Brazil Road Surface Classification from OSM and AI Predictions

A 2026 dataset from the Heidelberg Institute for Geoinformation Technology (HeiGIT) provides AI-derived road surface classifications for Brazil. It covers approximately 4.57 million kilometers of OpenStreetMap roads, distinguishing paved and unpaved surfaces using a hybrid deep learning model. The data is augmented with urban classification layers and Mapillary imagery to address missing surface information.

Use Cases

• Transportation planning based on paved vs. unpaved road classifications.
• Infrastructure analysis using combined surface classifications from OSM and deep learning predictions.
• Climate emissions assessment leveraging road surface and urban/rural classification data.
• GIS applications utilizing the geospatial road network with attributes like highway type and surface smoothness.

Strengths

• Covers approximately 4.57 million kilometers of roads in Brazil.
• Provides AI-derived predictions for 15.3% paved and 37.3% unpaved roads, addressing gaps in OSM data.
• Integrates multiple data sources: OSM, Mapillary imagery, and global human settlement layers.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Approximately 47.4% of road surface information is still missing from the core OSM data.

Provenance

Source

Heidelberg Institute for Geoinformation Technology (HeiGIT)

Collection Method

Derived from OpenStreetMap (OSM) and augmented with deep learning predictions based on Mapillary street-level imagery.

Time Range

null

Freshness

Last updated 2026-03-02 17:16:04.701527; freshness should be verified.

Geography

Brazil