Learning Occlusion-Aware Dense Correspondences for Multi-Modal Images | IEEE Conference Publication | IEEE Xplore