Tensorflow implementation for paper "A learned representation of artist specific colourisation"
Given a greyscale image we would like to produce an as realistic as possible colour image, in the style of the artist. Some examples using Conditional Instance Normalisation:
| Original | Greyscale input | Colourisation |
|---|---|---|
 |
 |
 |
 |
 |
 |
Dependencies
Dataset Painters by numbers
- Run data/make_tf_records.py on image data to create tfrecords
- Run count.py to gather statistics for re-weighting of colours
- Run train.py
Run eval.py with correct --source