[Paper Review] Street Sense: Learning from Google Street View

Suriyan Laohaprapanon, Kimberly Ortleb+1 more|arXiv (Cornell University)|Jul 10, 2018

Geographic Information Systems Studies3 citations

TL;DR

This paper proposes a scalable, low-cost method to assess public infrastructure quality using Google Street View imagery, combining random sampling of road segments with crowdsourced labeling via Amazon Mechanical Turk. The key finding is that Jakarta’s roads have nearly four times more potholes than other cities, with 23% of segments affected, while Wayne County, MI shows a clear income-quality gradient, where higher-income tracts have significantly fewer potholes.

ABSTRACT

How good are the public services and the public infrastructure? Does their quality vary by income? These are vital questions---they shed light on how well the government is doing its job, the consequences of disparities in local funding, etc. But there is little good data on many of these questions. We fill this gap by describing a scalable method of getting data on one crucial piece of public infrastructure: roads. We assess the quality of roads and sidewalks by exploiting data from Google Street View. We randomly sample locations on major roads, query Google Street View images for those locations and code the images using Amazon's Mechanical Turk. We apply this method to assess the quality of roads in Bangkok, Jakarta, Lagos, and Wayne County, Michigan. Jakarta's roads have nearly four times the potholes than roads of any other city. Surprisingly, the proportion of road segments with potholes in Bangkok, Lagos, and Wayne is about the same, between .06 and .07. Using the data, we also estimate the relation between the condition of the roads and local income in Wayne, MI. We find that roads in more affluent census tracts have somewhat fewer potholes.

Motivation & Objective

• To develop a scalable, cost-effective method for measuring public infrastructure quality in data-scarce urban environments.
• To investigate whether disparities in road quality correlate with local income levels, particularly in the U.S. context.
• To demonstrate the utility of Google Street View and crowdsourcing for large-scale, real-world infrastructure monitoring.
• To create a labeled dataset that can serve as a foundation for automated machine learning systems to monitor infrastructure at scale.

Proposed method

• The method uses OpenStreetMap data to identify and segment primary, secondary, tertiary, and trunk roads into 0.5 km segments.
• A random sample of these segments is selected using the open-source geo-sampling Python package for efficient spatial sampling.
• Google Street View API is queried at the starting coordinates of each sampled segment to retrieve street-level images.
• Crowdsourced labeling via Amazon Mechanical Turk is used to code images for road conditions, including potholes, cracks, litter, and sidewalk quality.
• The proportion of segments with each condition is computed and analyzed across cities and income groups.
• Statistical models are applied to assess the relationship between income quintiles and infrastructure quality, adjusting for road type.

Experimental results

Research questions

• RQ1 How does the quality of roads, sidewalks, and street cleanliness vary across major cities in the Global South and a U.S. county?
• RQ2 To what extent does road quality in Wayne County, Michigan correlate with local household income levels?
• RQ3 How does the presence of potholes and cracks vary by road type (e.g., primary, secondary, tertiary) across different cities?
• RQ4 Can Google Street View data be used to build a scalable, low-cost infrastructure monitoring system for public services?
• RQ5 What is the relationship between infrastructure quality and quality-of-life outcomes such as travel comfort and accessibility?

Key findings

• Jakarta has the worst road quality, with 23% of sampled road segments containing potholes—four times higher than Bangkok, Lagos, and Wayne County, MI, where the rate is 6–7%.
• In Wayne County, MI, higher-income census tracts have significantly fewer potholes: the proportion of segments with potholes decreases by 0.03 in the 23k–29k income quintile compared to the lowest income group.
• Crack prevalence is highest in Wayne County (62% of segments), followed by Jakarta (44%), Bangkok (24%), and Lagos (20%), with no strong correlation to road type.
• Litter is most prevalent in Jakarta (21% of segments), followed by Lagos (15%), while Bangkok and Wayne County each have 6–9% of segments with visible litter.
• Sidewalks are least paved in Lagos (30% of segments), compared to 51% in Bangkok, 49% in Jakarta, and 67% in Wayne County.
• Even after adjusting for road type, the cross-city differences in pothole and crack prevalence remain substantial, indicating that city-level disparities in infrastructure quality are not driven solely by road classification.