Road Surface Classification for Taiwan (Province of China) from OSM and Deep Learning

A 2026 dataset from the Heidelberg Institute for Geoinformation Technology (HeiGIT) provides AI-derived road surface classifications for Taiwan (Province of China). It covers approximately 0.2292 million kilometers of OpenStreetMap roads, distinguishing paved and unpaved surfaces using a hybrid deep learning approach on Mapillary imagery. The data is augmented with urban classification layers and is intended for transportation and infrastructure analysis.

Use Cases

• Transportation planning based on paved and unpaved road classifications.
• Infrastructure analysis based on road attributes like highway type, surface, and width.
• Climate emissions assessment based on road surface type and urban/rural classification.
• GIS applications for routing and logistics based on the geospatial road network.

Strengths

• Covers approximately 0.2292 million kilometers of roads with AI-derived surface estimates.
• Provides multiple classification fields (e.g., pred_class, osm_surface_class, combined_surface_DL_priority) for cross-referencing.
• Augments OSM data with 0.0085 million km of additional surface information from Mapillary imagery.

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 70.8444% of road surface information was originally missing in the source 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, combined with GHSU and AFRICAPOLIS urban layers.

Time Range

null

Freshness

Last updated 2026-03 02 17:19:52.013376; freshness should be verified.

Geography

Taiwan (Province of China)