How do you map multiple input images to a single mask (segmented image/ground truth) in tensorflow/keras?

Question

I know how to map single input image to single masked image in keras. You do that by having specific directory structure and in that each input image gets mapped to correspondin