VT-Grapher: Video Tube Graph Network with Self-Distillation for Human Action Recognition | IEEE Journals & Magazine | IEEE Xplore