Image colorization GAN

It is a conditional GAN that aims to colorize an image given user guidance.

Code

In the notebook Project_Final_Code.ipynb you may find the code for implementing the network as well as the training and testing.

Data

The dataset we have used is STL10. You can easily download the dataset using the DataLoader class MySTL10. Just make sure that the parameter download is set to True

Model

In the folder Resultsyou may find the models trained. The best model is called cgan_stl_10_100ep_60pts.ckpt.