Robust Estimation of 3D Human Poses from a Single Image

Human pose estimation is a key step to action recogni-tion. We propose a method of estimating 3D human poses from a single image, which works in conjunction with an existing 2D pose/joint detector. 3D pose estimation is chal-lenging because multiple 3D poses may correspond to the same 2D pose after projection due to the lack of depth in-formation. Moreover, current 2D pose estimators are usu-ally inaccurate which may cause errors in the 3D estima-tion. We address the challenges in three ways: (i) We rep-resent a 3D pose as a linear combination of a sparse set of bases learned from 3D human skeletons. (ii) We enforce limb length constraints to eliminate anthropomorphically implausible skeletons. (iii) We estimate a 3D pose by mini-mizing the L1-norm error between the projection of the 3D pose and the corresponding 2D detection. The L1-norm loss term is robust to inaccurate 2D joint estimations. We use the alternating direction method (ADM) to solve the op-timization problem efficiently. Our approach outperforms the state-of-the-arts on three benchmark datasets. 1