Cross-modal body representation based on visual attention by saliency | IEEE Conference Publication | IEEE Xplore