How to prepare and train on medical images?

[yiluzhou]

I wonder if this is suitable for segmentation on grayscale 2D medical images. How to do the data preparation? It looks like I need to prepare the medical dataset using exactly the same file structure as ADE20k dataset?
For training, do I still have to use "--dataset ade20k" argument if I prepare my customized training dataset?
Any other suggestions? Many thanks!