A GAN to generate images of digits based on MNIST Dataset
Generative Adversarial Network is a framework for estimating generative models using an adversarial process.
The following diagram provides a hight level perspective of the work-flow of a GAN.
The loss function we are trying to optimize:
You can find the output generated after each epoch in the output folder.
I have also accumulated them and created a video named Output_GAN.mp4.
Output after 300 epochs :-
