Integration of ontology reasoning-based monocular cues in deep learning modeling for single image depth estimation in urban driving scenarios