[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84051-en":3,"doc-seo-84051-105":29,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},84051,1374391975076,"Riley","https://ap-avatar.wpscdn.com/avatar/14000253ca4ec9f6853?x-image-process=image/resize,m_fixed,w_180,h_180&k=1783305029341752051",8,"Research & Report","Uncertainty Aware Cross Modal Remote Sensing Image Text Retrieval via Evidential Learning","Cross-modal remote sensing image–text retrieval faces unreliable results when test images and texts deviate from curated benchmarks due to sensor and atmosphere degradations and text-side vocabulary heterogeneity. Existing methods assume full certainty and fail to model query-dependent uncertainty. The proposed evidential learning-based approach models inter-modal correspondences as Dirichlet distributions, derives uncertainty per query, and aligns it with retrieval correctness. A deferral-threshold strategy returns confident results and refines uncertain ones using RS-aware test-time augmentation, improving robustness under RS-specific perturbations.","Uncertainty-Aware Cross-Modal Remote Sensing Image–Text Retrieval via Evidential Learning  \nZhuoyue Wang, Xueqian Wang, Member, IEEE, Gang Li, Senior Member, IEEE, Chengxi Li, Member, IEEE, Yongpan Liu, Senior Member, IEEE, and Yifang Ban, Senior Member, IEEE  \nAbstract—In cross-modal remote sensing image–text retrieval (CMRSITR), test-time remote sensing (RS) images and textual descriptions may deviate from well-curated benchmark conditions because of sensor- and atmosphere-related image degradations and text-side RS-vocabulary heterogeneity. Under such non-ideal conditions, existing CMRSITR methods may produce unreliable retrieval results because they usually perform retrieval with full certainty for each query and do not distinguish the varying uncertainty introduced across queries. To address this issue, we propose an evidential learning-based CMRSITR (ELC) method for uncertainty-aware retrieval. During the training phase of ELC, evidential learning (EDL) is employed to model the inter-modal correspondences between RS images and textual descriptions as Dirichlet distributions, from which the uncertainty of each query can be obtained. Based on the EDL outputs, uncertainty–correctness alignment learning (UCL) is introduced to align the estimated uncertainty with retrieval correctness, encouraging high uncertainty for incorrect retrieval and low uncertainty for correct retrieval. Furthermore, intramodal relationship learning (RL) distills the intra-modal similarity structure from pretrained mentor encoders for the trainable encoders, thereby making the Dirichlet distributions modeled by EDL more discriminative. In the test phase of ELC, the estimated uncertainty is compared with a threshold determined by a fixed deferral ratio, where low-uncertainty queries are directly returned and high-uncertainty queries are refined by RS-aware test-time augmentation (RS-TTA). Experimental results demonstrate that our proposed ELC achieves competitive retrieval performance in comparison with existing state-of-the-art CMRSITR methods and provides stronger robustness under the evaluated RS-specific degradations, including sensor- and atmosphere-related image perturbations and RS-vocabulary heterogeneity.  \nIndex Terms—Cross-modal remote sensing image–text retrieval, evidential learning, remote sensing data management  \nThis work was supported by National Key R&D Program of China under Grant 2025YFF0514602, National Natural Science Foundation of China under Grants 62471274, 62341130, and Beijing Natural Science Foundation under Grant JQ2501 . Part of this paper has been published by International Geoscience and Remote Sensing Symposium 2025 [1] . Corresponding Author: Xueqian Wang.  \nZ. Wang and Y. Liu are with the Department of Electronic Engineering, Tsinghua University, Beijing 100084, China.  \nX. Wang and G. Li are with the Department of Electronic Engineering, Tsinghua University, Beijing 100084, China, and also with the State Key Laboratory of Space Network and Communications, Tsinghua University, Beijing 100084, China ([e-mail: wangxueqian@mail.tsinghua.edu.cn](e-mail: wangxueqian@mail.tsinghua.edu.cn)).  \nC. Li is with the School of Electrical Engineering and Computer Science, KTH Royal Institute of Technology.  \nY. Ban is with the Division of Geoinformatics, School of Architecture and the Built Environment, KTH Royal Institute of Technology.  \nI. INTRODUCTION  \nWith the rapid development of technology [2]-[4], an enormous  \nEarth observation volume of remote  \nsensing (RS) data, including images and textual descriptions, is continuously generated by numerous satellite sensors. RS images provide rich spatial and spectral details but often lack explicit semantic meaning. In contrast, textual descriptions provide semantic context and high-level interpretations but may lack detailed spatial precision. Effectively leveraging both visual and textual modalities is crucial for improving the accessibility and interpretability of RS data, whic","cbCailgZ7fYJWPva","https://ap.wps.com/l/cbCailgZ7fYJWPva","pdf",1560865,1,16,"English","en",105,"# Introduction\n## Cross-modal remote sensing image–text retrieval task\n## Deep learning approaches and categories\n## Challenges under non-ideal test conditions","[{\"question\":\"Why do existing CMRSITR methods become unreliable in real-world scenarios?\",\"answer\":\"Test images and text descriptions often differ from curated benchmarks due to sensor- and atmosphere-related degradations and text-side RS vocabulary heterogeneity. Most methods treat every query with full certainty and do not model varying uncertainty, leading to unreliable retrieval.\"},{\"question\":\"How does evidential learning estimate uncertainty for each query in the proposed method?\",\"answer\":\"During training, evidential learning models inter-modal correspondences between remote sensing images and textual descriptions as Dirichlet distributions. The outputs of this modeling provide an uncertainty estimate for each query.\"},{\"question\":\"How are low-uncertainty and high-uncertainty queries handled at test time?\",\"answer\":\"A threshold determined by a fixed deferral ratio is used to compare with estimated uncertainty. Low-uncertainty queries are returned directly, while high-uncertainty queries are refined using RS-aware test-time augmentation (RS-TTA).\"}]",1784192249,40,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"uncertainty-aware-cross-modal-remote-sensing-image-text-retrieval-via-evidential-learning","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/uncertainty-aware-cross-modal-remote-sensing-image-text-retrieval-via-evidential-learning/84051/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why do existing CMRSITR methods become unreliable in real-world scenarios?","Question",{"text":75,"@type":76},"Test images and text descriptions often differ from curated benchmarks due to sensor- and atmosphere-related degradations and text-side RS vocabulary heterogeneity. Most methods treat every query with full certainty and do not model varying uncertainty, leading to unreliable retrieval.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does evidential learning estimate uncertainty for each query in the proposed method?",{"text":80,"@type":76},"During training, evidential learning models inter-modal correspondences between remote sensing images and textual descriptions as Dirichlet distributions. The outputs of this modeling provide an uncertainty estimate for each query.",{"name":82,"@type":73,"acceptedAnswer":83},"How are low-uncertainty and high-uncertainty queries handled at test time?",{"text":84,"@type":76},"A threshold determined by a fixed deferral ratio is used to compare with estimated uncertainty. Low-uncertainty queries are returned directly, while high-uncertainty queries are refined using RS-aware test-time augmentation (RS-TTA).","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,119,122,127,130,134],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":28,"slug":118},7,"Healthcare","healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":45,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":45,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":45,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":45,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]