This project is my first hands-on with Generative Adversarial Networks (GANs). The goal was to generate really realistic MNIST digit images starting from random noise.
-
Keras
-
Custom training loops
-
Matplotlib (for visualization)
-
CNNs
The generator creates digit-like images from random noise, while the discriminator tries to distinguish between real and fake MNIST digits. Over time, both networks improve and the generated images start looking surprisingly realistic.