A Pilot Deployment of an Online Tool for Large-Scale Virtual Auditing of Urban Accessibility
Manaswi Saha, Kotaro Hara, Soheil Behnezhad, Anthony Li, Michael Saugstad, Hanuma Maddali, Sage Chen, Jon E. Froehlich · 2017 · Proceedings of the 19th International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS '17) · doi:10.1145/3132525.3134775
Summary
This demonstration paper introduces Project Sidewalk, a web-based tool that enables anyone with an internet connection to remotely audit urban sidewalk accessibility by virtually walking through city streets using Google Street View. The tool addresses a significant gap in geographic information systems: while platforms like Google Maps contain vast data about roads and points of interest, they include almost no information about physical accessibility. Existing crowdsourced accessibility platforms like Wheelmap.org and Axsmap.com suffer from severe data sparseness — Ding et al. found that only 1.6% of points of interest in Wheelmap had accessibility data entered. These platforms are limited because they require local users with physical experience of a place. Project Sidewalk removes this constraint by leveraging Street View imagery, opening participation to a global pool of volunteers. The system builds on five years of prior research into virtual auditing interfaces, including tools like CANVAS, Spotlight, Bus Stop CSI, and Tohme, which demonstrated that virtual audits have high concordance with traditional physical audits. Project Sidewalk extends this work by adding interactive onboarding tutorials, gamification elements (mission-based tasks, progress dashboards, completion sounds), and open-world exploration to train, engage, and sustain non-expert volunteers.
Key findings
The pilot deployment in Washington DC attracted 581 users who contributed 71,873 accessibility labels across 504 miles of streets — nearly 50% of all streets in the city. Users completed an average of 3.6 missions (2.9 miles audited per user), taking about 16.9 minutes per 1,000 feet of auditing. Of the 581 users, 54.2% chose to register, and registered users contributed the vast majority of labels (58,887 vs. 12,986 from anonymous users). The most commonly labeled features were curb ramps (64.3%), followed by missing sidewalks (17%), missing curb ramps (7.2%), obstacles in path (5.6%), and surface problems (4.9%). Missing curb ramps received the highest average severity rating (3.8 out of 5). Prior work suggests 80-90% labeling correctness. The team estimated that auditing a city the size of DC using paid Mechanical Turk workers at minimum wage would cost approximately $14,000 — orders of magnitude less than traditional physical audits. Multiple city departments of transportation expressed interest in deploying the tool in their cities.
Relevance
Project Sidewalk represents a scalable, cost-effective approach to a problem that directly affects millions of people with mobility impairments: the lack of accessible route information for pedestrian navigation. Traditional sidewalk accessibility audits are prohibitively expensive and time-consuming, leaving most cities without comprehensive data on sidewalk conditions. This tool demonstrates that non-expert volunteers can be trained through gamification to perform meaningful accessibility audits remotely, democratizing both the data collection process and the resulting accessibility information. For accessibility practitioners and urban planners, the project offers a model for how crowdsourcing combined with Street View imagery can generate actionable data about curb ramps, sidewalk obstacles, and missing infrastructure. The planned suite of downstream tools — AccessScore for neighborhood ratings, AccessRoute for personalized routing, and AccessModel for socioeconomic analysis — points toward practical applications that could transform how wheelchair users and others with mobility impairments plan their travel.
Tags: crowdsourcing · urban accessibility · GIS · sidewalk accessibility · curb ramps · mobility impairment · virtual auditing · citizen science · gamification
Standards referenced: ADA