Potholes are a pervasive road hazard, often causing accidents and posing risks to drivers. Early identification and timely repair are critical for accident prevention. However, existing manual methods for detecting potholes are labor-intensive and inefficient. To address this, we propose a real-time detection system using the YOLOv9 algorithm combined with an ultrasonic sensor attached to a vehicle’s suspension. The sensor detects vibrations caused by potholes, and a Node MCU transmits the data when significant fluctuations occur. A dashcam captures street footage, which is processed by YOLOv9 for pothole identification. A dataset of over 4000 pothole images, including those from IIT Goa, was used to train the model. After annotation, 77.25% of the images were allocated for training, 21% for validation, and 1.75% for testing. This approach aims to enhance real-time detection for improved road safety.