Deep colorizer based on unet architecture (Encoder decoder with skip connections) without a GAN. Trained with flickr-8k dataset for 50 epochs and 10k samples from Celeb dataset for 2 epochs.
python colorizer_gan.py --data_path path_to_flickr_8k_data --model_path path_to_save_the_model
unet-color is based on an encoder decoder architecture with skip connections. Skips ensure that high frequency features trickle down to the base layers of network.
1] http://lmb.informatik.uni-freiburg.de/people/ronneber/u-net/ 2] https://github.com/phillipi/pix2pix