Overview
- Niantic Spatial, which launched Tuesday, introduced a revamped Scaniverse platform and VPS 2.0 for global visual positioning.
- Users capture spaces with phones, 360-degree cameras or drones, then the cloud turns video into 3D maps using Gaussian splatting and a depth model that aims for lidar-level surfaces.
- The system reads what the camera sees to set exact location and orientation, reaching near centimeter accuracy in mapped areas and offering positioning without a prior scan by correcting GPS errors.
- Customers can import interior or private scans and refresh maps over time, a bottoms-up model that differs from car-based Street View mapping.
- The launch follows reports that training drew on about 30 billion crowdsourced Pokémon GO images, which the company says were opt-in and anonymized under GDPR-level rules.