Niantic tests object recognition for Pokémon Go engine

Niantic has acquired British machine vision company Matrix Mill. The goal is to give Niantics Real World Platform, the engine behind Pokémon Go, a better understanding of objects in the physical world.

The Matrix Mill team comes from University College London and specializes in developing deep neural networks that can derive information from scans of environments with a camera. Niantic says it focuses strongly on computer vision, machine learning and developing deep neural networks. The company wants to further develop its Real World Platform and allow these objects to be recognized. This should provide more natural images of the ar platform and digital objects should also be able to interact with the physical world.

For example, if the Real World Platform software understands that the camera is pointing at a bunch of flowers, it can decide itself to make a bee appear, or a duck by a pond. Niantic also shows a proof-of-concept where a virtual Pikachu runs behind physical planters, a technique called occlusion. The company is also working on lowering latency to enable ar multiplayer.

Niantic opens its engine to a small group of developers so that third parties can experiment with it as well. Earlier this year, the Pokemon Go developer acquired Escher Reality, which develops services for cross-platform mobile ar games.