France Road Surface Data: 15.7M km of OSM and AI-Derived Attributes

HeiGIT produced this geospatial dataset in 2026 covering 15.76 million km of roads in France, combining OpenStreetMap (OSM) attributes with deep learning predictions from Mapillary imagery. It classifies road surfaces as paved or unpaved, specifically addressing the 58% of OSM roads in the region that lack surface tags. The data integrates AI-derived estimates with crowdsourced OSM tags to provide a more complete picture of road infrastructure.

Use Cases

• Logistics optimization by filtering for pred_class to identify and avoid unpaved routes for heavy vehicles
• Urban development analysis using the urban flag and urban_area attributes to compare infrastructure quality across regions
• Climate emissions modeling by calculating rolling resistance based on surface and smoothness values

Strengths

• Covers 15.76 million km of mapped road network
• Augments crowdsourced data with 0.6775 million km of AI-predicted surface labels
• Includes standardized urban classification via GHSU and AFRICAPOLIS layers

Limitations

• High missingness rate remains as 58.19% of the network still lacks surface information
• Model predictions in pred_class are dependent on the availability and quality of Mapillary street-level imagery
• Temporal staleness may occur due to varying update frequencies between OSM and Mapillary sources

Provenance

Source

HeiGIT (Heidelberg Institute for Geoinformation Technology)

Collection Method

Hybrid deep learning on Mapillary street-level imagery combined with OpenStreetMap data extraction and GHSU/AFRICAPOLIS urban layers.

Freshness

Last updated March 2026; update frequency is not specified.

Geography

France