{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/style-separation-and-synthesis-via-generative","title":"Style Separation and Synthesis via Generative Adversarial Networks","arxiv_id":"1811.02740","date":"2018-11-07","proceeding":null,"authors":["Rui Zhang","Sheng Tang","Yu Li","Junbo Guo","Yongdong Zhang","Jintao Li","Shuicheng Yan"],"abstract":"Style synthesis attracts great interests recently, while few works focus on\nits dual problem \"style separation\". In this paper, we propose the Style\nSeparation and Synthesis Generative Adversarial Network (S3-GAN) to\nsimultaneously implement style separation and style synthesis on object\nphotographs of specific categories. Based on the assumption that the object\nphotographs lie on a manifold, and the contents and styles are independent, we\nemploy S3-GAN to build mappings between the manifold and a latent vector space\nfor separating and synthesizing the contents and styles. The S3-GAN consists of\nan encoder network, a generator network, and an adversarial network. The\nencoder network performs style separation by mapping an object photograph to a\nlatent vector. Two halves of the latent vector represent the content and style,\nrespectively. The generator network performs style synthesis by taking a\nconcatenated vector as input. The concatenated vector contains the style half\nvector of the style target image and the content half vector of the content\ntarget image. Once obtaining the images from the generator network, an\nadversarial network is imposed to generate more photo-realistic images.\nExperiments on CelebA and UT Zappos 50K datasets demonstrate that the S3-GAN\nhas the capacity of style separation and synthesis simultaneously, and could\ncapture various styles in a single model.","url_abs":"http://arxiv.org/abs/1811.02740v1","url_pdf":"http://arxiv.org/pdf/1811.02740v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"style-separation-and-synthesis-via-generative","repo_url":"https://github.com/nirey10/S3-pytorch","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"style-separation-and-synthesis-via-generative","repo_url":"https://github.com/nirey10/S3-Keras","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"tf","reach":{"status":"ok"}}],"tasks":[{"task_slug":null,"task_name":"Generative Adversarial Network"},{"task_slug":"object","task_name":"Object"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1811.02740","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}