{"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/190408241","title":"Deep Anomaly Detection for Generalized Face Anti-Spoofing","arxiv_id":"1904.08241","date":"2019-04-17","proceeding":null,"authors":["Daniel Pérez-Cabo","David Jiménez-Cabello","Artur Costa-Pazo","Roberto J. López-Sastre"],"abstract":"Face recognition has achieved unprecedented results, surpassing human\ncapabilities in certain scenarios. However, these automatic solutions are not\nready for production because they can be easily fooled by simple identity\nimpersonation attacks. And although much effort has been devoted to develop\nface anti-spoofing models, their generalization capacity still remains a\nchallenge in real scenarios. In this paper, we introduce a novel approach that\nreformulates the Generalized Presentation Attack Detection (GPAD) problem from\nan anomaly detection perspective. Technically, a deep metric learning model is\nproposed, where a triplet focal loss is used as a regularization for a novel\nloss coined \"metric-softmax\", which is in charge of guiding the learning\nprocess towards more discriminative feature representations in an embedding\nspace. Finally, we demonstrate the benefits of our deep anomaly detection\narchitecture, by introducing a few-shot a posteriori probability estimation\nthat does not need any classifier to be trained on the learned features. We\nconduct extensive experiments using the GRAD-GPAD framework that provides the\nlargest aggregated dataset for face GPAD. Results confirm that our approach is\nable to outperform all the state-of-the-art methods by a considerable margin.","url_abs":"http://arxiv.org/abs/1904.08241v1","url_pdf":"http://arxiv.org/pdf/1904.08241v1.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":"190408241","repo_url":"https://github.com/aoru45/Deep-Anomaly-Detection-for-Generalized-Face-Anti-Spoofing","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"anomaly-detection","task_name":"Anomaly Detection"},{"task_slug":"face-anti-spoofing","task_name":"Face Anti-Spoofing"},{"task_slug":"face-recognition","task_name":"Face Recognition"},{"task_slug":"metric-learning","task_name":"Metric Learning"},{"task_slug":null,"task_name":"Triplet"}],"methods":[{"method_slug":"focal-loss","method_name":"Focal Loss"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1904.08241","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}