/hooligan

Generative Adversarial Network for MNIST / CIFAR-10 dataset.

Primary LanguagePython

HooliGAN

Generative Adversarial Network for MNIST / CIFAR-10 dataset.

Download repository to your local computer, train the model by typing

python3 gan.py

... and enjoy the beautiful images that your generator produces.

If you want to tune hyperparameters, open and modify config.py file.

You can also do it by providing flags with the same names as parameters, e.g.

python3 gan.py --batch_size=256 --epochs=10 --gen_lr=0.0003

Here are samples taken after training for a few hours:

Contact

If you want to contribute / give a feedback / say hello, don't hesitate and mail me at:

gmail: tomwesolowski92