I am trying image segmentation for aerial images. The problem is that the training images are very big (5000px*5000px), but I want to train the model with smaller (1000px*1000px) images.
To be precise, I want my training set to have 25 images per given original train image.
My original training images look like this one below:
I want to have my training dataset to have 25 images per original image as shown below:
One of the methods is to manually crop all the images in the train dataset. I want to know if it is possible to do this using transforms in fastai.