Preparation of training samples for U-Net

Hello. In the U-Net based segmentation example, we saw the examples of the pre-segmented training images as outputs. How does this marking process happen in the real world? Who marks the human, cars and trees with different colors? Or is it done automatically? If yes, then how it can be done before the training? If no, then how such a time-consuming routine is handled?

Thanks in advance,

Hi jhasanov,

You can find an interesting paper on this topic here.