Deep Cross-Modal Representation Learning and Distillation for Illumination-Invariant Pedestrian Detection | IEEE Journals & Magazine | IEEE Xplore