A Resnet-based Text to Image Conversion Method Using Word2Vec and Generative Adversarial Networks

Abstract

Image generation from natural language is one of the main applications of recent conditional generation model. In addition to testing our ability to model conditional, high-dimensional distributions, text-to-image synthesis has many exciting and practical applications, such as photo editing or computer-aided content creation. Recent progress has been made in the use of Generated Countermeasure Networks (GAN). Although Generative Adversarial Networks (GANs) have shown remarkable success in various tasks, they still face challenges in generating high quality images. In this paper, we propose a generative adversarial networks (GAN) based text-to-image generating method. In many natural language processing tasks, which word expressions are determined by their term frequency – inverse document frequency scores. Word2Vec is a type of neural network model that, in the case of an unlabeled corpus, we produces a vector that expresses semantics for words in the corpus and an image is generated by GAN training according to the obtained vector. Thanks to the understanding of the word we can generate higher and more realistic images. Our GAN structure is based on deep convolution neural networks and pixel recurrent neural networks. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. Comparing the generated image with the real image, we get about 92% similarity on the Oxford-102 flowers dataset.