{"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/learning-modulated-transformation-in-gans","title":"Learning Modulated Transformation in GANs","arxiv_id":null,"date":"2023-09-21","proceeding":"NeurIPS 2023 11","authors":[],"abstract":"The success of style-based generators largely benefits from style modulation,\nwhich helps take care of the cross-instance variation within data. However, the\ninstance-wise stochasticity is typically introduced via regular convolution, where\nkernels interact with features at some fixed locations, limiting its capacity for\nmodeling geometric variation. To alleviate this problem, we equip the generator\nin generative adversarial networks (GANs) with a plug-and-play module, termed\nas modulated transformation module (MTM). This module predicts spatial offsets\nunder the control of latent codes, based on which the convolution operation can\nbe applied at variable locations for different instances, and hence offers the model\nan additional degree of freedom to handle geometry deformation. Extensive\nexperiments suggest that our approach can be faithfully generalized to various\ngenerative tasks, including image generation, 3D-aware image synthesis, and\nvideo generation, and get compatible with state-of-the-art frameworks without\nany hyper-parameter tuning. It is noteworthy that, towards human generation on\nthe challenging TaiChi dataset, we improve the FID of StyleGAN3 from 21.36 to\n13.60, demonstrating the efficacy of learning modulated geometry transformation.\nCode and models are available at https://github.com/limbo0000/mtm.Submission Number: 3123","url_abs":"https://openreview.net/forum?id=h8vJVABiBP","url_pdf":"https://openreview.net/pdf?id=h8vJVABiBP","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":"learning-modulated-transformation-in-gans","repo_url":"https://github.com/limbo0000/mtm","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":null}],"tasks":[],"methods":[{"method_slug":"convolution","method_name":"Convolution"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}