This API is an implementation of the CyclicGAN been presented in the original paper.
CylicGAN is a special type of GAN network which works with Unpaired Data to extract mapping between the
two domains (X and Y). We generally have two Generator networks to generate images and two Discriminator networks to criticize
the work of the Generators (declare the images are fake or real). Apart from the Adversarial loss, we also have
cyclic consistency loss which makes sure that any mapping of the image from the domain X to domain Y has
the same reverse mapping.
This API aims to makes it easy for developers to train and infer from a CyclicGAN model.
Before installing the package, install the dependencies specified in requirements.txt file.
This package is also available on Pypi and thus one can directly pip install it.
$ pip install cyclicganIf you wanna contribute or build the package from repository, you can build the package as follows:
- Clone the repo.
- Run
python setup.py install - API has been installed as a python package and now you can directly use its functionalities.
You can follow this tutorial notebook to get an idea of the functionality of this API with a dataset.