LDCAP: Robust Image Captioning via Latent Compression and Dynamic Decoder Conditioning

LDCAP: Robust Image Captioning via Latent Compression and Dynamic Decoder Conditioning | 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 LDCAP: Robust Image Captioning via Latent Compression and Dynamic Decoder Conditioning Veerababu Reddy, Seetharam Poola, Musharaf Shaik, Sai Durga Kistaparapu, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9480848/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 Generating accurate natural language descriptions from images remains a challenging task, particularly when input images are captured under poor lighting conditions such as dim indoor environments, nighttime scenes, or strongly backlit settings. Under such conditions, region-level visual features extracted by standard object detection networks become noisy and unreliable, causing attention mechanisms to focus on uninformative image regions and ultimately degrading caption quality. This paper presents LDCAP (Latent Compression and Dynamic Decoder Conditioning for Image Captioning), a transformer-based captioning model that addresses visual feature degradation directly within the network architecture, without requiring any external image preprocessing or enhancement step. LDCAP incorporates three targeted architectural modifications over its SCAP baseline. First, the encoder is redesigned using a Recurrent Interface Network (RIN), which compresses variable-length region features into a fixed set of 64 learnable latent tokens through iterative cross-attention, forming a structured information bottleneck that naturally suppresses noise-dominated feature dimensions. Second, Feature-wise Linear Modulation (FiLM) layers are integrated into each decoder block, enabling global scene context to dynamically condition hidden representations at every caption generation step, complementing the local cross-attention mechanism. Third, a two-stage training strategy is employed, combining cross-entropy pre-training with self-critical sequence training (SCST), which directly aligns the optimisation objective with standard captioning evaluation metrics. The complete model contains 31.8M parameters, remaining compact relative to large-scale vision-language pre-trained models while achieving competitive performance. Experimental evaluation on the MS-COCO 2014 benchmark demonstrates that LDCAP achieves a CIDEr score of 134.2 (±0.4), improving upon the SCAP baseline of 131.7, with consistent gains across METEOR (32.9), ROUGE-L (62.8), BLEU-1 (85.2), and BLEU-4 (39.8). Zero-shot evaluation on Flickr30k confirms that the improvements generalise across datasets, with LDCAP reaching a CIDEr score of 135.4 compared to 132.8 for SCAP. The advantage of LDCAP is most pronounced under degraded illumination, where it outperforms SCAP by 4.4 CIDEr points under low-light conditions versus 2.2 points under normal lighting. Controlled experiments with synthetic gamma degradation at four severity levels confirm that the performance gap widens monotonically as illumination deteriorates, and that internal architectural robustness consistently outperforms CLAHE-based external preprocessing at every degradation level. Ablation experiments validate that all three proposed components contribute independently and positively to overall performance, and attention visualisations demonstrate that LDCAP produces more focused and semantically meaningful attention patterns under challenging visual conditions. The source code is publicly available at : doi.org/10.5281/zenodo.19626529}{doi.org/10.5281/zenodo.19626529. Image captioning Recurrent Interface Networks Feature wise Linear Modulation Visual feature robustness Self critical sequence training Attention mechanism Full Text Additional Declarations No competing interests reported. Supplementary Files InputOutput.docx GraphicalAbstract.png 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-9480848","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":626847465,"identity":"7a99a077-fbf3-429f-97b2-c4b24711a22a","order_by":0,"name":"Veerababu Reddy","email":"data:image/png;base64,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","orcid":"","institution":"Vignan’s Lara Institute of Technology and Science","correspondingAuthor":true,"prefix":"","firstName":"Veerababu","middleName":"","lastName":"Reddy","suffix":""},{"id":626847466,"identity":"ffe6a272-3fa7-47c5-83ae-c19d5623305e","order_by":1,"name":"Seetharam Poola","email":"","orcid":"","institution":"Vignan’s Lara Institute of Technology and Science","correspondingAuthor":false,"prefix":"","firstName":"Seetharam","middleName":"","lastName":"Poola","suffix":""},{"id":626847467,"identity":"2f281b94-7792-41ff-8d81-bb837e4f5b0f","order_by":2,"name":"Musharaf Shaik","email":"","orcid":"","institution":"Vignan’s Lara Institute of Technology and Science","correspondingAuthor":false,"prefix":"","firstName":"Musharaf","middleName":"","lastName":"Shaik","suffix":""},{"id":626847468,"identity":"4ae8c449-fb79-43ae-9213-7faefc49aecd","order_by":3,"name":"Sai Durga Kistaparapu","email":"","orcid":"","institution":"Vignan’s Lara Institute of Technology and Science","correspondingAuthor":false,"prefix":"","firstName":"Sai","middleName":"Durga","lastName":"Kistaparapu","suffix":""},{"id":626847469,"identity":"a218586b-51b7-492c-ad41-6fc2ec762a26","order_by":4,"name":"Vignesh Sajja","email":"","orcid":"","institution":"Vignan’s Lara Institute of Technology and Science","correspondingAuthor":false,"prefix":"","firstName":"Vignesh","middleName":"","lastName":"Sajja","suffix":""}],"badges":[],"createdAt":"2026-04-21 08:08:44","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9480848/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9480848/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108493402,"identity":"8605aac3-903c-49f8-bcab-416303b70b3b","added_by":"auto","created_at":"2026-05-05 10:00:16","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":6211455,"visible":true,"origin":"","legend":"","description":"","filename":"LDCAP.