The research of virtual face based on Deep Convolutional Generative Adversarial Networks using TensorFlow

摘要

Since Generative Adversarial Nets (GANs) has been proposed in 2014, it has become one of the most popular hot topics. Deep Convolutional Generative Adversarial Networks (DCGAN) is greatly promoted the development and application of GANs. In this paper, we have made an in-depth exploration for the most popular DCGAN at present via utilizing TensorFlow deep learning framework, using the open CelebA face dataset of The Chinese University of Hong Kong as the data source. By comparing DCGAN unconstrained and DCGAN constrained, the experimental results show that the DCGAN model significantly improves the virtual face generation model after adding constraints in the training phase, which enhance the ability of the generator to deceive the discriminator. Finally, we have evaluated the proposed model from the perspective of TensorBoard and achieved the desired experimental results. (C) 2019 Elsevier B.V. All rights reserved.