[Machine Learning] nerf and adop can dramatically improve the way humans make 3d models from pictures and video