Multi-class Segmentation in UNet

Hi, I am new to PyTorch. I need to train Unet model in PyTorch. The dataset has 3 classes, let’s call them A,B,C. The dataset directory consists of an images folder and three other folders which contain masks for class A,B and C. How do I load these masks for training? Should I use separate dataloaders and train the model for each class one by one or is there any way to do it simultaneously

Shouldn’t there be one mask for every image instead of 3? Each output mask will contain the necessary classes, right?