Human Face Generation with Generative Adversarial

Abstract

Generating adversarial Networks(GANs) are all about constructing something new, like drawing a human face or creating a portrait. It's quite different and unique than other fields of deep learning. But in every way, it contracts human closer to understanding intelligence. In this research project we proposed a GAN model where we normalize human facial images, then create Generator and Discriminator networks and lastly train the networks which generate new faces. The interesting part of this model is, like real life critics here the art critic is Discriminator, who gives effort to individualize between real and fake images. On the other hand, Generator is the artist, whose main focus is to fool the discriminator by drawing new human faces. So, it turns out that this model works more attractive than any other neural nets models. Many researchers have worked on the creation of human faces using generative adversarial networks but no one worked with images of Bangladeshi people. The purpose of doing this research project is to work not only with the images of foreign people but also Bangladeshi people. So that we can create some new type facial structure people images which we can use in different sectors in our country like game development and animation. For developing different games or animation's characters, we need human faces. To create real-looking human face, production artists need to hire. By applying this GANs model this task can be solved easily.