Warp-Refine Propagation: Semi-Supervised Auto-labeling via Cycle-consistency
    Aditya Ganeshan1 * Alexis Vallet2 Yasunori Kudo2 Shin-ichi Maeda2 Tommi Kerola2
              Rares, Ambrus, 3 Dennis Park3 Adrien Gaidon3
      1
       Brown University 2 Preferred Networks, Inc. 3 Toyota Research Institute (TRI)

               Abstract
  Deep learning models for semantic segmentation rely on
expensive, large-scale, manually annotated datasets. La-
belling is a tedious process that can take hours ...