This GitHub repository contains the PyTorch implementation of PaletteNet: Image Recolorization with Given Color Palette (Cho et al., 2017).
I used a Jupyter Notebook to demonstrate the implementation for the following two reasons:
- The authors did not provide the dataset for their experiments, so my team crawled images and their palettes from Dribbble. My intention here is not to create a data pipeline here; instead, I want to showcase the models quickly.
- There are many helper functions needed for successful training of the models, such as extracting the luminance of an image. I believe putting all the helper functions in one place with the model training process (instead of compartmentalizing the utility functions) makes reading my codes simpler.
I have commented on the codes inside the notebook, so it's easy to follow my implementation.