Building a street-level imagery program

In 2015, Detroit was in the thick of an asset management problem. The city hit its lowest-ever rating by the U.S. Insurance Services Office, largely due to a lack of information on the whereabouts and status of its 30,000 fire hydrants. This information gap posed emergency management issues and affected residents with skyrocketing home insurance rates. Detroit already had the most expensive auto insurance rates in the nation, making heightened home insurance rates all the more critical to address.

The City of Detroit (CoD) identified that street level imagery (SLI) could be key to improving their situation, serving as a database of their assets and a way to monitor current conditions. They first turned to vendors, targeting SLI on a project-by-project basis. They quickly encountered a problem: caveats in imagery licenses prohibited imagery sharing between city departments (or with the public)—meaning that the city had to pay for multiple sets of SLI for an area so that multiple teams could use it.

Realizing this would not scale, Detroit set out to find a cost-effective, centralized platform for street level imagery that would support all of its 30 departments—and that’s when they found Mapillary!

Dexter Slusarski, a former GIS Analyst for the CoD, pioneered Mapillary in Detroit. His pilot runs consisted of strapping GoPro cameras to a van and driving around the city. The resulting imagery was presented at a meeting with the mayor of Detroit, who was impressed by the ease at which they could view the city.

Looking to take the project further, Dexter applied to and won a Mayor’s Innovation Grant in 2018. This funded the purchase of a van and the Trimble MX2 mobile mapping system, enabling them to capture comprehensive 360° imagery. A full time driver was hired to do just that in November 2018.

Gaining momentum

Detroit’s focus was initially on imagery of physical space, with employees sharing Mapillary images across departments to provide evidence on the status of Detroit roads and buildings for various initiatives. As demand for SLI grew, the GIS team began to extract Mapillary-generated fire hydrants, traffic signs, cross walks, and other map features. They ingested these point features into their own GIS tools to further synthesize and document their asset data internally.

The CoD remained very involved with the Mapillary team. They contributed imagery for our Mapillary research datasets and were champions of our (since deprecated) Verification Projects feature, a service we offered to Mapillary organizations to improve both their generated map data and our models as we refined our detection algorithm in our earlier days. The Mapillary Verification Projects brought a competitive spin to team efforts in improving Detroit’s asset database—the top employees on the Verification tool got huge bragging rights around the office!

By the time we wrote our 2019 case study on Detroit, just one year after the program's inception, over half of Detroit’s 2,900+ miles of roadway had been mapped via 1.48 million Mapillary images!

2024 update

From an imagery contribution standpoint alone, it is clear that Detroit remains strongly committed to open SLI—as of November 2024, 99.8% of all public roads in Detroit have been mapped, most several times over, making Detroit the best mapped city on Mapillary! In addition to roads, the city has extensive coverage of parking lots, development complexes, and alleys which are typically missed by commercial SLI providers.

In total, the City of Detroit has contributed over 12,760,000 images; the codgis account is the #24 all-time contributor to Mapillary.

To learn more about how Detroit’s SLI program has grown in the past five years, we sat down with two key leaders of the City of Detroit’s SLI program: Tamara Fant, Enterprise Applications Manager, and Ted Schultz, Enterprise GIS Team Supervisor.

Tamara and Ted shared that Detroit is more invested than ever in SLI. In 2020 the CoD GIS Team was awarded a $400,000 grant from General Motors to upgrade their capture platform. In 2023 they used this grant to acquire a Trimble MX50, which has both improved 360° image quality and LiDAR capabilities. Additionally, they’ve grown their capture team to two full time drivers. Between the two drivers, they have someone out in the field collecting imagery every day, weather permitting!

Detroit has continued to find value in Mapillary imagery and map features across a spectrum of usages. Having a consistent record of SLI over the years has enabled the CoD to do the following:

• Document changing landscapes, detect blight, identify redevelopment opportunities, and showcase infrastructure and public works improvements
• Extract stop sign, manhole, and pole map features to identify signage gaps, determine asset quality, and identify camera locations for public safety
• Partner with the Frederick Douglass Academy GIS Pathways Program to encourage GIS careers and expose students to geospatial field work

The CoD GIS Team is also working to develop in-house computer vision models to extract additional intelligence from SLI. Ted is leading the team developing these models; they’re currently using feature detection algorithms within bounding boxes to detect urban blight indicators such as broken windows, vandalism, and trash. Ted said it best: “if you capture everything, you can extract almost anything.”

Mapillary as a means to a census challenge

We’d be remiss if we didn’t share what is arguably the most unique way that Detroit has used Mapillary: Detroit SLI hosted on Mapillary was used when the city challenged the Census Bureau’s undercount of its population in the annual population estimates released following the 2020 Census.

The 2020 Census count of Detroit was hampered not only by the COVID pandemic, but also because many houses were removed from the Census Bureau’s tally of the city’s housing stock due to poor condition or undelivered mail. Using SLI to investigate, analyze, and identify miscounted homes (by examining exterior indicators like windows, doors, and roof conditions), the city was able to present compelling evidence that the city’s population was underestimated.

Detroit’s diligent approach to fresh imagery was key to its success. The fact that the city had captured SLI recently before and after the 2020 Census enumeration window (which would not have been the case if relying on third party SLI) was crucial to support the city’s claims.

Detroit ultimately persuaded the Census Bureau to change some of its estimating methods and experiment with using more “administrative data” such as imagery documentation. The initial adjusted numbers indicate that Detroit’s population has grown for the first time since 1957! This anecdote is a great example of the power that an organization can harness by routinely capturing and hosting SLI on Mapillary.

So, what's next for Detroit?

Detroit is committed to the continued success of their program and are looking to take it one step further by adding UTVs (equipped with action cameras) into their capture fleet. This expansion would allow them to capture imagery of their greenways, paths, riverfront, and other car-restricted areas. As pedestrian spaces continue to grow in an ever-evolving Detroit, there is an increasing opportunity to document such spaces.

To conclude, we’ll end with a quote from Tamara on why Mapillary is Detroit’s preferred SLI platform:

We are incredibly thankful for our partnership with Tamara, Ted, and the City of Detroit over the past decade. If you haven’t yet, be sure to check out the Mapillary imagery in Detroit, the most comprehensive coverage of a city in Mapillary—it is truly remarkable.

To learn more about Detroit’s SLI program, visit the DetroitStreetView website.

/Marcie