Abstract: Establishing 3-D perception capabilities for self-driving cars is a key research problem. Recent research has differentially “lifted” features from multicamera images onto a 2-D ground plane ...