This repo is the pytorch implementation of Colorful Image Colorization. In ECCV, 2016.

Version info: pytorch 0.3.0, python 3.5

Installation

git clone https://github.com/chuchienshu/Colorization.git

Training and validation

training from scratch:

python main.py 

or resume training from specified time point:

python main.py -R ***

Both above commands default to support cross validation, which can be switch down by adding option -V False. More detailed configure setting ,please refer to config.py

Inference

python main.py -T False -I ***  # *** means the inference step. 

Results on DAVIS 2017

training loss curve

NOTE: As you can see,I didnot train enough time(since I do not have NVIDIA's support...).And the initial lr is 3.16e-4, which is 3.16e-5 in original.

Acknowledgments

Parts of this code were derived, as noted in the code, from richzhang

Reference

@inproceedings{zhang2016colorful,
  title={Colorful Image Colorization},
  author={Zhang, Richard and Isola, Phillip and Efros, Alexei A},
  booktitle={ECCV},
  year={2016}
}