anishreddy3 / Fashion-MNIST-GAN-Keras

Deep convolutional generative adversarial networks on Fashion MNIST dataset

Deep Convolutional Generative Adversarial Networks

Trained on Fashion_MNIST dataset using Keras Sequential API and Tensorflow gradient tape training loop

Generative Adversarial Networks have two components:

A Generator and Discriminator which are trained simultaneosuly using adversarial crafting process.

A Generator network : It behaves as an artist trying to generate images without any knowledge about the true and learns by interacting with the Discriminator.

A Discriminator network : It behaves as an art critic trying to determine whether a given image is real or fake. It uses the output of generator as training data

Generator images Animation :

The training for generator began by giving input as some random noise and after 60 epochs it starts to generate images that resemble images from the Fashion MNIST dataset.