MNIST GAN

A PyTorch implementation of a Generative Adversarial Network (GAN) trained on the MNIST dataset. Have a fun with it here

This project implements a GAN to generate handwritten digits similar to those in the MNIST dataset. The architecture includes:

• A Generator that creates fake images from random noise
• A Discriminator that tries to distinguish between real and fake images

torch>=2.0.0
torchvision>=0.15.0
matplotlib>=3.5.0
tqdm>=4.65.0

1. Clone the repository:

git clone https://github.com/tveshas/mnist-gan.git
cd mnist-gan

2. Install dependencies:

pip install -r requirements.txt

3. Run the training:

python my_gan.py

• my_gan.py: Main implementation file containing GAN architecture and training loop
• my_gan.ipynb: Jupyter notebook version with visualizations and explanations

• Generator: Multi-layer perceptron with batch normalization and ReLU activation
• Discriminator: Multi-layer perceptron with LeakyReLU activation
• Input dimension: 784 (28x28 MNIST images)
• Latent dimension: 64

MIT License

This implementation is based on the original GAN paper by Goodfellow et al.