BNUDC: A Two-Branched Deep Neural Network for Restoring Images from Under-Display Cameras

This is the implementation of BNUDC: A Two-Branched Deep Neural Network for Restoring Images from Under-Display Cameras (CVPR, 2022).

Now constructing....

How to run

You can train or test the BNUDC by changing options; exp_name; and data_root in options/base_options.py, and run train.py or test.py.

Dataset

The datasets are available at POLED containing the pre-processed UDC images, TOLED and SYNTH

Pretrained

Three pre-trained weights for these datasets can be downloaded at CHECKPOINTS.

Citation

@InProceedings{Koh_2022_CVPR,
    author    = {Koh, Jaihyun and Lee, Jangho and Yoon, Sungroh},
    title     = {BNUDC: A Two-Branched Deep Neural Network for Restoring Images From Under-Display Cameras},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2022},
    pages     = {1950-1959}
}