{"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/dense-intrinsic-appearance-flow-for-human","title":"Dense Intrinsic Appearance Flow for Human Pose Transfer","arxiv_id":"1903.11326","date":"2019-03-27","proceeding":"CVPR 2019 6","authors":["Yining Li","Chen Huang","Chen Change Loy"],"abstract":"We present a novel approach for the task of human pose transfer, which aims\nat synthesizing a new image of a person from an input image of that person and\na target pose. We address the issues of limited correspondences identified\nbetween keypoints only and invisible pixels due to self-occlusion. Unlike\nexisting methods, we propose to estimate dense and intrinsic 3D appearance flow\nto better guide the transfer of pixels between poses. In particular, we wish to\ngenerate the 3D flow from just the reference and target poses. Training a\nnetwork for this purpose is non-trivial, especially when the annotations for 3D\nappearance flow are scarce by nature. We address this problem through a flow\nsynthesis stage. This is achieved by fitting a 3D model to the given pose pair\nand project them back to the 2D plane to compute the dense appearance flow for\ntraining. The synthesized ground-truths are then used to train a feedforward\nnetwork for efficient mapping from the input and target skeleton poses to the\n3D appearance flow. With the appearance flow, we perform feature warping on the\ninput image and generate a photorealistic image of the target pose. Extensive\nresults on DeepFashion and Market-1501 datasets demonstrate the effectiveness\nof our approach over existing methods. Our code is available at\nhttp://mmlab.ie.cuhk.edu.hk/projects/pose-transfer","url_abs":"http://arxiv.org/abs/1903.11326v1","url_pdf":"http://arxiv.org/pdf/1903.11326v1.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":"dense-intrinsic-appearance-flow-for-human","repo_url":"https://github.com/ly015/intrinsic_flow","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"pose-transfer","task_name":"Pose Transfer"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1903.11326","atlas_url":"https://app.syntology.ai/?focus=1903.11326","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}