WIP
Data preparation for CV tasks (super-resolution, denoising, others):
- LR generation with openCV-based composable transformation (https://github.com/victorca25/opencv_transforms): (TBD)
- Extraction of realistic image kernels with KernelGAN (http://www.wisdom.weizmann.ac.il/~vision/kernelgan/): modified to work with Tensorflow 2.0 and PyTorch >= 1.3.0 without generating warnings, the resulting kernels are saved as numpy arrays (npy). To obtain the kernels from images in an "images" directory, run: python kgan/train.py --input-dir "../images" --X4 -o "../results"
- Extraction of noise patches from real images (https://openaccess.thecvf.com/content_cvpr_2018/papers/Chen_Image_Blind_Denoising_CVPR_2018_paper.pdf). To extract the patches, edit "noisepatches/noise_patches.py" and modify the "images_path" and "noise_patches_path" variables to set the input images and output directory for the results.