05-18-23 | News

New AI Tool Allows for Sidewalk Mapping

With Machine Learning, a New Tool can Scan Sidewalks from Satellite Images
by Staff

This could allow for cities' pedestrian infrastructure improvements to be fundamentally improved, as many cities still use manual documentation.

Researchers, Maryam Hosseini and Andres Sevtusk, from the Department of Urban Studies and Planning, Massachusetts Institute of Technology (MIT), Fabio Miranda, from the Department of Computer Science, University of Illinois at Chicago (UIC), Roberto M. Cesar Jr., from the University São Paulo, and Claudio T. Silva, from the Department of Computer Science and Engineering, New York University, and the Center for Data Science, New York University, have recently unveiled a new tool that uses satellite images to map out sidewalks and crosswalks using machine learning. Titled Tile2Net, the researchers have stated that the program is "the first open-source scene classification model for pedestrian infrastructure from sub-meter resolution aerial tiles," which means that they use Google Maps to get accurate measurements of sidewalks and crosswalks.

Reportedly, the tool has 90% accuracy at identifying those sidewalks when compared against maps that cities developed through the conventional method of manual auditing and tracing of infrastructure. Members of the research team even mentioned that the process of auditing sidewalks is so arduous that many cities just choose to forgo it entirely.

As a result, it is left to the citizen that has to use the sidewalk to report when it is in need of repair and improvement and that it is essentially impossible to create a new sidewalk plan unless canvassing of the entire city is completed first. This leaves many sidewalks in disarray, causing problems for citizens that live and use them, especially those with disabilities that need their sidewalks in working order just to get around. And worse, if they want to get them fixed, there is often a long bureaucratic process that may simply result in fining the property owner until they fix the pedestrian walkway.

The researchers that developed this machine learning tool hope that cities will be able to look at accurate data of where their sidewalks are, and then use that data to plan out pedestrian walkways more efficiently and with modern planning trends in mind. This will make it easier and less costly for cities to partake in the pedestrian-oriented streets that so many planners and pedestrians themselves are currently pushing for.


Sign up for
LAWeekly newsletter. Get exclusive content today.