Noise2Noise

This is an unofficial and partial Keras implementation of "Noise2Noise: Learning Image Restoration without Clean Data" [1].

There are several things different from the original paper (but not a fatal problem to confirm the noise2noise training framework):

• Training dataset (orignal: ImageNet, this repository: [2])
• Model (original: RED30 [3], this repository: SRResNet [4])

Dependencies

• Keras, TensorFlow, NumPy, OpenCV

Train Noise2Noise

Download Dataset

mkdir dataset
cd dataset
wget https://cv.snu.ac.kr/research/VDSR/train_data.zip
wget https://cv.snu.ac.kr/research/VDSR/test_data.zip
unzip train_data.zip
unzip test_data.zip
cd ..

Train Model

python3 train.py --image_dir dataset/291 --test_dir dataset/Set5

TODOs

• Compare (noise, clean) training and (noise, noise) training
• Add different noise models
• Write readme

References

[1] J. Lehtinen, J. Munkberg, J. Hasselgren, S. Laine, T. Karras, M. Aittala, T. Aila, "Noise2Noise: Learning Image Restoration without Clean Data," in Proc. of ICML, 2018.

[2] J. Kim, J. K. Lee, and K. M. Lee, "Accurate Image Super-Resolution Using Very Deep Convolutional Networks," in Proc. of CVPR, 2016.

[3] X.-J. Mao, C. Shen, and Y.-B. Yang, "Image Restoration Using Convolutional Auto-Encoders with Symmetric Skip Connections," in Proc. of NIPS, 2016.

[4] C. Ledig, et al., "Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network," in Proc. of CVPR, 2017.