This project showcases a real-time traffic detection system built on low-power, edge-based hardware. It leverages a Raspberry Pi 5 paired with a Hailo-8L AI accelerator to run a YOLOv object detection model efficiently.
A camera continuously captures live video from a roadway or intersection. The video stream is processed entirely on the Raspberry Pi, with the Hailo-8L providing hardware acceleration for neural network inference. The YOLOv model identifies objects such as cars, trucks, bicycles, and pedestrians in real time.
All processing occurs locally on the device—no cloud services or external servers are required. This ensures fast, private, and reliable traffic analysis, even in remote or low-connectivity environments.
Importantly, no images or video are recorded or stored, either locally or in the cloud. The system operates purely on live video for real-time detection, ensuring a strong level of privacy and data security.
Please note: the system relies on a clear visual feed and functions best during daylight hours. Its performance may be reduced in low-light conditions or when visibility is obstructed by adverse weather, such as fog, heavy rain, or snow.
The goal of this project is to demonstrate how affordable, off-the-shelf components can power practical computer vision applications like traffic monitoring. The system is flexible and can be adapted for use in data collection, research, or educational settings.