IDR: Self-Supervised Image Denoising via Iterative Data Refinement

Yi Zhang¹, Dasong Li¹, Ka Lung Law², Xiaogang Wang¹, Hongwei Qin², Hongsheng Li^1
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).

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

SenseNoise dataset

Downloads Drive | Baidu Netdisk

The released dataset is what we used in our paper. Thanks to the advice from the anonymous reviewers, we are still working on improving the quality of the dataset.

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 [email protected] if you have any questions.

Acknowledgments

• N2N, N2V, bm3d, basicsr