Gantools is a set of classes written in Tensorflow to work with Generative Adversarial Netork (GAN). Their exist many other framework to do this. The reason I coded this one is to have a full control over every parts of the network.
Warning: the code is insuficiently documented, not entirely tested and may contains some bugs.
In the repository, ... , you will find a demonstration notebook.
Since this is intented as a package for developpement, we highly suggest to install it with a simlink.
git clone git@github.com:nperraud/gantools.git
cd gantools
pip install -r requirements.txt
pip install -e .
The code is composed of a package named gantools. It is composed of the following submodules:
- gansystem: implement the basic training and generating system for a gan
- modely: contains the different network architecture
- data: data module
- blocks: basic tensorflow units
- utils: useful functions
- metrics: computation of the different error functions
- plot: helper for the different plots