How to set up annotation cvs file when each label image contains multiple categories?

I am very new to Pytorch. I am trying to set up a custom dataset where I want to use Unet and train it for semantic segmentation. I have 41 images and 41 label images where each label image may contain 5 different categories. The label images are png images with a specific grayscale for background, category 1, category 2 etc. How should I set up my annotation csv file?

Thanks for any help.