Is image as patches and train with crop size same in Neural network training?

Question

I am using semantic segmentation model U-net for training a classification task for a project. My image size is 1536*1024. Since the GPU capability cannot handl