Skip to content

Probability distribution learning with GANs - DCGAN, CycleGAN

Notifications You must be signed in to change notification settings

remydeshayes/GANs

Repository files navigation

Deep Learning project

DCGAN and cycleGAN

Authors : Rémy DESHAYES, Maxime BERILLON

When studying an unsupervised learning problem, data at hand supposedly follow a distribution that we want to discover and learn.

Our report - please have a look at it - follows 4 steps. First, section 1 introduces two approaches to tackle the problem of probability distribution learning and make a brief introduction to a model - the GAN - that leverages one of these approaches. Then, section 2 builds on the remarks made in the previous part and introduces the Wasserstein GAN. Section 3 and section 4 propose an in-depth review of two GANs and their implementations, namely the DCGAN and the CycleGAN.

Implementations can be found here or on Google Colab Open In Colab

alt text