Modifiy image prossing part only in U-net Model and adding cross valadation

Hi I have U-net model working fine in Keras and tensor flow, I need someone to modify the way I read the images.
the one I have now is reading images and mask from different directory (image folder and mask folder), my new images are in same directory saved as numpy fils and image and it is mask saved in same file with different channels, more details will provide in chat