Joint 2D-3D temporally consistent semantic segmentation of street scenes | IEEE Conference Publication | IEEE Xplore