Learning Cross-Modal Deep Representations for Robust Pedestrian Detection | IEEE Conference Publication | IEEE Xplore