YOLOv3 Algorithm with additional convolutional neural network trained for traffic sign recognition

The ability of perception and understanding all static and dynamic objects around vehicle in various driving and environmental conditions represent one of the main requirements for autonomous vehicles and most of Advanced Driving Assistance Systems (ADAS). Current promise to deliver safe ADAS in modern cars could be achieved by convolutional neural network (CNN). In this paper we present a software based on YOLO that is extended with a CNN for traffic sign recognition. Since real time detection is required for safe driving, YOLO network used in this paper is pre trained for detection and classification of only five objects which are separated in categories such as cars, trucks, pedestrians, traffic signs, and traffic lights. Detected traffic signs are further passed to CNN which can classify them in one of 75 categories. We demonstrate the high level of classification confidence by accurately recognition more than 99.2% of examined signs in quite diverse conditions.