Sparse Semantic Map-Based Monocular Localization in Traffic Scenes Using Learned 2D-3D Point-Line Correspondences | IEEE Journals & Magazine | IEEE Xplore