Learnable contextual regularization for semantic segmentation of indoor scene images | IEEE Conference Publication | IEEE Xplore