Transform class for pairwise data augmentation, flip, random crop and random rotation, ColorJitter (possibly useful for denoising, e.g., dehazing, deraining, etc.), you can add your transformation also.
This code is very useful if you work with pair of images as input ou output, you can make the same data augmentation on both images, if you have three images you can easily extend the code for that situation.
Code was made for: deep learning using PyTorch
You need: PyTorch and PIL