From Pokémon GO to physical AI: Niantic Spatial unveils its global 3D mapping platform

Niantic Spatial, the company spun out of Pokémon GO maker Niantic, is launching a revamped version of its Scaniverse platform for businesses and a new global visual positioning system (VPS), in its biggest release since becoming an independent company last year.

The San Francisco-based company, which maintains an engineering team of about 30 people in Bellevue, Wash., developed key parts of the release at the Seattle-area office, including the latest version of its developer toolkit. Niantic previously operated a significant Bellevue engineering office focused on Pokémon GO and its core AR platform.

Scaniverse lets users capture physical spaces with a smartphone or 360-degree camera and generate detailed 3D models. The VPS, trained in part on billions of images crowdsourced from Pokémon GO players, pinpoints a device’s location and orientation using visual cues rather than GPS.

Applications include robotics, augmented reality, construction and industrial inspection.

“When we spun out, it was still very much purpose-built for augmented reality and games, even though there are many, many more applications,” said Tory Smith, director of product management at Niantic Spatial. Physical AI creates “an enormous new beachhead.”

Niantic sold its gaming business, including Pokémon GO, to Scopely for $3.5 billion last year and spun off its geospatial AI division as Niantic Spatial, led by CEO John Hanke, and funded with $250 million.

History and privacy: The spatial technologies at the core of the revamped platform trace back to data collected from Pokémon GO and Ingress players, who opted in to submit scans of real-world locations in exchange for in-game rewards.

An MIT Technology Review report last month detailed how Niantic Spatial trained its visual positioning system using some 30 billion images crowdsourced from Pokémon GO players, raising questions about whether gamers understood how their data would ultimately be used. 

Smith said the data collection was always opt-in, with no background collection, and that all data has been anonymized with GDPR-level protections applied globally.

“I would want to put aside any rumors that there was clandestine data collection or anything like that in any of our products,” he said in an interview with GeekWire this week.

How it works: The company’s visual positioning system determines a device’s exact location and orientation using what the camera sees, rather than relying on GPS. In mapped areas, it delivers centimeter-level accuracy. The new version, VPS 2.0, extends coverage globally, using additional data sources to correct for GPS errors even in areas that haven’t been scanned.

Google offers similar visual positioning through its ARCore Geospatial API, built on its vast Street View database. Smith said the key difference is that Niantic Spatial lets customers bring their own data into the platform, enabling higher-fidelity positioning in interior spaces and private areas that Google’s public Street View coverage doesn’t reach. 

Customers can also keep their maps updated over time, which Google doesn’t support.

Availability: Scaniverse will be offered through a free tier for basic scanning and positioning, with paid subscription plans adding premium features such as 360-degree camera support. 

The company said it plans to add semantic understanding capabilities later this year, enabling AI to reason about the objects and environments it encounters, not just navigate through them.