Raspberry Pi Hailo accelerated YOLO

This example project was created in June 2025 to explore how a Raspberry Pi can power AffectiLink’s vision preprocessing while offloading heavy inference to the Hailo AI Hat+. This project captures high-quality video from a Pi Camera v3 using Picamera2 and Libcamera, and routes each frame through HailoRT for YOLO object detection, keeping the onboard CPU available for other work.

Under the hood, the pipeline initializes a Picamera2 stream with Libcamera to leverage the camera’s autofocus features if available, then forwards frames to the Hailo 8 accelerator via HailoRT and your chosen YOLO model and label set. Detected bounding boxes and labels are overlaid in real time with OpenCV, producing a smooth, low-latency preview window.

This project fills a gap by providing a simple, end-to-end example of running YOLO on a Hailo AI Hat+ with the Raspberry Pi Camera. Something I couldn't find in the official documentation or examples. It serves as a foundation for building more complex computer vision applications on Raspberry Pi with Hailo.