[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-155411-105":3,"detail-sidebar-cat-0-en-105":80,"doc-detail-155411-en":130},{"code":4,"msg":5,"data":6},0,"ok",{"site_id":7,"language":8,"slug":9,"title":10,"keywords":11,"description":12,"schema_data":13,"social_meta":73,"head_meta":75,"extra_data":77,"updated_unix":79},105,"en","semantic-manga-character-sketch-generation-journal-of-image-and-graphics","Semantic Manga Character Sketch Generation - Journal of Image and Graphics","","A comic is an effective medium for information and knowledge transfer, and manga’s key element is the character, whose illustration requires experience. Earlier GAN-based works can synthesize character images but provide limited controllability over the generated results. This study proposes conditional GANs guided by semantic masks to control posture, anatomy, and basic clothing style, and to systematically specify desired character attributes. Human evaluation shows the method can competently synthesize sketch-like manga character outputs.",{"@graph":14,"@context":72},[15,34,55],{"@type":16,"itemListElement":17},"BreadcrumbList",[18,23,27,31],{"item":19,"name":20,"@type":21,"position":22},"https://docshare.wps.com","Home","ListItem",1,{"item":24,"name":25,"@type":21,"position":26},"https://docshare.wps.com/document/","Document",2,{"item":28,"name":29,"@type":21,"position":30},"https://docshare.wps.com/document/research-report/","Research & Report",3,{"item":32,"name":10,"@type":21,"position":33},"https://docshare.wps.com/document/semantic-manga-character-sketch-generation-journal-of-image-and-graphics/155411/",4,{"url":32,"name":10,"@type":35,"image":36,"author":41,"headline":10,"publisher":44,"fileFormat":47,"inLanguage":8,"description":12,"dateModified":48,"datePublished":49,"encodingFormat":47,"isAccessibleForFree":50,"interactionStatistic":51},"DigitalDocument",{"url":37,"@type":38,"width":39,"height":40},"https://docshare.wps.com/thumbnails/semantic-manga-character-sketch-generation-journal-of-image-and-graphics/155411.png","ImageObject",300,407,{"name":42,"@type":43},"Levi","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-10-17","2026-08-28",true,{"@type":52,"interactionType":53,"userInteractionCount":33},"InteractionCounter",{"@type":54},"ViewAction",{"@type":56,"mainEntity":57},"FAQPage",[58,64,68],{"name":59,"@type":60,"acceptedAnswer":61},"Why do vanilla GANs struggle with manga character generation controllability?","Question",{"text":62,"@type":63},"Vanilla GANs can produce high-quality images but cannot reliably control specific attributes of the synthesized character. This limits control over posture, anatomy, and clothing style during generation.","Answer",{"name":65,"@type":60,"acceptedAnswer":66},"How does the proposed method improve control over generated manga characters?",{"text":67,"@type":63},"It uses a conditional GAN with a semantic mask as the conditioning input. The semantic mask encodes components such as posture, anatomy, and basic dressing style, guiding the generator to produce a more targeted sketch.",{"name":69,"@type":60,"acceptedAnswer":70},"What semantic information is extracted and how is it used in training?",{"text":71,"@type":63},"A self-correction human parsing framework is retrained to extract semantic components like upper torso, legs, and hair from manga datasets. Pix2pix with SPADE then trains a generator to synthesize images based on these semantic masks.","https://schema.org",{"og:url":32,"og:type":74,"og:title":10,"og:site_name":45,"og:description":12},"article",{"robots":76,"canonical":32},"index,follow",{"doc_id":78,"site_id":7},155411,1787911616,{"code":4,"msg":81,"data":82},"success",[83,87,91,95,100,105,110,114,119,122,126],{"id":22,"doc_module":4,"doc_module_name":25,"category_name":84,"show_sort_weight":85,"slug":86},"Story & Novel",90,"story-novel",{"id":26,"doc_module":4,"doc_module_name":25,"category_name":88,"show_sort_weight":89,"slug":90},"Literature",80,"literature",{"id":33,"doc_module":4,"doc_module_name":25,"category_name":92,"show_sort_weight":93,"slug":94},"Exam",70,"exam",{"id":96,"doc_module":4,"doc_module_name":25,"category_name":97,"show_sort_weight":98,"slug":99},5,"Comic",60,"comic",{"id":101,"doc_module":4,"doc_module_name":25,"category_name":102,"show_sort_weight":103,"slug":104},6,"Technology",50,"technology",{"id":106,"doc_module":4,"doc_module_name":25,"category_name":107,"show_sort_weight":108,"slug":109},7,"Healthcare",40,"healthcare",{"id":111,"doc_module":4,"doc_module_name":25,"category_name":29,"show_sort_weight":112,"slug":113},8,30,"research-report",{"id":115,"doc_module":4,"doc_module_name":25,"category_name":116,"show_sort_weight":117,"slug":118},9,"Religion & Spirituality",20,"religion-spirituality",{"id":117,"doc_module":4,"doc_module_name":25,"category_name":120,"show_sort_weight":117,"slug":121},"World Cup","world-cup",{"id":123,"doc_module":4,"doc_module_name":25,"category_name":124,"show_sort_weight":123,"slug":125},10,"Lifestyle","lifestyle",{"id":127,"doc_module":4,"doc_module_name":25,"category_name":128,"show_sort_weight":96,"slug":129},19,"General","general",{"code":4,"msg":81,"data":131},{"doc_id":78,"user_id":132,"nickname":42,"user_avatar":133,"doc_module":4,"category_id":111,"category_name":29,"doc_title":10,"doc_description":12,"doc_content":134,"file_id":135,"file_url":136,"file_type":137,"file_size":138,"view_count":33,"is_deleted":4,"is_public":22,"is_downloadable":22,"audit_status":22,"page_count":101,"language":139,"language_code":8,"site_id":7,"html_lang":8,"table_of_contents":140,"faqs":141,"seo_title":142,"seo_description":12,"update_tm":79,"read_time":143,"preview_page_count":4},7971461740909,"https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d","Journal