{"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-symmetry-consistent-deep-cnns-for","title":"Learning Symmetry Consistent Deep CNNs for Face Completion","arxiv_id":"1812.07741","date":"2018-12-19","proceeding":null,"authors":["Xiaoming Li","Ming Liu","Jieru Zhu","WangMeng Zuo","Meng Wang","Guosheng Hu","Lei Zhang"],"abstract":"Deep convolutional networks (CNNs) have achieved great success in face\ncompletion to generate plausible facial structures. These methods, however, are\nlimited in maintaining global consistency among face components and recovering\nfine facial details. On the other hand, reflectional symmetry is a prominent\nproperty of face image and benefits face recognition and consistency modeling,\nyet remaining uninvestigated in deep face completion. In this work, we leverage\ntwo kinds of symmetry-enforcing subnets to form a symmetry-consistent CNN model\n(i.e., SymmFCNet) for effective face completion. For missing pixels on only one\nof the half-faces, an illumination-reweighted warping subnet is developed to\nguide the warping and illumination reweighting of the other half-face. As for\nmissing pixels on both of half-faces, we present a generative reconstruction\nsubnet together with a perceptual symmetry loss to enforce symmetry consistency\nof recovered structures. The SymmFCNet is constructed by stacking generative\nreconstruction subnet upon illumination-reweighted warping subnet, and can be\nend-to-end learned from training set of unaligned face images. Experiments show\nthat SymmFCNet can generate high quality results on images with synthetic and\nreal occlusion, and performs favorably against state-of-the-arts.","url_abs":"http://arxiv.org/abs/1812.07741v1","url_pdf":"http://arxiv.org/pdf/1812.07741v1.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":"learning-symmetry-consistent-deep-cnns-for","repo_url":"https://github.com/csxmli2016/SymmFCNet","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"face-recognition","task_name":"Face Recognition"},{"task_slug":"facial-inpainting","task_name":"Facial Inpainting"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/facial-inpainting-on-vggface2","task":"Facial Inpainting","dataset":"VggFace2","model":"SymmFCNet (Full)","rank_in_archive_order":1,"of":1,"metrics":{"PSNR":"27.81"},"uses_additional_data":false},{"leaderboard":"/sota/facial-inpainting-on-webface","task":"Facial Inpainting","dataset":"WebFace","model":"SymmFCNet (Full)","rank_in_archive_order":1,"of":1,"metrics":{"PSNR":"27.22"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/1812.07741","atlas_url":"https://app.syntology.ai/?focus=1812.07741","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}