Augments the KITTI with more instance pixel-level annotation for 8 categories.
Source: Amodal Instance Segmentation with KINS Dataset