A Mobile Application for Crowdsourced Acquisition of Urban Street-View Pedestrian Facility Data

In recent years, pedestrians have been dangerously overrepresented in traffic crashes, and the pedestrian fatality rate has steadily increased during the last decade. Additionally, studies have shown that the majority of pedestrian-involved traffic accidents occur in urban non-intersections, which suggests that a more well-connected pedestrian facility network in cities would lower the rate of pedestrian involvement in traffic accidents. One way to improve the pedestrian facility network coverage is to first have up-to-date, accurate, and thorough data regarding the presence of existing pedestrian facilities. However, state departments of transportation have stated that the current methods of acquiring this data are expensive and time consuming. In this project, we developed a mobile application prototype for crowdsourced acquisition of street-view images containing pedestrian facilities, or more specifically, crosswalks. The resulting application used modern full-stack development techniques and is a native Android application that allows the user to take pictures using their mobile devices and automatically upload those pictures, along with relevant metadata (such as location data), to a server where they are classified using a machine learning model that was trained to recognize the presence of crosswalks in images