Turns out, a GoPro camera is not just for recording extreme sports. It can also zero in on potholes.
Or at least, that’s how officials in San Jose, Calif., used the devices as part of a pilot project to record the streetscape, and use AI technology to identify road hazards.
The Al-powered Road Safety Conditions Pilot, which launched in 2023, correctly identified potholes 97 percent of the time, and managed an 88 percent accuracy when watching for trash or debris in the roadways. The cameras were mounted on a parking enforcement vehicle, Stephen Caines, who serves as chief innovation officer and city budget director, and senior adviser to Mayor Matt Mahan, said.
The pilot project worked with several technology companies, including City Rover, and Blue Systems.
“We were able to get a lot of interest from folks in the private industry who wanted to collaborate on this project, and it was real exciting to see the high level of accuracy we received,” Caines said.
The next phase of the initiative will expand the scope to a wider area of study, and will mount the cameras on street sweepers, offering a more road-level view, taking in travel lanes as well as bike lanes and sidewalks. Funding for Phase 2 will come, in part, from a grant by the Toyota Mobility Foundation.
The project will establish what city officials describe as “the nation’s first open-source AI training database,” Caines said. San Jose was a founding member of the GovAI Coalition, which aims to serve the public sector in developing policies and procurement practices to effectively vet and deploy AI tools. Formed in 2023, the coalition is a network of some 850 government agencies and 2,500 members, including public- and private-sector workers.
An open source AI database allows San Jose to share its own learnings from the pilot with other cities.
“In order to really reap the benefits of AI we need to have good data governance practices in place,” Leila Doty, a privacy and AI analyst with the city, said during a webinar panel in August, organized by the National Academies of Sciences, Engineering, and Medicine.
“If we want to actually use AI effectively, we need to get these foundational, core data governance practices in place,” she said.
By expanding the project to include bike lanes and sidewalks, city officials hope to get a better understanding of some of the larger issues impacting roadway safety.
“I would say that traffic safety has been a very pervasive issue in San Jose,” Caines said, as he recalled a recent downtown scooter accident which resulted in a fatality. “And just at a basic level, sidewalk maintenance has been a consistent issue.”
The system is not yet connected to the city’s 311 system, but the plan is to link them — allowing problems identified to initiate a 311 request, which can then be addressed by the appropriate department.
“At the moment we are focusing primarily on accuracy and also having the background conversations with departments as to what exactly they want to detect, and why,” Caines said. “We are taking on administrative and workflow elements first, and then we’ll transition into the technical integration into our back end, where then, yes, it would dispatch a ticket, and then there would be a city worker that would be dispatched specifically.”
San Jose to Expand AI-Powered Road Safety Pilot
The effort to identify road hazards and potholes with vehicle-mounted GoPro cameras and AI has been so successful that the city is planning to deploy the tech on more of its fleet vehicles in more areas.
