Pokémon Go Maker Used Billions Of Images To Train An AI Map Of The World

Niantic has been selling map data collected from Pokémon Go players, and one client is using that information to train AI map models for delivery robots. Niantic Spatial, the AI-focused offshoot formed last year after Niantic was acquired by Scopely, revealed it used Pokémon Go’s map data to create a model that can pinpoint a person’s location within centimeters.

The model is intended for companies such as Coco Robotics to improve navigation in areas where GPS is less reliable, like cities with higher rates of signal interference. Niantic Spatial has been building a visual positioning system that uses images and videos instead of relying solely on GPS coordinates.

The system was trained on over 30 billion images captured in urban environments, though players must opt in to contribute, particularly at hotspots like gyms where photos are taken from many angles and at different times of day. “We had a million-plus locations around the world where we can locate you precisely,” said Niantic Spatial CTO Brian McClendon.

niantic, pokémon go, niantic spatial, scopely, coco robotics, visual positioning, map data, gps reliability, 30 billion, location accuracy