Mapillary, a hosting platform for street-level imagery, has some 40 million images from 270,000 miles of roadways from state transportation departments in Utah, Florida, Arizona, Connecticut and Vermont. The images have been processed with "computer vision” to identify key items like signage and striping. The data is used by DOTs to examine any given stretch of highway for safety, suitability and other criteria.
“Instead of investing in the reprogramming of an outdated viewer system, our GIS team worked with Mapillary to host ADOT’s historical imagery free of charge,” said Doug Nintzel, a spokesman for the Arizona Department of Transportation. “Our goal, in addition to saving tax dollars, is to be transparent in the sharing of this public data.”
The Arizona DOT began working with Mapillary in 2017.
Part of the strength of Mapillary, said Janine Yoong, vice president of business development at Mapillary, is the platform’s ability to apply its computer vision to any number of images uploaded to the system.
“I think what we’re really trying to showcase with these different participants is that anyone, with any camera — whether it’s professional grade, whether you have millions of images, or a few thousand — you can contribute to the vision that we have of collaborative mapping,” said Yoong. “With computer vision, it’s possible for us to do a large amount of extraction in ways that we couldn’t have before. … But the real key to unlock this, is the fact that we are really focused on collaborative mapping, opening up the ability for anyone to contribute."
Arizona has about 6,800 miles of roadways on Mapillary’s system, said Nintzel, adding that the imagery is used to analyze state-owned assets like signs and pavement striping, but also roadway characteristics like the number of lanes on a given stretch of highway.
"That imagery is helpful for surveying purposes, allowing for analysis in an office rather than sending a crew into the field to verify,” he added.
If the state DOTs are submitting millions of images to Mapillary, along with the data filed by individual communities, “we are actually building a database of imagery that can be used to create the best maps in the world,” said Yoong. “It’s not just enough to have fresh, accurate data at the highway level, but also to have fresh, accurate data at the sidewalk or the curb level.”
And it’s not just transportation officials who can collect images. In the coming days of autonomous vehicles — which include technology to visually assess the landscapes they are driving through — those cars could be a source for imagery.
“We also speak with a lot of forward-thinking DOTs that are interested in collaborating with automotive OEMs [original equipment manufacturers] on autonomous vehicles,” said Yoong. “What we are able to do is play a very specific role where we say, 'Listen, we have this platform where we can surface this data, whether the imagery is collected by you — the state DOT — or you, the automotive OEM.' What we’re able to do is collect and compile all of this imagery and really provide information that hopefully will help improve safety standards.”
