Responding to legal pressure, the Manhattan Borough President in 2014 organized a survey of curb cuts along Broadway \cite{mbpo2015}. Requiring one year to complete, the Broadway Curb Cut Survey detailed the presence and condition of safety features along the 53-kilometer street. Forty trained volunteers measured ramps at over 100 intersections on Broadway, and their efforts are captured within the Broadway Curb Cut Survey Dataset on the NYC OpenData platform. The survey, which was highly labor intensive to produce, offers a thorough picture of the accessibility status of a major urban thoroughfare.

The survey revealed that approximately 80% of the city's curb cuts do not meet ADA standards. This is despite some $243 million being spent over a 15-year period on sidewalk accessibility \cite{hogan2017}. Of 1,209 curb cuts surveyed, 28% were too steep, 60% were uneven where the cut meets the street, and 24% had crumbling concrete. Of particular concern to the visually impaired, nearly 90% lacked detectable warning features such as bumps or textured stripes.
Pressure to address the issue has come from civil society groups, notably a major lawsuit from the Eastern Paralyzed Veterans Association, which resulted in a 2002 settlement in which the city endeavored (albeit without binding timelines or monitoring) to upgrade curb cuts to ADA compliance (Hogan, op cit). 

2. Previous work

A growing research literature addresses the physical form of city streets using sensing technologies. Previous studies at TU Delft employed a photogrammetry algorithm from UMG to measure street geometries \cite{aarsen2015}. For our work, we similarly considering applying Structure from Motion (SfM) algorithms to images collected at the study site, but concluded that other methods suit the project requirements better. Curb detection methods that fuse 3D Lidar and imagery data for Autonomous Land Vehicles (ALV) are achieving advances in precise navigation of urban environments \cite{Tan_2014}. Detection of mobility challenges has also been addressed through computer vision applied to Google Street View imagery, achieving detection of missing curb cuts through a pipeline of computer-assisted human input \cite{froehlich2014}.
Complementing these sensing approaches are recent efforts to harness the potential of crowdsourcing - especially as facilitated by widespread smartphone use - to fill information gaps on urban accessibility. Specialized disability-focused crowdsourcing efforts include WheelMap, a Berlin-based non-profit with international operations, and J'Accede, a French non-profit initiative. These and similar UK and US-based initiatives have built communities of volunteers who upload user-generated information on restaurant, public building or transit accessibility. Despite success and recognition, these crowdsourcing initiatives have attracted fewer users than commercial crowd-sourced resources such as TripAdvisor and Yelp \cite{captain2017}. Given multiple demands on volunteers' time, a key consideration to build larger crowdsourcing communities is to reduce the time required for data collection.