of Image and Graphics, Vol. 10, No. 4, December 2022  \nSemantic Manga Character Sketch Generation  \nKittinun Aukkapinyo and Seiji Hotta  \nDepartment of Computer and Information Science, Tokyo University of Agriculture and Technology, Tokyo, Japan  \n[Email: kittinun.auk@gmail.com](Email: kittinun.auk@gmail.com), [s-hotta@cc.tuat.ac.jp](s-hotta@cc.tuat.ac.jp)  \nAbstract—A comic is one of the approaches in information and knowledge transfer. Manga is a unique and attractive comic style that originated in Japan. One of the important components in each manga is a character. It requires experience and knowledge in illustrating a manga character. Several research studies employed Generative Adversarial Networks (GANs) to illustrate a character. Although unconditional GANs could produce a high-quality image, it still lacks controllability over synthesized character. This research proposed an approach to employ conditional GANs with a semantic mask to control posture, anatomy, and basic dressing style during the synthesizing process. It also introduces an approach to systematically specify desired character’s style and parse a character into a semantic mask. As a result of a human-based evaluation, this research can competently employ a semantic mask as a condition to synthesize a decent sketch version of a manga character.  \nIndex Terms—image generation, comic computing, deep learning  \nI. INTRODUCTION  \nA. Background  \nManga is a comic that originated in Japan, which can contribute to information transfer. It can improve recognition of particular information, such as academic knowledge and marketing content, through its attractiveness. It can also serve as a tool to transfer complicated knowledge to others. A character is one of the most important components in each manga creation. Although character drawing is an easy task for professional human artists, it is considerably difficult for inexperienced humans. In addition, it is also a challenging task for a machine to creatively synthesize a manga character.  \nA generative adversarial network [1] is one of the unsupervised machine learning techniques that can be used to train a computer to synthesize new contents from training data such as images, videos, and sounds. It consists of two deep neural networks architecture - a generator and a discriminator. A manga character can be represented as an image composed of strokes and color information, such features allow a generator to learn to illustrate a character. However, a vanilla GAN has an issue of controllability of a generated image. So, a conditional GAN (cGAN) is introduced to provide some conditions along with input during training. It is to control output with  \nManuscript received May 6, 2022; revised September 20, 2022.  \na certain condition such as a less information version of the final image.  \nThere are many applications of generative adversarial networks in rendering and translating real-world images, such as black-and-white image colorization. It is an uncomplicated task when compared to manga characters due to their surrealism. In addition, there is an extremely high variance in definitions of real-world things in each manga. For example, a character’s anatomy and fantasy clothing style. It is more challenging for a generative model to learn the extreme differences in features as to what defines a particular object.  \nSeveral research studies employed a Generative Adversarial Network (GAN) to generate a character ranging from coarse-to-fine solutions where users provide some conditions, such as a sketch or real-world image [2] -[8] . Chen et al. [2], introduced a framework, CartoonGAN, to train a computer to transform an input image of realworld scenery into a comic style. Another one is Autopainter [3] . This paper proposes a methodology to train a computer to paint and generate comic drawings from its sketch version using a GAN. In addition, Li et al. [8] generated a character from a random noise vector. Alth","cbCaibuFNMHwhokP","https://ap.wps.com/l/cbCaibuFNMHwhokP","pdf",1769033,"English","# Abstract\n# I. Introduction\n## Background\n## Contribution","[{\"question\":\"Why do vanilla GANs struggle with manga character generation controllability?\",\"answer\":\"Vanilla GANs can produce high-quality images but cannot reliably control specific attributes of the synthesized character. This limits control over posture, anatomy, and clothing style during generation.\"},{\"question\":\"How does the proposed method improve control over generated manga characters?\",\"answer\":\"It uses a conditional GAN with a semantic mask as the conditioning input. The semantic mask encodes components such as posture, anatomy, and basic dressing style, guiding the generator to produce a more targeted sketch.\"},{\"question\":\"What semantic information is extracted and how is it used in training?\",\"answer\":\"A self-correction human parsing framework is retrained to extract semantic components like upper torso, legs, and hair from manga datasets. Pix2pix with SPADE then trains a generator to synthesize images based on these semantic masks.\"}]","Semantic Manga Character Sketch Generation - Journal of Image and Graphics | PDF",15]