11/27/2024
In today’s travel news… Thank you all for playing Pokémon Go…
Niantic is leveraging player-contributed scans from games like Pokémon GO to develop a Large Geospatial Model (LGM), an advanced AI system designed to understand and interact with physical spaces. This approach uses voluntarily collected 3D scans of public locations to train models that support spatial intelligence and enable precise augmented reality (AR) experiences.
How Pokémon GO Contributed to Data Collection:
Niantic uses player-uploaded scans of real-world locations to build detailed, pedestrian-level 3D maps. These scans are optional and require users to visit specific public sites and actively scan the area using their device. This data feeds into their Visual Positioning System (VPS), which determines precise device location and orientation. Unlike passive gameplay, such as walking around while playing Pokémon GO, this scanning activity directly trains AI models.
Building the Large Geospatial Model:
The LGM extends the capabilities of VPS by consolidating local 3D models into a global system. This system:
Uses billions of images and scans tied to specific geographic locations.
Comprehends how individual locations relate to others worldwide, even when only partially scanned.
Captures 3D spatial data in metric scale for precise mapping, unlike traditional 3D models which lack geographic anchoring.
Applications:
Enhanced AR Experiences: Persistent and interactive AR objects (e.g., Pokémon Playgrounds where Pokémon remain at specific locations for others to see).
Navigation & Guidance: Real-time assistance using AR glasses or audio-based interfaces.
Spatial Planning & Design: Advanced tools for urban planning, logistics, and real-world modeling.
Future Technologies: Integration with robotics, autonomous systems, and content creation.
Broader Implications:
The LGM combines 3D vision, spatial intelligence, and geospatial data to enable real-world applications far beyond gaming. This includes next-generation maps, enhanced human-environment interaction, and new AR-enabled wearable technologies.
Niantic’s approach highlights the transition from AI understanding 2D and static data to interpreting 3D, dynamic, and spatially-rooted environments, signaling a significant advancement in geospatial intelligence.
At Niantic, we are pioneering the concept of a Large Geospatial Model that will use large-scale machine learning to understand a scene and connect it to millions of other scenes globally.
