CAHIR: Co-Attentive Hierarchical Image Representations for Visual Place Recognition | IEEE Conference Publication | IEEE Xplore