Face Generation from Textual Description

Abstract

The main object of this project is to generate a human face based on a given description. Generating images from text plays a huge role for public safety. This is a sub domain of text-to-image synthesis. This project is a part of deep learning. This project uses two latest architectures for generating images from text. StackGAN and ProGAN for synthesis of the image from the given description. We are using a dataset named ‘Face2Text’, which contains 400 facial images and textual captions for each of the image. The key idea is to grow both the generator and discriminator progressively: starting from a low resolution, we add new layers that model increasingly fine details as training progresses. The final generated output is not that high quality but comparable with the given input.