Deep state-space modeling for explainable representation, analysis, and generation of professional human poses