If I have two separate datasets declared with separate sets of torch.transforms, and then I concatenate the datasets using torch.utils.data.ConcatDataset to create the final dataset, will the different sets of transforms be preserved when the dataloader calls the images, or will all the images follow only one/common set of torch.transforms?
I have an original dataset, and I am trying to create an augmented dataset using some torch.transforms, and then feed both of them to the network as one. I expect the original images to go through their own transforms, and also separately go through the augmentation transforms to form the augmented dataset, and ultimately unite them to make the final dataset.
