Automated framework for extracting sidewalk dimensions from images using deep learning

State and local governments are required by federal state laws to provide maintain accessibility on their sidewalks pedestrian facilities. They need conduct frequently update self-evaluation assess the compliance of facilities with requirements identify any barriers that limit or deny access for people disabilities public programs, services, activities. This paper presents development an automated framework is capable (1) providing a cost-effective practical methodology conducting self-evaluations using sidewalk images, (2) creating 3D models existing can be used in analyzing conditions, (3) automatically extracting dimensions geometry from input images. A case study small network includes 830 m was analyzed test performance demonstrate its novel capabilities.