“Roadrunner” is a smart pothole detection and classification system designed to improve road safety and assist in infrastructure maintenance. This project integrates stereo computer vision, embedded systems, and cloud-based machine learning into a cohesive solution for identifying and responding to road surface damage.

In this system, a ZED stereo camera connected to a Raspberry Pi captures live stereo images of the road. The Raspberry Pi processes these images to generate disparity maps and detect potholes using blob and contour analysis techniques. Upon detecting a pothole, the Raspberry Pi records its GPS location using a dedicated GPS module and sends the image and location data to a Firebase database. Simultaneously, it sends a signal to a PIC18F4550 microcontroller, which activates vibration motors embedded in the driver’s seat cushion to provide immediate haptic feedback and alert the user of any danger. A server-side machine

learning model later classifies the pothole types and informs the relevant authorities for timely road repair and maintenance.