Skip to content

neuronest/custom-cyclegan

Repository files navigation

Custom CycleGAN

Custom CycleGAN is a pure and concise TensorFlow 2 implementation of the Cycle-Consistent Adversarial Networks paper : https://arxiv.org/pdf/1703.10593.pdf.

Prerequisites

A ready to use Conda package manager.
You may be interested by Miniconda and the minimal installation it requires.

Install

git clone git@github.com:neuronest/custom-cyclegan.git
cd custom-cyclegan
source install.sh

Run TensorBoard server

tensorboard --port 6006 --logdir tensorboard

Note: TensorBoard internally restricts to 10 the number of distinct images displayed per tag.
You can work around this limitation by adding the --samples_per_plugin images=n argument, with n a higher number, analogous to the maximum number of epochs.
For example:

tensorboard --port 6006 --logdir tensorboard --samples_per_plugin images=500

Run a training experiment

python -m src.main

About

Pure TensorFlow 2 CycleGAN implementation

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors