Using AI and Machine Learning to Proactively Address Urban Blight

October 28, 2022 - (4 min read)

Google Cloud is helping state and local governments empower their workforce and improve the lives of their constituents with our secure, interoperable, intelligent platform. Whether your organization is looking to build new applications in the cloud or transform your current infrastructure, we can help modernize service delivery.

Municipalities across the country are finding new ways to combine technology with existing resources to combat age-old challenges. Road maintenance and urban blight are two examples of those challenges many cities find difficult to address. Both issues have a deep impact on the experience of living and traveling within a city. The City of Memphis, Tennessee, like many cities, struggled with these issues. They needed new ways to track and manage disrepair and improve the quality of life for their 652,000 residents.

Potholes are often one of the most visible signs of a city’s ability to provide services efficiently. Striving to make the city a better place to live and drive, Memphis was working hard to fix potholes in the 6,800 miles of road lanes they handle. Fewer potholes would mean safer streets and less paid out in city claims for unaddressed potholes, but they were having trouble finding the potholes that needed their attention. Residents rarely report potholes, and only 20% were reported on 311 systems. Considered a minor nuisance, most potholes were left unreported and, therefore, unrepaired and worsening over time. The city needed an efficient, systematic way to find potholes so they would be able to address road maintenance in a more proactive way.

Further impacting the quality of life for residents, vacant properties contributed to urban blight. The city believed that some of these properties were owned by out-of-town investors while others were simply abandoned. With no consistent way of identifying distressed properties, they were solely reliant on residents calling in to report them. In an attempt to combat this problem, Memphis enlisted a small army of 200 volunteers to try and track abandoned properties. It didn’t help; the project took too long, and the results were inconsistent in determining which areas needed assistance.

Realizing they needed to try a different approach, the City of Memphis engaged Google Cloud and SpringML, a Google Cloud Partner, to create a machine learning (ML) proof-of-concept (POC). Working together, they created a solution that could identify potholes quickly and accurately, as well as identify whether property had overgrown grass or tall weeds. To supply the models with the needed data, teams gathered and analyzed 30 days of video from a city bus and 360-degree high-resolution video from a code enforcement vehicle. They began by training TensorFlow models for ML object detection using pre-configured artificial intelligence (AI) Platform Deep Learning VM Images on Compute Engine.

“The City of Memphis has been a proud partner with Google Cloud and SpringML in developing these incredible and groundbreaking capabilities,” said Robert Knecht, the Director of Public Works for the City of Memphis. “This project will enable Public Works to be a revolutionary leader in efforts to proactively tackle municipal challenges in a more strategic and effective way like never before.”

The teams then layered this video with paving data, geolocation data from ArcGIS and Google Maps, city property records, tax records, and 311 reports that had been imported and analyzed in BigQuery. This created a model showing exactly which areas of the city needed repair. As this model evolved, it could consistently differentiate between potholes, maintenance holes, and other roadside objects.

Armed with the data they needed, Memphis began addressing the disrepair within the city. They reduced the number of potholes on their streets, creating a better driving experience for residents and visitors. Within three months, more than 800 potholes were approved for action, ultimately identifying 75 percent more potholes than before. They also identified more than 100 potholes on state-maintained roads and referred those to the Tennessee DOT. The new models also empowered Memphis Code Enforcement to make data-driven decisions on how to deal with abandoned or derelict properties. It allows social services to be more targeted in their approach to blighted areas.

For the City of Memphis, potholes and abandoned properties changed from unsolvable issues to trackable, workable challenges they can meet head-on. Their story serves as an example of what cities can accomplish by combining AI and ML technologies with their existing resources to optimize and modernize service delivery. This single project has had a ripple effect on the community, making Memphis a more desirable place to live. The best part? Other cities can do it too.

Learn More

To see how your city can apply AI and ML with a Google Cloud partnership to improve services and solve unique challenges, connect with one of our specialists today.

Get in touch