The goal of this benchmark is to segment and parse an image into different image regions associated with semantic categories, such as sky, road, person, and bed. It is similar to semantic segmentation ...