Researchers have introduced TOL, a new benchmark and framework for text-to-OpenStreetMap (T2O) localization. This approach estimates a 2D position in urban environments solely from textual descriptions, without relying on geometric observations or GNSS. The TOL benchmark comprises approximately 121,000 textual queries paired with OSM map tiles across Boston, Karlsruhe, and Singapore. The proposed TOLoc framework uses a coarse-to-fine method to model semantics and directional information, achieving improved localization performance. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
IMPACT Introduces a new method for geospatial localization using text and OpenStreetMap, potentially enabling new location-aware applications.
RANK_REASON The submission is an academic paper introducing a new benchmark and localization framework.