A Lightweight Convolutional Neural Network for Face Recognition

This paper presents a lightweight Convolutional Neural Network (CNN) model for face recognition. The network structure has three convolutional layers, three pooling layers, one flattening layer and two fully connected layers, the pooling layer uses average pooling, the back propagation process uses the Adam model, and the training set used for network training is the FERET face database. The single-layer or two-layer convolutional neural network is widely used for simple image recognition. This structure expands the convolutional layer and the pooling layer to three layers, which makes the network capable of recognizing face images and has a high recognition accuracy rate, which can reach over 98%. Compared with large-scale deep neural networks, this network has a simpler structure, takes less time to train, and is easy to implement in hardware.