Image Transformation Based on Generative Adversarial Networks

Abstract

Abstract Image transformation is a hot topic in the field of computer vision. Its purpose is to use the existing similar or semblable images to learn the mapping between the input image and the output image. Generative Adversarial Networks model is a powerful generative model with the idea of Zero-sum game theory, Co-training the two through the confrontation learning method of the generator and the discriminator, so as to estimate the potential distribution of data samples and generate new data samples. This paper is based on the generative adversarial networks and the existing network, the image transformation is realized by combining the two network structures of DCGAN and Cycle GAN. Experimental results show that this method not only effectively solves the problem that paired images are not easy to obtain, but also fully demonstrates the superiority of the generated adversarial networks in image transformation.