Image Synthesis Using Generative Adversarial Network

Abstract

Generative Adversarial Networks (GANs) are a deep learning based generative model. GANs are a model for training a generative model and it is common to use deep learning models. Generative Adversarial Network(GANs) are a powerful class of neural networks that are used for unsupervised learning. GANs are basically made up of two competing neural network models which compete with each other and are able to analyze, capture and copy the variations within dataset. GANs achieve high level realism by pairing a generator which learns to produce a target output with a discriminator which learns to distinguish true data from the output of the generator. GANs used for Image Synthesis generates high resolution images. Text to face generation is a sub domain of text to image synthesis, and it has a huge impact along with the wide range of applications on public safety domain. Our proposed fully trained GAN outperformed by generating the good quality images with accordance to the input sentence. StackGAN aims at generating high resolution photorealistic images. The Stage-I GAN sketches the primitive shape and colors of a scene based on a given text description, yielding low-resolution images. The Stage-II GAN takes Stage-I results and the text description as inputs, and generates high-resolution images with photo-realistic details.