C2N: Practical Generative Noise Modeling for Real-World Denoising - Official PyTorch release

This is an official PyTorch release of the paper "C2N: Practical Generative Noise Modeling for Real-World Denoising" from ICCV 2021.

If you find C2N useful in your research, please cite our work as follows:

@InProceedings{Jang_2021_ICCV,
    author    = {Jang, Geonwoon and Lee, Wooseok and Son, Sanghyun and Lee, Kyoung Mu},
    title     = {C2N: Practical Generative Noise Modeling for Real-World Denoising},
    booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
    month     = {October},
    year      = {2021},
    pages     = {2350-2359}
}

[PDF] [Supp] [arXiv]

Setup

Dependencies

Python 3.9.6
numpy >= 1.16.4
cudatoolkit >= 10. (if using GPU)
PyTorch 1.2.0
opencv-python
scikit-image >= 0.15.0
tqdm
pillow
pyyamml
imutils

Data

You can place any custom images in ./data and image datasets in subdirectory ./data/[name_of_dataset]

For the SIDD and DND benchmark images, you can find them at SIDD Benchmark and DND Benchmark. Convert them into .png images and place them in each subdirectory.

Pre-trained Models

Download following pre-trained models:

Generator	Clean	Noisy	config	Pre-trained
C2N	SIDD	SIDD	C2N_DnCNN	model
C2N	SIDD	DND	C2N_DnCNN	model

Denoiser	Generator	Clean	Noisy	Clean (denoiser train)	config	Pre-trained
DnCNN	C2N	SIDD	SIDD	SIDD	C2N_DnCNN	model
DIDN	C2N	SIDD	SIDD	SIDD	C2N_DIDN	model
DIDN	C2N	SIDD	DND	SIDD	C2N_DIDN	model

Demo (Quick start)

test_generate.py:

config: Name of the configuration.
ckpt: Name of the checkpoint to load. Choose between 'C2N-SIDD_to_SIDD' and 'C2N-DND_to_SIDD' depending on the noisy images it is trained on.
mode: 'single' or 'dataset'.
data: Filename of clean image if mode is 'single', dataset of clean images if mode is 'dataset'.
gpu: GPU id. Currently this demo only supports single-GPU or CPU device.

Examples:

# Generate on single clean image
python test_generate.py --ckpt C2N-SIDD_to_SIDD.ckpt --mode single --data clean_ex1.png --gpu 0
python test_generate.py --ckpt C2N-DND_to_SIDD.ckpt --mode single --data clean_ex2.png --gpu 0

# Generate on clean images in a dataset
python test_generate.py --ckpt C2N-SIDD_to_SIDD.ckpt --mode dataset --data SIDD_clean_examples --gpu 0
python test_generate.py --ckpt C2N-DND_to_SIDD.ckpt --mode dataset --data SIDD_clean_examples --gpu 0

test_denoise.py:

config: Name of the configuration. Choose between 'C2N_DnCNN' and 'C2N_DIDN' depending on the denoiser to be used.
ckpt: Name of the checkpoint to load.
- Name format: '[denoiser model]-[C2N train noisy set]to[C2N train clean set]-on_[denoiser train set]'
  - denoiser model: 'DnCNN' or 'DIDN'.
  - C2N train noisy set: Dataset that the noisy images for C2N training are sampled from.
  - C2N train clean set: Dataset that the clean images for C2N training are sampled from.
  - denoiser train set: Dataset that the clean->noisy images are generated from, to train the denoiser.
mode: 'single' or 'dataset'.
data: Filename of noisy/generated image if mode is 'single', dataset of noisy/generated images if mode is 'dataset'.
gpu: GPU id. Currently this demo only supports single-GPU or CPU device.