Ridding The Country's Roads Of Potholes, With An Army Of Online Inspectors

Using Google Streetview and Mechanical Turk, a new project is listing roads in need of repair around the country.

Our roads are in a sorry state. The American Society of Civil Engineers (whose members wouldn’t mind being hired to fix them) gives our roads a D- for the poor conditions costing U.S. motorists $67 billion a year in repairs and operating costs, about $333 per motorist.

Since we have 2.6 million miles of paved roads in the U.S., no government agency will reasonably be able to fix or even keep tabs on them all. But that may not be true for the many denizens of the Internet.

University of Maryland researchers, possibly sick of their bumpy ride to work, have shown it’s possible to enlist hundreds of online workers to cruise Google Street View and identify streets in need of repair.

Typically, municipalities audit their roads with slow and costly drive-bys using crews who may need to make multiple trips to finalize a fix. The U.S. Department of Transportation, interested in a cheaper tool to clean up the country’s roads, encouraged Maryland professor Jon Froehlich and his team to develop software using crowdsourcing and Google Street View to solve the same problem.

The team enlisted six volunteers as well as 402 workers on Amazon’s Mechanical Turk, a crowd sourcing platform where tasks may cost as little as one cent, to label problem areas without any special training aside of an instructional video. The workers—looking at images collected by Google’s camera-equipped cars—were able to identify and label about 14,000 road problems such as potholes, obstructed pathways, broken curbs, and missing ramps (researchers are focusing first on street access). Turkers accurately spotted access issues about 80% of the time and could hit 93% with quality control measures in place. There are still problems to work out—how to select the right image point of view, identify all types of road issues, and others—but the potential seems promising. The researchers say they’ll be looking to scale the technology next.

And they’re not alone. A number of crowdsourced road repair apps and initiatives are already on the streets. SeeClickFix, FixMyStreet.com in London, the Road Damage Assessment System by Carnegie Mellon University, and Boston’s Street Bump take different approaches to giving drivers and pedestrians an easier time of it on the rough roads out there.

Add New Comment


  • Jon Froehlich

    Hi Michael,

    I am the principal investigator of this research and though we appreciate the media attention, I'm afraid our work has been misrepresented by the media. Our goal is not to use crowdsourcing to find and assess roadway quality (e.g., potholes) but rather to utilize crowds to find, label, and assess sidewalk accessibility problems. Indeed, the title of the paper you refer to in this article is "Combining Crowdsourcing and Google Street View to Identify Street-Level Accessibility Problems." 

    There are currently few mechanisms for people with mobility impairments to determine accessible areas of a city a priori (that is, before they leave their house). We are experimenting with new scalable methods to collect data on physically inaccessible areas of cities. 

    Here's a short 30 second preview of our research:

    A longer video: 

    And you can download the full paper here:

    Thanks again for taking the time to cover our work. I hope this clarifies things.


  • sanchezjb

    It's great that 14,000 road problems were identified but where's the discussion on outcomes?  How many of these road problems have been fixed?  While there are lots of discussions about crowdsourcing challenges we need to start having discussions about the difference-making outcomes that are resulting from these initiatives.