{"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/training-deep-face-recognition-systems-with","title":"Training Deep Face Recognition Systems with Synthetic Data","arxiv_id":"1802.05891","date":"2018-02-16","proceeding":null,"authors":["Adam Kortylewski","Andreas Schneider","Thomas Gerig","Bernhard Egger","Andreas Morel-Forster","Thomas Vetter"],"abstract":"Recent advances in deep learning have significantly increased the performance\nof face recognition systems. The performance and reliability of these models\ndepend heavily on the amount and quality of the training data. However, the\ncollection of annotated large datasets does not scale well and the control over\nthe quality of the data decreases with the size of the dataset. In this work,\nwe explore how synthetically generated data can be used to decrease the number\nof real-world images needed for training deep face recognition systems. In\nparticular, we make use of a 3D morphable face model for the generation of\nimages with arbitrary amounts of facial identities and with full control over\nimage variations, such as pose, illumination, and background. In our\nexperiments with an off-the-shelf face recognition software we observe the\nfollowing phenomena: 1) The amount of real training data needed to train\ncompetitive deep face recognition systems can be reduced significantly. 2)\nCombining large-scale real-world data with synthetic data leads to an increased\nperformance. 3) Models trained only on synthetic data with strong variations in\npose, illumination, and background perform very well across different datasets\neven without dataset adaptation. 4) The real-to-virtual performance gap can be\nclosed when using synthetic data for pre-training, followed by fine-tuning with\nreal-world images. 5) There are no observable negative effects of pre-training\nwith synthetic data. Thus, any face recognition system in our experiments\nbenefits from using synthetic face images. The synthetic data generator, as\nwell as all experiments, are publicly available.","url_abs":"http://arxiv.org/abs/1802.05891v1","url_pdf":"http://arxiv.org/pdf/1802.05891v1.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":"training-deep-face-recognition-systems-with","repo_url":"https://github.com/unibas-gravis/parametric-face-image-generator","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"training-deep-face-recognition-systems-with","repo_url":"https://github.com/Arneli/image-generator-for-BScThesis","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"Apache-2.0"}}],"tasks":[{"task_slug":"face-model","task_name":"Face Model"},{"task_slug":"face-recognition","task_name":"Face Recognition"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1802.05891","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}