Feature Extraction of Dual-convolutional Network with LBP for Face Anti-Spoofing

Abstract

With the promotion of information technology, network security issues have received more and more attention. As the most unique and widely used biological feature, human face is the primary verification method that people choose. But it has also become the preferred target of criminals. Based on the research and analysis of algorithms that have performed well in recent years, this paper proposes a dual-convolution multi-scale feature extraction network that combines central differential convolution and depth separable convolution. At the same time, we also added traditional manual feature LBP to Improve the robustness of the network. The proposed network mainly for high -definition photo and video replay attacks.