Modernizing Old Photos Using Multiple References via Photorealistic Style Transfer

We present a Multi-Reference-based Old Photo Modernization (MROPM) framework that can modernize old photos using multiple references by performing stylization and enhancement in a unified manner. MROPM initially stylizes old photos using multiple references through photorealistic style transfer. Then, it further enhances the results to produce modern-looking images. Compared to the previous state-of-the-art (SOTA), the proposed method can perform modernization on real old photos even without using old photos during training. It depends on the selection of references, which need to have similar semantics to the old photo.

Modernizing Old Photos Using Multiple References via Photorealistic Style Transfer
Agus Gunawan, Soo Ye Kim, Hyeonjun Sim, Jae-Ho Lee, Munchurl Kim
CVPR 2023

Model Sources

Repository: [https://github.com/KAIST-VICLab/old-photo-modernization]
Paper: [https://arxiv.org/pdf/2304.04461]
Project Page: [https://kaist-viclab.github.io/old-photo-modernization/]

Model Description

Our framework named as Multi-Reference-based Old Photo Modernization (MROPM) consists of two key ideas: 1) MROPM-Net to modernize old color photos by changing their styles and enhancing them to look modern and 2) synthetic data generation scheme to train the network in a self-supervised manner

Please check our paper and project page for more details.

Citation

If you find our work inspiring or use our codebase in your research, consider giving a star ⭐ and a citation.

@inproceedings{gunawan2023modernizing,
  title={Modernizing Old Photos Using Multiple References via Photorealistic Style Transfer},
  author={Gunawan, Agus and Kim, Soo Ye and Sim, Hyeonjun and Lee, Jae-Ho and Kim, Munchurl},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={12460--12469},
  year={2023}
}