Traffic light Detection using OpenCV and YOLO

Traffic lights detection importance is increasing every year as the world is focusing on self-driven cars. Traffic lights detection is a vital task of any autonomous vehicles to ensure safety and effectiveness. In this paper, two models are built to detect traffic lights using images and videos. A dataset is created to contain data collected from Bahrain streets. Both models are evaluated and compared to introduce the most effective one. The first model is based on analyzing still images while the second model is used to analyze video frames using YOLO algorithm weights and OpenCV library. It was found that the second model was able to detect all traffic lights in video frames with high accuracy as compared with first model that analyzes still image.