IDR: Self-Supervised Image Denoising via Iterative Data Refinement

Yi Zhang¹, Dasong Li¹, Ka Lung Law², Xiaogang Wang¹, Hongwei Qin², Hongsheng Li¹<br> ¹CUHK-SenseTime Joint Lab, ²SenseTime Research

---

This repository is the official PyTorch implementation of IDR. It also includes some personal implementations of well-known unsupervised image denoising methods (N2N, etc).

Update

• 2023.08.01: We release the sensenoise dataset v4, which contains both Raw images, sRGB images and meta information (e.g. lens shading, ccm).

SenseNoise dataset

V4 Downloads:

• Raw images (OneDrive | Baidu)
• sRGB images (OneDrive | Baidu)
• Meta data (OneDrive | Baidu)

V3 Downloads: OneDrive | Baidu Netdisk

Thanks to the advice from the anonymous reviewers, we are still working on improving the quality of the dataset.

Training

Slurm Training. Find the config name in configs/synthetic_config.py.

sh run_slurm.sh -n config_name

Example of training IDR for Gaussian denoising:
sh run_slurm.sh -n idr-g

Testing

The code has been tested with the following environment:

pytorch == 1.5.0
bm3d == 3.0.7
scipy == 1.4.1 

• Prepare the datasets. (kodak | BSDS300 | BSD68)
• Download the pretrained models and put them into the checkpoint folder.
• Modify the data root path and noise type (gaussian | gaussian_gray | line | binomial | impulse | pattern).

python -u test.py --root your_data_root --ntype gaussian 

Citation

@inproceedings{zhang2021IDR,
      title={IDR: Self-Supervised Image Denoising via Iterative Data Refinement},
      author={Zhang, Yi and Li, Dasong and Law, Ka Lung and Wang, Xiaogang and Qin, Hongwei and Li, Hongsheng},
      booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
      year={2022}
}

Contact

Feel free to contact zhangyi@link.cuhk.edu.hk if you have any questions.

Acknowledgments

• N2N, N2V, bm3d, basicsr