Awesome

Learning sRGB-to-Raw-RGB De-rendering with Content-Aware Metadata @CVPR'22

[Paper] [arXiv] [Supplemental]

Usage

• The code runs with Python 3.7 and PyTorch 1.8.
• Install python packages: torch, torchvision, numpy, scikit-image, opencv-python
• Download the dataset, which is composed of Samsung, Sony, and Olympus cameras.
• Fix the dataset path in each .sh file in ./scripts.
• Run train_*.sh for training, and test_*.sh for inference.
• (Optional) download pre-trained models, and fix the model path in test_*.sh files in ./scripts.

Citation

Please cite our paper when you use this code.

@InProceedings{Nam_2022_CVPR,
    author    = {Nam, Seonghyeon and Punnappurath, Abhijith and Brubaker, Marcus A. and Brown, Michael S.},
    title     = {Learning sRGB-to-Raw-RGB De-Rendering With Content-Aware Metadata},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2022},
    pages     = {17704-17713}
}

Contact

Please contact snam0331 AT gmail.com if you have any question about this work.