Abstract: We propose a multi-view weakly-supervised 3D human pose estimation network for depth maps. In this network, we introduce the Sum of Gaussians (SoG) model to represent the human body and ...