ugis22 / creatingDCGAN

Show how DCGAN works. Contains some code about the generator, discriminator and training step.

******** 🚧🚧 UNDER CONSTRUCTION 🚧🚧 ************

Deep Convolutional Generative adversarial network

Introduction

Generative Adversarial Networks (GAN) has been described in 2014 by Ian Goodfellow and other researchers belonging to the University of Montreal. The most interesting point about GAN is that they can learn how to mimic the distribution of certain data and through that, generate images, music, speechs, texts, and much more. Briefly, GAN is composed of two neural networks: a convolutional network called "The Generator" which will generate images from a noise vector and another flipped convolutional network called "The discriminator" that determines if the image came from real data or was generated.

Objective

The aim of this project is to build a deep convolutional GAN that takes as an input images and generates similar new images.

How GAN works

• Generator
• Discriminator

******** 🚧🚧 UNDER CONSTRUCTION 🚧🚧 ************