{"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/virtual-conditional-generative-adversarial","title":"Virtual Conditional Generative Adversarial Networks","arxiv_id":"1901.09822","date":"2019-01-25","proceeding":null,"authors":["Haifeng Shi","Guanyu Cai","Yuqin Wang","Shaohua Shang","Lianghua He"],"abstract":"When trained on multimodal image datasets, normal Generative Adversarial\nNetworks (GANs) are usually outperformed by class-conditional GANs and ensemble\nGANs, but conditional GANs is restricted to labeled datasets and ensemble GANs\nlack efficiency. We propose a novel GAN variant called virtual conditional GAN\n(vcGAN) which is not only an ensemble GAN with multiple generative paths while\nadding almost zero network parameters, but also a conditional GAN that can be\ntrained on unlabeled datasets without explicit clustering steps or objectives\nother than the adversary loss. Inside the vcGAN's generator, a learnable\n``analog-to-digital converter (ADC)\" module maps a slice of the inputted\nmultivariate Gaussian noise to discrete/digital noise (virtual label),\naccording to which a selector selects the corresponding generative path to\nproduce the sample. All the generative paths share the same decoder network\nwhile in each path the decoder network is fed with a concatenation of a\ndifferent pre-computed amplified one-hot vector and the inputted Gaussian\nnoise. We conducted a lot of experiments on several balanced/imbalanced image\ndatasets to demonstrate that vcGAN converges faster and achieves improved\nFrech\\'et Inception Distance (FID). In addition, we show the training byproduct\nthat the ADC in vcGAN learned the categorical probability of each mode and that\neach generative path generates samples of specific mode, which enables\nclass-conditional sampling. Codes are available at\n\\url{https://github.com/annonnymmouss/vcgan}","url_abs":"http://arxiv.org/abs/1901.09822v1","url_pdf":"http://arxiv.org/pdf/1901.09822v1.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":"virtual-conditional-generative-adversarial","repo_url":"https://github.com/annonnymmouss/vcgan","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":null}],"tasks":[{"task_slug":"clustering","task_name":"Clustering"},{"task_slug":"conditional-image-generation","task_name":"Conditional Image Generation"},{"task_slug":"decoder","task_name":"Decoder"},{"task_slug":"image-generation","task_name":"Image Generation"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}