Moreover, under various LiDAR-camera calibration errors, the proposed method reduced the depth estimation MAE to 0. 34-0. 93 times from previous depth completion methods.
Our experiments show that our method can outperform previous unsupervised and semi-supervised depth completion methods in terms of accuracy.
Low level features like edges and textures play an important role in accurately localizing instances in neural networks.
Weakly-supervised instance segmentation, which could greatly save labor and time cost of pixel mask annotation, has attracted increasing attention in recent years.