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