{"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/jointly-optimize-data-augmentation-and","title":"Jointly Optimize Data Augmentation and Network Training: Adversarial Data Augmentation in Human Pose Estimation","arxiv_id":"1805.09707","date":"2018-05-24","proceeding":"CVPR 2018 6","authors":["Xi Peng","Zhiqiang Tang","Fei Yang","Rogerio Feris","Dimitris Metaxas"],"abstract":"Random data augmentation is a critical technique to avoid overfitting in\ntraining deep neural network models. However, data augmentation and network\ntraining are usually treated as two isolated processes, limiting the\neffectiveness of network training. Why not jointly optimize the two? We propose\nadversarial data augmentation to address this limitation. The main idea is to\ndesign an augmentation network (generator) that competes against a target\nnetwork (discriminator) by generating `hard' augmentation operations online.\nThe augmentation network explores the weaknesses of the target network, while\nthe latter learns from `hard' augmentations to achieve better performance. We\nalso design a reward/penalty strategy for effective joint training. We\ndemonstrate our approach on the problem of human pose estimation and carry out\na comprehensive experimental analysis, showing that our method can\nsignificantly improve state-of-the-art models without additional data efforts.","url_abs":"http://arxiv.org/abs/1805.09707v1","url_pdf":"http://arxiv.org/pdf/1805.09707v1.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":[],"tasks":[{"task_slug":"data-augmentation","task_name":"Data Augmentation"},{"task_slug":"pose-estimation","task_name":"Pose Estimation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/pose-estimation-on-leeds-sports-poses","task":"Pose Estimation","dataset":"Leeds Sports Poses","model":"Residual Hourglass + ASR + AHO","rank_in_archive_order":4,"of":18,"metrics":{"PCK":"94.5%"},"uses_additional_data":false},{"leaderboard":"/sota/pose-estimation-on-mpii-human-pose","task":"Pose Estimation","dataset":"MPII Human Pose","model":"Residual Hourglass +ASR+AHO","rank_in_archive_order":18,"of":46,"metrics":{"PCKh-0.5":"91.5"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/1805.09707","atlas_url":"https://app.syntology.ai/?focus=1805.09707","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}