Norway Road Surface Data: 1.9 Million KM with AI-Derived Classifications

HeiGIT produced this dataset in 2026, covering 1.9661 million km of road segments in Norway. It integrates OpenStreetMap attributes with AI-derived surface classifications generated from Mapillary street-level imagery to identify paved and unpaved segments.

Use Cases

• Logistics planning by filtering pred_class to identify paved routes for heavy vehicle transport
• Infrastructure gap analysis by comparing osm_surface_class with combined_surface_DL_priority to update local maps
• Environmental impact assessment using predicted_length and surface type to estimate dust or runoff levels

Strengths

• Covers 1.9661 million km of road infrastructure
• Provides dual-priority fields (combined_surface_osm_priority and combined_surface_DL_priority) for flexible data integration
• Includes urban classification attributes like urban_area and urban (0/1) flags

Limitations

• Over 50% of the road network still lacks surface information despite AI augmentation
• AI-derived surface information only accounts for 1.58% of the total road length
• Model predictions are dependent on the availability and quality of Mapillary street-level imagery

Provenance

Source

Heidelberg Institute for Geoinformation Technology (HeiGIT)

Collection Method

Hybrid deep learning model applied to Mapillary imagery combined with OpenStreetMap data extraction.

Freshness

Last updated March 2026

Geography

Norway