineering \u0026 Technology","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":"Anomaly detection, Thermal proxy, Physics-informed, CLIP features, Laplacian residuals, RGB inspection","lastPublishedDoi":"10.21203/rs.3.rs-9385007/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9385007/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThermal infrared imaging is physically well-suited to industrial anomaly detection. However, calibrated thermal cameras cost thousands of dollars per device, making it inaccessible for most industrial environments. In this paper, we directly address this barrier. We prove that Kirchhoff's law of thermal radiation, establishes thermal emissivity and optical reflectance are complementary quantities at thermal equilibrium. Since luminance approximates reflectance for Lambertian surfaces, the synthetic thermal proxy ε = 1 – L follows directly from standard RGB input. This proxy is validated against physics-based IR references using r = 0.785 average Pearson correlation across a range of industrial surfaces. We propose ThermalCLIP, which combines this proxy with frozen CLIP ViT-B/16 patch features and PDE-based multiscale residuals. The core finding is that on surface categories where thermal diffusion structure is the primary discriminative signal, highly repetitive textures where all appearance-based patch features are near-identical, the physics-informed Laplacian residual provides information that appearance-based methods are structurally unable to recover. On the grid category of MVTec AD, ThermalCLIP achieves an image-level AUROC of 0.9942 compared to 0.6817 for PatchCore under identical evaluation conditions, a gap that no appearance-based method is positioned to close. Across all 15 MVTec AD categories, ThermalCLIP achieves a mean image level AUROC of 0.9441, and 0.8724 in zero-shot transfer to the VisA dataset. The complete framework runs at 7.5 FPS on a consumer-grade GPU with no infrared sensor required.\u003c/p\u003e","manuscriptTitle":"ThermalCLIP: Hardware-Agnostic Industrial Anomaly Detection via Physics-Informed Thermal Proxies from RGB","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-14 06:50:27","doi":"10.21203/rs.3.rs-9385007/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":"8d3c27ed-4391-4811-a6ca-b5df93891e17","owner":[],"postedDate":"April 14th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-04-21T14:39:43+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-14 06:50:27","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9385007","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9385007","identity":"rs-9385007","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}