RILS: Masked Visual Reconstruction in Language Semantic Space | IEEE Conference Publication | IEEE Xplore