Well, you can use one or two cameras and a lidar, and use that to generate data to train a depth-map.