{"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/deep-identity-aware-transfer-of-facial","title":"Deep Identity-aware Transfer of Facial Attributes","arxiv_id":"1610.05586","date":"2016-10-18","proceeding":null,"authors":["Mu Li","WangMeng Zuo","David Zhang"],"abstract":"This paper presents a Deep convolutional network model for Identity-Aware\nTransfer (DIAT) of facial attributes. Given the source input image and the\nreference attribute, DIAT aims to generate a facial image that owns the\nreference attribute as well as keeps the same or similar identity to the input\nimage. In general, our model consists of a mask network and an attribute\ntransform network which work in synergy to generate a photo-realistic facial\nimage with the reference attribute. Considering that the reference attribute\nmay be only related to some parts of the image, the mask network is introduced\nto avoid the incorrect editing on attribute irrelevant region. Then the\nestimated mask is adopted to combine the input and transformed image for\nproducing the transfer result. For joint training of transform network and mask\nnetwork, we incorporate the adversarial attribute loss, identity-aware adaptive\nperceptual loss, and VGG-FACE based identity loss. Furthermore, a denoising\nnetwork is presented to serve for perceptual regularization to suppress the\nartifacts in transfer result, while an attribute ratio regularization is\nintroduced to constrain the size of attribute relevant region. Our DIAT can\nprovide a unified solution for several representative facial attribute transfer\ntasks, e.g., expression transfer, accessory removal, age progression, and\ngender transfer, and can be extended for other face enhancement tasks such as\nface hallucination. The experimental results validate the effectiveness of the\nproposed method. Even for the identity-related attribute (e.g., gender), our\nDIAT can obtain visually impressive results by changing the attribute while\nretaining most identity-aware features.","url_abs":"http://arxiv.org/abs/1610.05586v2","url_pdf":"http://arxiv.org/pdf/1610.05586v2.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":"attribute","task_name":"Attribute"},{"task_slug":"denoising","task_name":"Denoising"},{"task_slug":"face-hallucination","task_name":"Face Hallucination"},{"task_slug":"hallucination","task_name":"Hallucination"},{"task_slug":"image-to-image-translation","task_name":"Image-to-Image Translation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/image-to-image-translation-on-rafd","task":"Image-to-Image Translation","dataset":"RaFD","model":"DIA","rank_in_archive_order":2,"of":4,"metrics":{"Classification Error":"4.10%"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/1610.05586","atlas_url":"https://app.syntology.ai/?focus=1610.05586","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}