Sri Lanka Road Surface Classification from OpenStreetMap and AI

HeiGIT provides a dataset detailing road surfaces in Sri Lanka, distinguishing paved and unpaved roads. It covers approximately 163,700 km of roads mapped in OpenStreetMap, with AI-derived estimates classifying 17.2% as paved and 11.8% as unpaved. The dataset was last updated on March 2, 2026.

Use Cases

• Transportation planning based on road surface type and highway classification.
• Infrastructure analysis using attributes like width, lanes, and bridge status.
• Climate emissions assessment leveraging the distinction between paved and unpaved road networks.
• GIS applications for routing and accessibility studies based on urban/rural classification.
• Validating and augmenting OpenStreetMap data with AI-derived surface predictions.

Strengths

• Includes AI-derived surface predictions for approximately 163,700 km of roads.
• Augments OSM data with deep learning predictions from Mapillary imagery.
• Provides urban/rural classification using GHSU and AFRICAPOLIS layers.
• Offers multiple classification outputs (OSM-priority and DL-priority).

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Approximately 70.98% of road surface information is reported as missing in the base OSM data.

Provenance

Source

OpenStreetMap (OSM), augmented with deep learning predictions from Mapillary imagery and urban layers from GHSU and AFRICAPOLIS.

Collection Method

Surface classification derived using a hybrid deep learning approach, as detailed in the referenced arXiv paper.

Time Range

The mean timestamp of predictions used is included in the 'DL_mean_timestamp' attribute.

Freshness

Last updated 2026-03-02 17:18:08.843280; freshness should be verified.

Geography

Sri Lanka.