By Shuangfan Zhou, Chu Zhou, Youwei Lyu, Heng Guo, Zhanyu Ma, Boxin Shi, Imari Sato
- python=3.8.18
- pytorch=2.0.1
- cv2
- numpy
- tqdm
- scikit-image
We publicly release our dataset, which contains a total of 138 images.
data/
├── train/
│ ├── 001/
│ │ ├── RGB_0.png
│ │ ├── RGB_45.png
│ │ ├── RGB_90.png
│ │ └── RGB_135.png
│ ├── 002/
│ │ ├── RGB_0.png
│ │ ├── RGB_45.png
│ │ ├── RGB_90.png
│ │ └── RGB_135.png
│ └── .../
└── test/
├── 001/
│ ├── RGB_0.png
│ ├── RGB_45.png
│ ├── RGB_90.png
│ └── RGB_135.png
├── 002/
│ ├── RGB_0.png
│ ├── RGB_45.png
│ ├── RGB_90.png
│ └── RGB_135.png
└── .../
We provide the pre-trained models for inference
- You can run inference on the example raw images using the following command:
python infer.pythe results will be saved in the infer/ directory
- or you can use your own dataset by specifying the
--infer_pathargument:
python infer.py --infer_path /path/to/your/datapython train.pyIf you find this work helpful to your research, please cite:
@inproceedings{pidsr2025zhou,
title = {PIDSR: Complementary Polarized Image Demosaicing and Super-Resolution},
author = {Shuangfan, Zhou and Chu, Zhou and Youwei, Lyu and Heng, Guo and Zhanyu, Ma and Boxin, Shi and Imari, Sato},
year = {2025},
booktitle = CVPR,
}
This code is built on Restormer. We thank the authors for sharing their codes.