Abstract: Estimating three-dimensional (3-D) human poses from a given two-dimensional (2-D) shape is still an inherently ill-posed problem in computer vision. This paper proposes a method called ...