A novel Deep CNN based LDnet model with the combination of 2D and 3D CNN for Face Liveness Detection

Face anti-spoofing plays a very important role in security applications using face recognition in their system. Numbers of imposters are involved in spoofing methods. To ensure the presence of real human face in the security system, liveness detection is developed. Recently, many research studies have focused on printed/ photo imposter faces. But this work concentrated on finding face anti-spoofing using both photo and video imposters. In this paper, a novel Deep Convolution Neural Network (CNN) based Liveness Detection network (LDnet) using the combination of both 2D CNN and 3D CNN is proposed. As a pre-processing step, HOG (Histogram of Oriented Gradients) face detector is used to extract the face region from the given input images. The extracted face images are fed into the proposed LDnet model with 15 CNN layers to learn the features while training operation. Finally, the learned features are tested for face liveness detection to predict the real and imposter faces correctly. It is found that the performance of the proposed LDnet model using ROSE-YOUTU database gives better accuracy of 99.79% and HTER (Half Total Error Rate) of 0.08% than other existing methods.