The idea of semantic segmentation is to recognize and understand what is in an image at the pixel-level.