/mnistgan

A simple Keras implementation of a Generative Adversarial Network, used to generate images of handwritten digits

Primary LanguagePythonMIT LicenseMIT

GAN handwritten digits generation

Generation examples

A simple Keras implementation of a Generative Adversarial Network, used to generate images of handwritten digits. Based on the tutorial by Stefan Hosein at https://towardsdatascience.com/demystifying-generative-adversarial-networks-c076d8db8f44

Install

After getting a working Conda distribution, install the environment as

conda env install -f environment.yml

If you do not have an nVidia GPU available, edit the environment.yml to change the keras-gpu package to just keras.

Then log into the environment

source activate mnistgan

Run

Train the network with

python mnistgan.py