GAN Final Project for CS152

Repository explanation

The code implement GAN and WGAN and tested them on the MNIST dataset. We also computed the FID and SSIM scores on the generated images.

File structure

GAN_params - GAN parameters (dis-params-n is the paramters for the discriminator after n epochs and gen-params-n is the paramters for the generator after n epochs).

plots - Saved plots of loss and discrimiantor output.

saved_loss - Saved losses as NumPy arrays.

WGAN_params - GAN parameters (dis-params-n is the paramters for the discriminator after n epochs and gen-params-n is the paramters for the generator after n epochs).

evaluating_GAN_generator.ipynb - Evaluate GAN using FID and SSIM scores.

evaluating_WGAN_generator.ipynb - Evaluate WGAN using FID and SSIM scores.

GAN.py - Python script to train GAN.

nn_helper.py - Helper file that store the classes for the generator and discriminator and loss functions.

plotting.ipynb - Python notebook for plotting the losses, discriminator outputs, and generated images.

WGAN.py - Python script to train WGAN.