“…Since pose estimation is much better-posed in 2D than in 3D, a popular way to infer joint positions is to use a generative model to find a 3D pose whose projection aligns with the 2D image data. In the past, this usually involved inferring a 3D human pose by optimizing an energy function derived from image information, such as silhouettes [6,14,21,22,25,31,44,49,60], trajectories [74], feature descriptors [58,62,63] and 2D joint locations [2,3,5,20,36,51,57,68,69]. Another class of approaches retrieve the pose from a dictionary of 3D poses based on similarity with the 2D image evidence [18,26,39,41,42].…”