Theres no reason it needs a DL model. There's a lot of software that calculates tie points and creates point clouds from pictures, which is what they are almost certainly going to do here. DL to go from orthoimages to point cloud, if it is a thing, is probably still in the feasibility steps.