This project implemented a Generative Adversarial Network (GAN) with MNIST data set. This project was completed in Python 3 using Keras.
- Fully-Connected, Feed-Forward Genreator and Discriminator - Showing Very Good results below
- Convolutional Genreator and Discriminator - Untested, needs additional work
The Feed-Forward model setup showed good results for generating MNIST digits. It is clear there is still some mode collapse, however I amd satisfied with the results of the Genreator.
The image below shows a sample of 100 random digits generated.
