It is easier than ever to examine maps of any location you choose to visit by automobile. On foot is a another thing. The majority of American cities and towns lack sidewalk maps, leaving pedestrians to fend for themselves. Can you walk to the eateries on the opposite side of the highway from your hotel? Is there a quicker route between downtown and the sports arena? And how exactly do you get to that bus stop?
Now MIT researchers, together with colleagues from many other universities, have built an open-source program that leverages aerial images and image-recognition to produce full maps of sidewalks and crosswalks. The tool can aid planners, lawmakers, and urbanists who seek to increase pedestrian infrastructure.
“There is a significant need in the urban planning and urban policy disciplines,” says Andres Sevtsuk, an associate professor at MIT and co-author of a recent study describing the capabilities of the tool. “Most municipal administrations in the United States know virtually little about their sidewalk systems. It contains no data. The business sector has not undertaken the mapping effort. That felt like a really vital technology to create, especially in an open-source approach that may be utilized by other places.”
The program, named TILE2NET, has been built using a few U.S. locations as initial sources of data, but it may be modified and adapted for usage everywhere.
Maryam Hosseini, a postdoc in MIT’s City Form Lab in the Department of Urban Studies and Planning (DUSP), whose research has been centered on the development of the tool, explains, “We believed we needed a system that could be utilized in many cities and be scalable.”
The research, “Mapping the Walk: A Scalable Computer Vision Method for Creating Sidewalk Network Datasets from Aerial Imagery,” appears online in the journal Computers, Environment and Urban Systems. The authors are Hosseini; Sevtsuk, who is the Charles and Ann Spaulding Career Development Associate Professor of Urban Science and Planning in DUSP and director of MIT’s City Form Lab; Fabio Miranda, an assistant professor of computer science at the University of Illinois at Chicago; Roberto M. Cesar, a professor of computer science at the University of Sao Paulo; and Claudio T. Silva, Institute Professor of Computer Science and Engineering at New York University (NYU)
Substantial research for the project was undertaken at NYU while Hosseini was a student, in collaboration with Silva.
There are several approaches of mapping sidewalks and other pedestrian paths in urban areas. Planners could create maps manually, which would be accurate but time-consuming; they could use roads and make assumptions about the extent of sidewalks, which would reduce accuracy; or they could attempt to track pedestrians, which would likely fall short of revealing the complete extent of walking networks.
Instead, the study team employed computational image-recognition techniques to create a tool that can visually identify sidewalks, crosswalks, and pathways. To do this, the researchers utilized 20,000 aerial pictures from Boston, Cambridge, New York City, and Washington – areas for which complete pedestrian maps already existed — as a starting point. By training the image-recognition model on such precisely defined items and utilizing sections of these cities as a starting point, scientists were able to determine how effectively TILE2NET would function in other areas of these cities.
In the end, the technology was effective, identifying at least 90 percent of all sidewalks and crosswalks in Boston and Cambridge, for example. Having been educated visually on these cities, the program may now be deployed to other metropolitan regions; individuals in other locales can now enter their aerial photos into TILE2NET.
Hosseini explains, “We wanted to make it easier for towns in other regions of the world to accomplish this without having to undertake the hard work of training the tool.” Together, we will hopefully make it better and better as we progress.
Sevtsuk, whose research focuses on pedestrian and nonmotorized mobility in cities and who has developed many types of pedestrian-mapping tools over his career, highlights the great need for such a tool. He observes that the networks of walkways and routes for pedestrians in the majority of cities are grossly deficient. Yet, it is difficult to efficiently extend these networks without mapping them.
“Imagine if automobile networks had the same gaps as pedestrian networks,” Sevtsuk explains. “After reaching an intersection, the road terminates abruptly. Maybe you cannot make a right turn because no road exists. This is a persistent obstacle for walkers, and we fail to see the significance of continuity for pedestrian networks.”
In a broader context, Sevtsuk notes that the continuance of climate change would need the expansion of cities’ infrastructure for pedestrians and bicycles, among other measures; transportation continues to be a significant source of carbon dioxide emissions.
“When cities discuss reducing carbon emissions, there is no alternative to addressing transportation,” adds Sevtsuk. “The quality of urban statistics for public transportation, pedestrians, and bicycles lags substantially behind that of car data.” This type of information is required for analyzing the viability of car-free cities.
On the bright side, according to Sevtsuk, the addition of pedestrian and bicycle infrastructure is being carried out more vigorously than in decades past. Throughout the 20th century, sidewalks were eliminated to create room for automobile roadways. We currently observe the opposite tendency. To make the greatest use of pedestrian infrastructure, it is crucial for towns to have network information about it. Now you can accurately determine how someone can reach a bus stop.”