First, I used this network to train the kitti dataset, but I didn't get the results from the author's paper. What parameters are not adjusted?
Second, use kitti_test.py to get a picture of how the picture looks black and white, but it is indeed a three-channel picture.
Thanks very much for your reply!