A baseline for semi-supervised learning of efficient semantic segmentation models - 42Papers