How to create an image mask for convolutional neural networks?

Question

Neural networks

Hello dear experts, I apologize in advance for a possibly stupid question :)
The fact is that I read one article where a fully convolutional neural network was used to determine the check area in the image. It used a mask of the following type:

where white is the desired area.
The question is, is it possible to make several masks for different parts and how to implement it? Or can we simply use another value from 0 to 255 for this, different from the existing ones? Thanks in advance.

Answer 1

For different classes (for example, background, paper, some objects), you need to use different channels, i.e. the target image must have a width, height, n_classes shape.