Get Full Government Meeting Transcripts, Videos, & Alerts Forever!
Get email alerts on the Infrastructure Technology topic
No spam. Unsubscribe anytime.
Memphis Public Works says AI system identified 6,600 potholes since 2022; department moving to clogged‑inlet and high‑grass detection next
Summary
Public Works reported the city is operating an AI/ML pothole detection program that has identified more than 6,600 potholes from vehicle‑mounted cameras since 2022; the system sends images to Google Cloud for processing and staff verify detections before creating service requests.
Memphis Public Works told the Public Works, Solid Waste & General Services Committee that an artificial intelligence/machine‑learning system is now operating citywide to detect potholes and other infrastructure needs from cameras mounted on city vehicles.
Director Scott Morgan said the program began with pilot phases in 2019–2022 and is now in operations with five vehicles in service and two more camera rigs due in the coming months. Video captured by vehicle‑mounted cameras uploads to a Google Cloud Platform instance where AI models analyze footage and flag likely potholes, clogged drainage inlets, broken curbs and high grass. The system returns images and a confidence rating; Public Works staff review the images and either approve or reject each detection before creating a service request.
From 2022 to the present, Morgan said the system identified roughly 6,600 potholes; about 4,900 detections were rejected by staff after review, leaving roughly 1,700 that resulted in crews filling potholes, with 520 referred to the Tennessee Department of Transportation as state‑owned issues. Morgan said the AI’s performance improves as the verifier feedback trains the model.
Public Works also reported its 2024 field crews filled more than 83,000 potholes through routine operations and service requests combined. The department said future steps include expanding AI detection for clogged inlets, damaged curb and gutter and high‑grass detection and integrating additional cameras on street sweepers.
Council members asked about 311 integration and rejected detections; Public Works said the AI system generates service requests internally and that verified detections are pushed into the city’s work order queue and become visible to 311. If a detection is rejected by staff, crews will still be dispatched to verify on site if the image is ambiguous, and residents can still call 311 for a follow‑up.
