This is an extension of Colorful Image Colorization by Zhang et al.. It is the final project for the Gothenburg University Machine Learning Course.
We attempt to create spatially coherent colorizations that do not compromise on color saturation.
Feel free to read the report if you would like to learn more about the ideas behind the implementation.
To run it on the university project server, first clone the project.
Then download the pretrained model.h5
from here.
Place it in a new folder called model
within the project.
Now you should be able to just run these commands to train and test:
python3.6 -m src.remote_runner
python3.6 -m src.remote_validator
You can find more information about these results in the report.