{"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/semi-supervised-adversarial-learning-to","title":"Semi-supervised Adversarial Learning to Generate Photorealistic Face Images of New Identities from 3D Morphable Model","arxiv_id":"1804.03675","date":"2018-04-10","proceeding":"ECCV 2018 9","authors":["Baris Gecer","Binod Bhattarai","Josef Kittler","Tae-Kyun Kim"],"abstract":"We propose a novel end-to-end semi-supervised adversarial framework to\ngenerate photorealistic face images of new identities with wide ranges of\nexpressions, poses, and illuminations conditioned by a 3D morphable model.\nPrevious adversarial style-transfer methods either supervise their networks\nwith large volume of paired data or use unpaired data with a highly\nunder-constrained two-way generative framework in an unsupervised fashion. We\nintroduce pairwise adversarial supervision to constrain two-way domain\nadaptation by a small number of paired real and synthetic images for training\nalong with the large volume of unpaired data. Extensive qualitative and\nquantitative experiments are performed to validate our idea. Generated face\nimages of new identities contain pose, lighting and expression diversity and\nqualitative results show that they are highly constraint by the synthetic input\nimage while adding photorealism and retaining identity information. We combine\nface images generated by the proposed method with the real data set to train\nface recognition algorithms. We evaluated the model on two challenging data\nsets: LFW and IJB-A. We observe that the generated images from our framework\nconsistently improves over the performance of deep face recognition network\ntrained with Oxford VGG Face dataset and achieves comparable results to the\nstate-of-the-art.","url_abs":"http://arxiv.org/abs/1804.03675v1","url_pdf":"http://arxiv.org/pdf/1804.03675v1.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":"semi-supervised-adversarial-learning-to","repo_url":"https://github.com/barisgecer/facegan","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}}],"tasks":[{"task_slug":"domain-adaptation","task_name":"Domain Adaptation"},{"task_slug":"face-generation","task_name":"Face Generation"},{"task_slug":"face-recognition","task_name":"Face Recognition"},{"task_slug":"face-verification","task_name":"Face Verification"},{"task_slug":"style-transfer","task_name":"Style Transfer"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"max-pooling","method_name":"Max Pooling"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"softmax","method_name":"Softmax"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/face-verification-on-ijb-a","task":"Face Verification","dataset":"IJB-A","model":"VGG + GANFaces","rank_in_archive_order":17,"of":17,"metrics":{"TAR @ FAR=0.001":"18.768","TAR @ FAR=0.01":"53.507%"},"uses_additional_data":false}],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1804.03675","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}