[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-82297-en":3,"doc-seo-82297-105":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},82297,13056703019662,"Evangeline","https://ap-avatar.wpscdn.com/avatar/be000253a8e92610077?_k=1778726343310543188",8,"Research & Report","TextileNet: Towards Zero-shot Text-style Segmentation of Manuscripts","Automatic writer identification has advanced, yet archival paleography still faces limited labeled training data, open scribe sets, and degraded image quality. TextileNet introduces a fully convolutional multi-task network trained only on synthetic data to generate dense pixel-level texture embeddings and transfer them zero-shot to historical manuscript analysis. A paleographic visual quiz with 80 pair and triplet questions establishes a human baseline for late medieval script-style discrimination. Zero-shot retrieval uses sub-word granularity for hand and gender identification, with results advising caution when inferring gender from handwriting.","arXiv :2607 .09299v1 [ cs .CV] 10 Jul 2026  \nTextileNet: Towards Zero-shot Text-style Segmentation of Manuscripts  \nAnguelos Nicolaou 1[0000−0003−3818−8718]⋆ , Antonella Ambrosio2[0000−0001−6944−2147], Desiree Di Donato2 , and Georg  \nVogeler 1[0000−0002−1726−1712]  \n1 University of Graz, Graz, Austria  \n[anguelos.nicolaou@gmail.com](anguelos.nicolaou@gmail.com) , [georg.vogeler@uni-graz.at](georg.vogeler@uni-graz.at)  \n2 Università degli Studi di Napoli Federico II, Naples, Italy  \n[aambrosio@unina.it](aambrosio@unina.it) , [desireedidonato40@gmail.com](desireedidonato40@gmail.com)  \nAbstract. Automatic writer identification systems have progressed remarkably in recent years, yet their deployment in archival paleography remains limited by the scarcity of labeled training data, open scribe sets, and degraded image quality. We present TextileNet, a fully convolutional multi-task network trained exclusively on synthetic data to produce dense pixel-level texture embeddings, which we transfer zeroshot to historical manuscript analysis. As an original contribution to evaluation methodology, we designed a paleographic visual quiz of 80 pair and triplet questions and administered it to a range from lay participants to senior paleographers under strict anonymity, establishing to our knowledge for the first time a human baseline for script-style discrimination on late medieval text. We employ TextileNet embeddings to perform zero-shot retrieval on sub-word granularity for hand and gender identification. Our experimental results help in building the credibility of TextileNet in the paleographic domain, but more than that demonstrate in experimental terms that the question of gender in handwriting needs to be treated with caution. 3  \nKeywords: writer identification · paleography · texture segmentation · multi-task learning · zero-shot learning · historical manuscripts  \n1 Introduction  \n1.1 Writer identification methods  \nAutomatic writer identification has been an active field in document image analysis for several decades, driven by applications ranging from forensic document examination to the study of historical manuscripts. Early approaches established the viability of local texture statistics and sparse coding as writer-specific descriptors [20,18,17] . Even when deep learning dominated other domains, writer  \n⋆  \n3  \nCorresponding author.  \nsource code repository available: [https://github.com/anguelos/textstyle](https://github.com/anguelos/textstyle)  \n2 A. Nicolaou et al.  \nidentification state-of-the-art was using vector embedding techniques on handcrafted features [7] or hybrid approaches [8] until eventually deep learning dominated. Recent years have seen substantial advances through self-supervised representation learning, which reduces dependence on labeled training data [24], and through efforts to make attributions interpretable rather than treating the classifier as a black box [23] . Tools designed with the paleographer as the intended user have begun to appear [13], reflecting growing recognition that deployment context shapes what a useful method looks like. A persistent gap, however, separates laboratory performance from real archival deployment: published evaluations typically assume clean, exhaustively labeled datasets with a fixed set of scribes, while actual manuscript work involves degraded images, disputed or partial annotations, and open-set attribution problems. Critically, existing systems offer their most confident output precisely where the answer is already largely known, whereas scholars most need computational assistance in cases of genuine uncertainty—when a hand is unattested, when two attributions are plausibly equivalent, or when the material evidence is ambiguous.  \n1.2 From writers to gender  \nThe work for identifying automatically the gender of writers has seen less attention than work looking into the particular identities of the hands. In the communities of pattern recognition, handwriting ana","cbCaicys9NmwO4fW","https://ap.wps.com/l/cbCaicys9NmwO4fW","pdf",9383956,1,17,"English","en",105,"# Abstract\n# 1 Introduction\n## 1.1 Writer identification methods\n## 1.2 From writers to gender","[{\"question\":\"What is TextileNet and how is it trained?\",\"answer\":\"TextileNet is a fully convolutional multi-task network trained exclusively on synthetic data to produce dense pixel-level texture embeddings, which are then transferred zero-shot to historical manuscript analysis.\"},{\"question\":\"Why is writer identification in archival paleography still challenging?\",\"answer\":\"Deployments are limited by scarce labeled training data, open scribe sets, and degraded image quality, while evaluations often assume clean, exhaustively labeled datasets and a fixed set of writers.\"},{\"question\":\"How does the work assess script-style discrimination and what baseline is provided?\",\"answer\":\"The authors design a paleographic visual quiz using 80 pair and triplet questions and administer it under strict anonymity to participants ranging from lay readers to senior paleographers, establishing a human baseline for late medieval script-style discrimination.\"}]",1784179456,43,{"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},"textilenet-towards-zero-shot-text-style-segmentation-of-manuscripts","",{"@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/textilenet-towards-zero-shot-text-style-segmentation-of-manuscripts/82297/",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","2026-07-22","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},"What is TextileNet and how is it trained?","Question",{"text":75,"@type":76},"TextileNet is a fully convolutional multi-task network trained exclusively on synthetic data to produce dense pixel-level texture embeddings, which are then transferred zero-shot to historical manuscript analysis.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"Why is writer identification in archival paleography still challenging?",{"text":80,"@type":76},"Deployments are limited by scarce labeled training data, open scribe sets, and degraded image quality, while evaluations often assume clean, exhaustively labeled datasets and a fixed set of writers.",{"name":82,"@type":73,"acceptedAnswer":83},"How does the work assess script-style discrimination and what baseline is provided?",{"text":84,"@type":76},"The authors design a paleographic visual quiz using 80 pair and triplet questions and administer it under strict anonymity to participants ranging from lay readers to senior paleographers, establishing a human baseline for late medieval script-style discrimination.","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},[93,97,101,105,110,115,120,123,128,131,135],{"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":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":45,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":45,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]