Generative Adversarial Networks

Authors: Daniel Vigild Juhász, Frederik Kromann Hansen & Lau Johansson

This repository contains hand-in assignment for the DTU course 02514 Deep learning in computer vision.

Read our article here

Simple GAN (fully connected) - MNIST

This notebook is a implementation of a simple GAN trained on MNIST dataset. Look at the implemented model here

Overview of GAN model

alt text

Fake generated digits

alt text

CNN GAN - MNIST

This notebook is a implementation of a CNN GAN trained on MNIST dataset. Look at the implemented model here

Fake generated digits

alt text

Interpolation between two digits

alt text

CycleGAN - Horse2Zebra

This notebook is a implementation of a CycleGAN trained on images of horses and zebras. Look at the implemented model here

Overview of CycleGAN model:

alt text

Horses2Zebras results

alt text