Wrivinder: Towards Spatial Intelligence for Geo-locating Ground Images onto Satellite Imagery
Wrivinder is a zero-shot framework that aligns ground photographs with satellite maps by reconstructing a 3D scene from multiple ground images. It uses geometric reconstruction (SfM and 3DGS) and semantic cues to infer camera positions and generate a zenith view, which is then matched to satellite imagery using self-supervised alignment, enabling geolocation without paired training data.