This project is a basic Generative Adversarial Network (GAN) implemented in PyTorch on the MNIST Database
This is one of my initial steps towards GANs in general. This mostly follows from the idea of GAN first published in arXiv:1406.2661 [stat.ML] by GoodFellow et.al.
Follow the instructions below to get our project running on your local machine.
- Clone the repository and make sure you have Python 3 to run the project.
- Go to
src/PyTorch/
and runpython gan-mnist-pytorch.py
- All the outputs and related plots can be found in
src/PyTorch/output
folder generated. - The various parameters that can be tweaked before run can be found at
python gan-mnist-pytorch.py --help
- PyTorch 0.4.0 or above
- CUDA 9.1 (or other version corresponding to PyTorch) to utilize any compatible GPU present for faster training
Images generated by Generator at various epochs -