Russian Federation Road Surface Data: 29.7 Million KM of OSM Networks

Delivering surface classifications for 29.75 million km of road networks across the Russian Federation, distinguishing between paved and unpaved segments. Created by the Heidelberg Institute for Geoinformation Technology (HeiGIT) in 2026, it integrates OpenStreetMap data with deep learning predictions derived from Mapillary street-level imagery.

Use Cases

• Infrastructure assessment using pred_class to identify unpaved segments
• Urban planning analysis using the urban and urban_area attributes
• GIS routing optimization based on the surface and smoothness columns

Strengths

• Covers 29.75 million km of road network
• Includes 1.33 million km of AI-augmented surface data not found in standard OSM
• Provides urban/rural classification via the urban attribute

Limitations

• 51.6% of roads lack surface information
• Potential prediction errors from the hybrid deep learning model
• Spatial bias toward areas with available Mapillary imagery

Provenance

Source

OpenStreetMap, Mapillary, GHSU Global Human Settlement Urban Layer 2019, and AFRICAPOLIS 2020

Collection Method

Hybrid deep learning approach combining street-level imagery analysis with OSM attribute extraction

Freshness

Last updated March 2026.

Geography

Russian Federation