Paved or unpaved? A deep learning derived road surface global dataset from mapillary street-view imagery

Abstract

Road surface information is essential for applications in urban planning, disaster routing or logistics optimization and helps to address various Sustainable Development Goals (SDGS): especially SDGs 1 (No poverty), 3 (Good health and well-being), 8 (Decent work and economic growth), 9 (Industry, Innovation and Infrastructure), 11 (Sustainable cities and communities), 12 (Responsible consumption and production), and 13 (Climate action). We have released an open dataset with global coverage that provides road surface characteristics (paved or unpaved). The data was derived by a GeoAI approach that utilized 105 million images from the world’s largest crowdsourcing-based street-view platform, Mapillary. We propose a hybrid deep learning approach which combines SWIN-Transformer based road surface prediction and CLIP-and-DL segmentation based thresholding for filtering of bad quality images. The road surface prediction results have been matched and integrated with OpenStreetMap (OSM) road geometries. Model validation against OSM surface data achieved strong performance, with F1 scores for paved roads varying between 91%–97% across continents. The dataset expands the availability of global road surface information by nearly four million kilometers compared to currently available information in OSM — now representing approximately 36% of the total length of the global road network. Most regions showed moderate to high paved road coverage (60%–80%), but significant gaps were noted in specific areas of Africa and Asia. Urban areas tend to have near-complete paved coverage, while rural regions displayed more variability. This information has the potential to derive more reliable estimations for indicators such as rural accessibility or regional economic development potential and to assist e.g. humanitarian actors in emergency logistic planning.