{"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/noisytwins-class-consistent-and-diverse-image","title":"NoisyTwins: Class-Consistent and Diverse Image Generation through StyleGANs","arxiv_id":"2304.05866","date":"2023-04-12","proceeding":"CVPR 2023 1","authors":["Harsh Rangwani","Lavish Bansal","Kartik Sharma","Tejan Karmali","Varun Jampani","R. Venkatesh Babu"],"abstract":"StyleGANs are at the forefront of controllable image generation as they produce a latent space that is semantically disentangled, making it suitable for image editing and manipulation. However, the performance of StyleGANs severely degrades when trained via class-conditioning on large-scale long-tailed datasets. We find that one reason for degradation is the collapse of latents for each class in the $\\mathcal{W}$ latent space. With NoisyTwins, we first introduce an effective and inexpensive augmentation strategy for class embeddings, which then decorrelates the latents based on self-supervision in the $\\mathcal{W}$ space. This decorrelation mitigates collapse, ensuring that our method preserves intra-class diversity with class-consistency in image generation. We show the effectiveness of our approach on large-scale real-world long-tailed datasets of ImageNet-LT and iNaturalist 2019, where our method outperforms other methods by $\\sim 19\\%$ on FID, establishing a new state-of-the-art.","url_abs":"https://arxiv.org/abs/2304.05866v1","url_pdf":"https://arxiv.org/pdf/2304.05866v1.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":"noisytwins-class-consistent-and-diverse-image","repo_url":"https://github.com/val-iisc/NoisyTwins","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"conditional-image-generation","task_name":"Conditional Image Generation"},{"task_slug":"diversity","task_name":"Diversity"},{"task_slug":"image-generation","task_name":"Image Generation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/conditional-image-generation-on-imagenet-lt","task":"Conditional Image Generation","dataset":"ImageNet-LT","model":"StyleGAN2 + NoisyTwins","rank_in_archive_order":1,"of":1,"metrics":{"FID":"21.29"},"uses_additional_data":false},{"leaderboard":"/sota/image-generation-on-inaturalist-2019","task":"Image Generation","dataset":"iNaturalist 2019","model":"StyeGAN2 + NoisyTwins","rank_in_archive_order":1,"of":2,"metrics":{"FID":"11.46"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2304.05866","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}