Abstract: We propose a multi-view weakly-supervised 3D human pose estimation network. In this network, we fuse multi-view features and generate the final 3D pose, supervising the 3D pose using human ...