A neural network can turn a selection of photographs of a scene into a 3D world that can then be viewed from any angle, and could eventually be used in the movie industry