Abstract: Accurate estimation of 3D human motion from monocular video requires modeling ... frame pose estimate and the next image frame to control a physically-simulated character to output the ...