Variational Context-Deformable ConvNets for Indoor Scene Parsing | IEEE Conference Publication | IEEE Xplore