[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-663725-105":59,"doc-detail-663725-en":130},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":123,"head_meta":125,"extra_data":127,"updated_unix":129},105,"en","learning-efficient-representations-for-image-based-patent-retrieval","Learning Efficient Representations for Image-Based Patent Retrieval","","Patent retrieval has long attracted research attention in intellectual property and information retrieval, yet most methods depend on text and metadata, leaving content-based image-based patent retrieval comparatively underexplored. Leveraging visual traits of patent drawing images, this work proposes a simple, lightweight model tailored to patent diagrams. Experiments on a large-scale benchmark show substantial improvements, raising mean average precision by 33.5% over existing approaches. The method scales to reach 93.5% mAP and attains first place in the ECCV 2022 Patent Diagram Image Retrieval Challenge.",{"@graph":69,"@context":122},[70,84,105],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":40,"@type":76,"position":81},"https://docshare.wps.com/document/research-report/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/learning-efficient-representations-for-image-based-patent-retrieval/663725/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/learning-efficient-representations-for-image-based-patent-retrieval/663725.png","ImageObject",300,407,{"name":92,"@type":93},"Adam","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-30","2026-10-28",true,{"@type":102,"interactionType":103,"userInteractionCount":81},"InteractionCounter",{"@type":104},"ViewAction",{"@type":106,"mainEntity":107},"FAQPage",[108,114,118],{"name":109,"@type":110,"acceptedAnswer":111},"Why is content-based image-based patent retrieval challenging?","Question",{"text":112,"@type":113},"Existing systems mostly rely on textual and metadata signals, while retrieval from patent drawing visuals remains an open challenge. Diagram images have structural characteristics different from natural images, and prior visual representations can be insufficiently discriminative at scale.","Answer",{"name":115,"@type":110,"acceptedAnswer":116},"What does the proposed method target in patent diagram images?",{"text":117,"@type":113},"The approach is designed by exploiting traits of patent drawing images. It uses a lightweight backbone plus a neck structure to produce low-dimensional representations for efficient retrieval, trained with a classification loss in a geometric angular space.",{"name":119,"@type":110,"acceptedAnswer":120},"What performance results and recognition does the paper report?",{"text":121,"@type":113},"On a large-scale benchmark, the method improves mean average precision by 33.5% over other approaches, and further scaling reaches 93.5% mAP. The method ranks first in the ECCV 2022 Patent Diagram Image Retrieval Challenge.","https://schema.org",{"og:url":83,"og:type":124,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":126,"canonical":83},"index,follow",{"doc_id":128,"site_id":62},663725,1793230679,{"code":4,"msg":5,"data":131},{"doc_id":128,"user_id":132,"nickname":92,"user_avatar":133,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":134,"file_id":135,"file_url":136,"file_type":137,"file_size":138,"view_count":81,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":139,"language":140,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":141,"faqs":142,"seo_title":143,"seo_description":67,"update_tm":129,"read_time":41},1374404737137,"https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d","arXiv :2308 . 13749v1 [ cs .CV] 26 Aug 2023  \nLearning Efficient Representations for Image-Based Patent Retrieval  \nHongsong Wang 1*(B) and Yuqi Zhang2*  \n1 Department of Computer Science and Engineering,  \nSoutheast University, Nanjing 210096, China  \n2 Baidu Inc  \n[hongsongwang@seu.edu.cn](hongsongwang@seu.edu.cn) , [zhangyuqi1991@gmail.com](zhangyuqi1991@gmail.com)  \nAbstract. Patent retrieval has been attracting tremendous interest from researchers in intellectual property and information retrieval communities in the past decades. However, most existing approaches rely on textual and metadata information of the patent, and content-based imagebased patent retrieval is rarely investigated. Based on traits of patent drawing images, we present a simple and lightweight model for this task.  \nWithout bells and whistles, this approach significantly outperforms other counterparts on a large-scale benchmark and noticeably improves the state-of-the-art by 33.5% with the mean average precision (mAP) score.  \nFurther experiments reveal that this model can be elaborately scaled up to achieve a surprisingly high mAP of 93 .5% . Our method ranks first in the ECCV 2022 Patent Diagram Image Retrieval Challenge.  \nKeywords: Image-Based Patent Retrieval · Patent Search · SketchBased Image Retrieval.  \n1 Introduction  \nWith a large and ever-increasing number of scientific and technical documents on the web every year, diagram image retrieval and analysis [33,1] become crucially important for intelligent document processing and understanding. Different from natural images that contain rich information about color, texture, and intensity, diagram images only involve line and shape. Although existing computer vision methods have achieved significant progress on natural image understanding (e.g. , classification, detection and retrieval) [12,26,16,31], diagram image understanding retains a challenging and less developed area.  \nImage-based patent retrieval [11,2,37], one of the most typical research topics of diagram image understanding, has drawn increasing interest from intellectual property and information retrieval communities. The need for effective and efficient systems of patent retrieval becomes inevitably crucial due to the tremendous amounts of patent data. As visual image plays an essential role in  \n* Equal Contribution. B Corresponding author.  \n2 H. Wang and Y. Zhang  \nFig. 1: An illustration of image-based patent retrieval.  \ninformation retrieval, patent image search can help people quickly understand the patent contents and check differences between patents. However, accurate patent retrieval based on visual content remains an open challenge. Commercial search engines such as Google and Baidu currently fail to retrieve relevant patent images with the query of a patent drawing [15] . An example of this task is illustrated in Fig. 1(a), where many unrelated drawing images with a similar style to the query are retrieved.  \nOne of the well-concerned problems in computer vision most related to imagebased patent retrieval is sketch-based image retrieval [9] . Although both are abstract drawings, there are subtle differences between sketch images and patent drawings. As shown in Fig. 1(b), patent drawings contain more detailed contoursand patterns about the object when compared with sketch images.  \nExisting patent search systems mostly use textual and metadata information. Only little published work studies content- and image-based patent retrieval. These approaches represent the patent drawing with relational skeletons [13], edge orientation autocorrelogram [27], contour description matrix [38], adaptive hierarchical density histogram [30], etc. For example, the patent retrieval system PATSEEK [27] applies shape-based retrieval and represents a patent image with an edge orientation autocorrelogram. However, these representations are lowlevel and handcrafted visual features, which are not discriminative enough for large-scale applicati","cbCairc8FYLVP9Hg","https://ap.wps.com/l/cbCairc8FYLVP9Hg","pdf",1657991,12,"English","# Abstract\n# Introduction\n## Problem background and motivation\n## Challenges in visual patent retrieval\n## Relation to sketch-based image retrieval\n# Related Work\n## Image-Based Patent Retrieval","[{\"question\":\"Why is content-based image-based patent retrieval challenging?\",\"answer\":\"Existing systems mostly rely on textual and metadata signals, while retrieval from patent drawing visuals remains an open challenge. Diagram images have structural characteristics different from natural images, and prior visual representations can be insufficiently discriminative at scale.\"},{\"question\":\"What does the proposed method target in patent diagram images?\",\"answer\":\"The approach is designed by exploiting traits of patent drawing images. It uses a lightweight backbone plus a neck structure to produce low-dimensional representations for efficient retrieval, trained with a classification loss in a geometric angular space.\"},{\"question\":\"What performance results and recognition does the paper report?\",\"answer\":\"On a large-scale benchmark, the method improves mean average precision by 33.5% over other approaches, and further scaling reaches 93.5% mAP. The method ranks first in the ECCV 2022 Patent Diagram Image Retrieval Challenge.\"}]","Learning Efficient Representations for Image-Based Patent Retrieval | PDF"]