Reference-Based Sketch Image Colorization using Augmented-Self Reference and Dense Semantic Correspondence[Lee+, CVPR20]

https://openaccess.thecvf.com/content_CVPR_2020/papers/Lee_Reference-Based_Sketch_Image_Colorization_Using_Augmented-Self_Reference_and_Dense_Semantic_CVPR_2020_paper.pdf

Note that this is an ongoing re-implementation and I cannot fully reproduce the results. Suggestions and PRs are welcome!

Requirements

• Python 3.6+
• PyTorch 0.4+

Usage

1. Download a Tag2Pix dataset from the officical repsitory..
2. Put it on ./datasets/tag2pix
3. Run bash scripts/train_tag2pix_xdog.sh baseline. The training using sketches by XDoG will run.
4. Run bash scripts/train_tag2pix_keras.sh baseline. The training using sketches by SketchKeras will run.

LICENCE

All code is licensed under the MIT license.

RELATED WORKS

• Re-implementation: https://github.com/SerialLain3170/Colorization/tree/master/scft
• https://github.com/MarkMoHR/Awesome-Image-Colorization

Acknowledgements

This repository is based on https://github.com/yunjey/stargan.

Additionally, if you use this repository, please cite original paper

@InProceedings{lee2020referencebased,
    title={Reference-Based Sketch Image Colorization using Augmented-Self Reference and Dense Semantic Correspondence},
    author={Junsoo Lee and Eungyeup Kim and Yunsung Lee and Dongjun Kim and Jaehyuk Chang and Jaegul Choo},
    year={2020},
    booktitle = {Proc. IEEE Computer Vision and Pattern Recognition (CVPR)}
}