Face recognition based on improved VGGNET convolutional neural network
Yang Zhiqi
Abstract
Currently, many famous convolutional neural networks depend on massive training samples and high-performance computers. To solve this problem, this paper improves the deep learning algorithm VGGNet for image classification and proposes a face recognition method called MicroFace. MicroFace uses CASIA WebFace database as the testing and training samples. Our research shows that compared with the original algorithm, the improved algorithm reduces the parameters, improves the pooling function, and increases the number of convolution kernels, which not only decreases the dependence on massive training samples and high-performance computers, but also achieves 96% recognition rate with good recognition performance and certain practicability.