COLORGAN

This is codes for Unsupervised Diverse Colorization via Generative Adversarial Networks

Requirement

tensorflow: 1.0+
numpy
scipy
Pillow
cv2

Usage

Prepare data

Download LSUN/bedroom_train data.
Maximum center crop the data and reshape into 64*64,
put reshaped data into data/lsun_64/bedroom_train/

Train

run train_lsun_wgan.sh for shortcut
or adjust settings by:

python main_wgan.py --is_train=True --some_paramter=some_value

Test

run test_lsun_wgan.sh for shortcut
or adjust settings by:

python main_wgan.py --is_train=False --some_paramter=some_value

Demo

Some demo results in demo/gan/