This is an implementation of a Relativistic Average GAN, introduced in "The relativistic discriminator: a key element missing from standard GAN". The original Pytorch code is found here.
- Download the (Fashion)-MNIST *.csv dataset from kaggle, and place in
./data/fashion-mnist_train.csv
. - Create the Logdir:
$ mkdir ./logs/
. - Run
$ python ragan.py
to train using the default hyperparameters, or run$ python ragan.py -h
for information about the tweakable parameters.
Below is the Fashion-MNIST (DC)RaGAN loss at 100 epochs, along with some example outputs.