Learning Robot Manipulation Skills From Human Demonstration Videos Using Two-Stream 2-D/3-D Residual Networks With Self-Attention | IEEE Journals & Magazine | IEEE Xplore