Optional scans from Pokémon Go and Ingress helped train Niantic Spatial’s visual-positioning model. The company now pairs it with Vantor’s Raptor platform for GPS-denied drones, vehicles and field teams.
Pokémon Go players who scanned landmarks for a mobile game helped build a visual map that Niantic now markets into defense and intelligence work.
Niantic told Ars Technica that optional player scans of public locations helped train its real-world foundation models. Ars reported that Niantic Spatial trained its Large Geospatial Model on 30 billion images, mostly around urban locations.
The resulting Visual Positioning System does not depend on satellite signals. It compares a live camera feed with a visual model of the world to locate the user or machine — the exact problem that surfaces when GPS gets jammed, spoofed or disappears.
“The models are the product of that training, not a copy of or a means of accessing the underlying scans.”— Niantic Spatial spokesperson, via Ars Technica
In December, Niantic Spatial and Vantor announced a defense partnership connecting Niantic’s ground-level system with Vantor’s Raptor software for aerial platforms. The firms said drones, vehicles, AR glasses and field teams could share coordinates without GPS.
A March update called the integration a “mission-ready joint capability” and reported early tests cut positioning error by up to 70%, reaching roughly 1.5 meters in many scenarios. Niantic now markets VPS to defense and intelligence customers, while Vantor pitches Raptor for GPS-denied battlefield reconnaissance and tactical operations.
Vantor says it does not receive Pokémon Go data. Niantic says raw scans are not shared. But the model trained from player scans is moving into the GPS-denied defense stack.