LVAR-CZSL: Learning Visual Attributes Representation for Compositional Zero-Shot Learning | IEEE Journals & Magazine | IEEE Xplore