Exploiting skeletal structure in computer vision annotation with Benders decomposition