Semantic Segmentation of Road Scenes