Face Anti-spoofing Based on Client Identity Information and Depth Map.

Face anti-spoofing (FAS) is an essential prerequisite for face recognition. In most methods, FAS is usually performed before face recognition and the client identity information is not utilized. Since presentation attacks (PAs) are always aimed at a certain client, the client identity information can provide useful clues for FAS task. In this paper, we propose a face anti-spoofing method based on client identity information using Siamese network. We applied FAS after face recognition to utilize the client identity information. As the real face and fake face have different properties, we use different weights for the two subnetworks of the Siamese network to extract features for real face and fake face, respectively. In addition, we employ depth map as auxiliary information to improve the performance. We perform experiments on SiW, CASIA-FASD and Replay-Attack datasets to demonstrate the validity of our method.