From street-level imagery to actionable mobility metrics: A deep learning framework for computing wheelchair accessibility from pedestrian sidewalk infrastructure features
Year: 2026
Type: article
Abstract: Urban space is not equally accessible to all city inhabitants. For individuals with mobility disabilities, physically open spaces can remain functionally inaccessible, as inadequate sidewalk surfaces, excessive slopes, or insufficient clearance can turn a seemingly traversable area into an impassable one. Although the need to quantify these disparities is increasingly recognised, significant gaps persist in producing reliable, scalable, and actionable accessibility data. This study introduces a processing pipeline that leverages street-level imagery and deep learning techniques to estimate key characteristics of pedestrian sidewalk infrastructure — including sidewalk segmentation, surface type and condition detection, effective clearance estimation, and slope calculation — and synthesises them into an actionable dataset followed by a composite sidewalk wheelchair accessibility index (WHAI). The Wheelchair Accessibility Dataset (WHAD) pipeline combines monocular metric depth estimation, 3D point cloud reconstruction, and terrain model analysis to derive measurements from publicly available Mapillary street-level imagery. Results are validated against unmanned aerial vehicles (UAVs) and aerial orthoimagery across areas of varying accessibility in Łódź, Poland, and Heidelberg, Germany. The resulting index enables the identification of localised inaccessible enclaves that conventional spatial analyses often overlook, providing an explainable and comparable foundation for evidence-based urban planning and improvements to pedestrian routing.
Authors: Maciej Adamiak(Data curation, Formal analysis, Investigation, Methodology, Project administration, Software, Validation, Visualization, Writing – original draft, Writing – review & editing), Gefei Kong(Conceptualization, Data curation, Formal analysis, Methodology, Software, Validation, Writing – original draft, Writing – review & editing), Danielle Gatland(Conceptualization, Software, Validation, Writing – original draft, Writing – review & editing), Andrzej K. Oleś(Conceptualization, Software, Writing – review & editing), Kirsten von Elverfeldt(Conceptualization, Supervision, Writing – review & editing), Alexander Zipf(Supervision, Writing – review & editing), Sascha Fendrich(Writing – review & editing)
Institution: Heidelberg University
Language: English
Subfield: Transportation
Field: Social Sciences
Domain: Social Sciences
Open Access status: hybrid
Estimated APC paid: $3,740
Funder: Klaus Tschira Stiftung
OpenAlex ID: W7235648998
