{"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/cvae-gan-fine-grained-image-generation","title":"CVAE-GAN: Fine-Grained Image Generation through Asymmetric Training","arxiv_id":"1703.10155","date":"2017-03-29","proceeding":"ICCV 2017 10","authors":["Jianmin Bao","Dong Chen","Fang Wen","Houqiang Li","Gang Hua"],"abstract":"We present variational generative adversarial networks, a general learning\nframework that combines a variational auto-encoder with a generative\nadversarial network, for synthesizing images in fine-grained categories, such\nas faces of a specific person or objects in a category. Our approach models an\nimage as a composition of label and latent attributes in a probabilistic model.\nBy varying the fine-grained category label fed into the resulting generative\nmodel, we can generate images in a specific category with randomly drawn values\non a latent attribute vector. Our approach has two novel aspects. First, we\nadopt a cross entropy loss for the discriminative and classifier network, but a\nmean discrepancy objective for the generative network. This kind of asymmetric\nloss function makes the GAN training more stable. Second, we adopt an encoder\nnetwork to learn the relationship between the latent space and the real image\nspace, and use pairwise feature matching to keep the structure of generated\nimages. We experiment with natural images of faces, flowers, and birds, and\ndemonstrate that the proposed models are capable of generating realistic and\ndiverse samples with fine-grained category labels. We further show that our\nmodels can be applied to other tasks, such as image inpainting,\nsuper-resolution, and data augmentation for training better face recognition\nmodels.","url_abs":"http://arxiv.org/abs/1703.10155v2","url_pdf":"http://arxiv.org/pdf/1703.10155v2.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":"cvae-gan-fine-grained-image-generation","repo_url":"https://github.com/One-sixth/CVAE-GAN_tensorlayer","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}},{"paper_slug":"cvae-gan-fine-grained-image-generation","repo_url":"https://github.com/Ram81/AC-VAEGAN-PyTorch","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"cvae-gan-fine-grained-image-generation","repo_url":"https://github.com/pranavbudhwant/ACVAEGAN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"attribute","task_name":"Attribute"},{"task_slug":"data-augmentation","task_name":"Data Augmentation"},{"task_slug":"face-recognition","task_name":"Face Recognition"},{"task_slug":null,"task_name":"Generative Adversarial Network"},{"task_slug":"image-generation","task_name":"Image Generation"},{"task_slug":"image-inpainting","task_name":"Image Inpainting"},{"task_slug":"super-resolution","task_name":"Super-Resolution"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"feature-matching","method_name":"GAN Feature Matching"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1703.10155","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}