{"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-augmentation-data-augmentation-via","title":"Style Augmentation: Data Augmentation via Style Randomization","arxiv_id":"1809.05375","date":"2018-09-14","proceeding":null,"authors":["Philip T. Jackson","Amir Atapour-Abarghouei","Stephen Bonner","Toby Breckon","Boguslaw Obara"],"abstract":"We introduce style augmentation, a new form of data augmentation based on\nrandom style transfer, for improving the robustness of convolutional neural\nnetworks (CNN) over both classification and regression based tasks. During\ntraining, our style augmentation randomizes texture, contrast and color, while\npreserving shape and semantic content. This is accomplished by adapting an\narbitrary style transfer network to perform style randomization, by sampling\ninput style embeddings from a multivariate normal distribution instead of\ninferring them from a style image. In addition to standard classification\nexperiments, we investigate the effect of style augmentation (and data\naugmentation generally) on domain transfer tasks. We find that data\naugmentation significantly improves robustness to domain shift, and can be used\nas a simple, domain agnostic alternative to domain adaptation. Comparing style\naugmentation against a mix of seven traditional augmentation techniques, we\nfind that it can be readily combined with them to improve network performance.\nWe validate the efficacy of our technique with domain transfer experiments in\nclassification and monocular depth estimation, illustrating consistent\nimprovements in generalization.","url_abs":"http://arxiv.org/abs/1809.05375v2","url_pdf":"http://arxiv.org/pdf/1809.05375v2.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-augmentation-data-augmentation-via","repo_url":"https://github.com/philipjackson/style-augmentation","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"classification-1","task_name":"Classification"},{"task_slug":"data-augmentation","task_name":"Data Augmentation"},{"task_slug":"depth-estimation","task_name":"Depth Estimation"},{"task_slug":"domain-adaptation","task_name":"Domain Adaptation"},{"task_slug":"classification","task_name":"General Classification"},{"task_slug":"monocular-depth-estimation","task_name":"Monocular Depth Estimation"},{"task_slug":"style-transfer","task_name":"Style Transfer"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1809.05375","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}