Deep Face Recognition based on an Optimized Deep Neural Network using ZFNet

Said Si Kaddour, Larbi Boubchir, Boubaker Daachi

Abstract: Facial recognition for security concepts allows the authentication and identification of persons by comparing their facial traits. Neural networks, in particular Convolutional neural networks (CNNs), have been widely and successfully used for image recognition. This paper provides a comparative study of CNN models applied for face recognition with different facial expressions and lighting conditions. It also presents a hybrid approach based on ZFNet architecture based on the optimization of hyperparameters to improve face recognition performance. Experiments conducted on the Yale Face database have shown that the proposed optimized ZFNet architecture allows achieving a high accuracy, and outperforms other well-known CNN models such as VGGNet and AlexNet architectures.