The recently acquired Matrix Mill team at Niantic has spent years building and perfecting deep neural networks that can infer information about the surrounding world from one or more cameras. This technology redefines how machines see and understand the 3D world and more importantly, how digital objects can interact with the real elements of it. Using computer vision and deep learning we are able to develop techniques to understand 3D space enabling much more realistic Augmented Reality (AR) interactions than are currently possible.
In the above AR experience of Codename: Niantic Occlusion, you can see Pikachu and Eevee weaving through and around different objects in the real world, dodging feet and hiding behind planters. This level of integration into the environment around us is a proof-of-concept that excites us about the future of AR.