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9480848/v1_covered_1ad2813b-d292-4a42-ae53-5e12e8c02476.pdf"},{"id":107489784,"identity":"bd147c43-8fef-4c5b-a57f-904a70fef0b4","added_by":"auto","created_at":"2026-04-22 02:48:57","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":1477963,"visible":true,"origin":"","legend":"","description":"","filename":"InputOutput.docx","url":"https://assets-eu.researchsquare.com/files/rs-9480848/v1/6dde726fec451c07bfc611e7.docx"},{"id":107489582,"identity":"540ec5be-911a-4fd7-85cc-08d535242af8","added_by":"auto","created_at":"2026-04-22 02:48:16","extension":"png","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":753706,"visible":true,"origin":"","legend":"","description":"","filename":"GraphicalAbstract.png","url":"https://assets-eu.researchsquare.com/files/rs-9480848/v1/c8f016fe9d2b1cc4f3e3b268.png"}],"financialInterests":"No competing interests reported.","formattedTitle":"LDCAP: Robust Image Captioning via Latent Compression and Dynamic Decoder Conditioning","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"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 captioning, Recurrent Interface Networks, Feature wise Linear Modulation, Visual feature robustness, Self critical sequence training, Attention mechanism","lastPublishedDoi":"10.21203/rs.3.rs-9480848/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9480848/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eGenerating accurate natural language descriptions from images remains a challenging task, particularly when input images are captured under poor lighting conditions such as dim indoor environments, nighttime scenes, or strongly backlit settings. Under such conditions, region-level visual features extracted by standard object detection networks become noisy and unreliable, causing attention mechanisms to focus on uninformative image regions and ultimately degrading caption quality. This paper presents LDCAP (Latent Compression and Dynamic Decoder Conditioning for Image Captioning), a transformer-based captioning model that addresses visual feature degradation directly within the network architecture, without requiring any external image preprocessing or enhancement step. LDCAP incorporates three targeted architectural modifications over its SCAP baseline. First, the encoder is redesigned using a Recurrent Interface Network (RIN), which compresses variable-length region features into a fixed set of 64 learnable latent tokens through iterative cross-attention, forming a structured information bottleneck that naturally suppresses noise-dominated feature dimensions. Second, Feature-wise Linear Modulation (FiLM) layers are integrated into each decoder block, enabling global scene context to dynamically condition hidden representations at every caption generation step, complementing the local cross-attention mechanism. Third, a two-stage training strategy is employed, combining cross-entropy pre-training with self-critical sequence training (SCST), which directly aligns the optimisation objective with standard captioning evaluation metrics. The complete model contains 31.8M parameters, remaining compact relative to large-scale vision-language pre-trained models while achieving competitive performance. Experimental evaluation on the MS-COCO 2014 benchmark demonstrates that LDCAP achieves a CIDEr score of 134.2 (±0.4), improving upon the SCAP baseline of 131.7, with consistent gains across METEOR (32.9), ROUGE-L (62.8), BLEU-1 (85.2), and BLEU-4 (39.8). Zero-shot evaluation on Flickr30k confirms that the improvements generalise across datasets, with LDCAP reaching a CIDEr score of 135.4 compared to 132.8 for SCAP. The advantage of LDCAP is most pronounced under degraded illumination, where it outperforms SCAP by 4.4 CIDEr points under low-light conditions versus 2.2 points under normal lighting. Controlled experiments with synthetic gamma degradation at four severity levels confirm that the performance gap widens monotonically as illumination deteriorates, and that internal architectural robustness consistently outperforms CLAHE-based external preprocessing at every degradation level. Ablation experiments validate that all three proposed components contribute independently and positively to overall performance, and attention visualisations demonstrate that LDCAP produces more focused and semantically meaningful attention patterns under challenging visual conditions. The source code is publicly available at : doi.org/10.5281/zenodo.19626529}{doi.org/10.5281/zenodo.19626529.\u003c/p\u003e","manuscriptTitle":"LDCAP: Robust Image Captioning via Latent Compression and Dynamic Decoder Conditioning","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-22 02:42:36","doi":"10.21203/rs.3.rs-9480848/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":"b3988f64-c7aa-4c86-bd18-1685f20b5173","owner":[],"postedDate":"April 22nd, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-05-03T01:23:34+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-22 02:42:36","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9480848","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9480848","identity":"rs-9480848","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}