How to create one hot representation of an Image

vision
1

How to use 3 channel and a N channel one hot vector to U-Net model.

2

Can you provide some kind of an example of what you are trying to do?