摘要
文章研究提出基于改进的MobileNet开展人脸反欺诈的方法,能够有效区分真实人脸和虚假人脸,发现人脸识别中的展示攻击现象。在人脸反欺诈公开数据集中验证可知,在本研究所实现的深度学习模型其所需要的判断时间不显著增加的情况下,指标明显提升,更加适合部署在硬件性能有限的移动端设备中。
This paper proposes a face anti-spoofing method based on the improved MobileNet,which can effectively distinguish real faces from fake faces and discover the display attack in face recognition.Validation in the public datasets of face anti-spoofing shows that the deep learning model proposed in this study requires little significant increase in judgment time,and the metrics are significantly improved,making it more suitable for deployment in mobile devices with limited hardware performance.
作者
商家衡
郝久月
SHANG Jia-heng;HAO Jiu-yue(People's Public Security University of China,Beijing 100038,China;First Research Institute of The Ministry of Public Security of PRC,Beijing 100048,China)
出处
《电脑与信息技术》
2023年第2期59-62,共4页
Computer and Information Technology