Low-Cost Embedded Multi-Object Detection Using Deep Neural Network

This paper presents a new method for detecting both vehicles and traffic signs, simultaneously. The proposed method is designed for self-driving cars that need low-cost and low processing power embedded systems like Raspberry Pi. Also, high-performance object detection algorithms of deep learningare hired. This paper uses three models: SSD MobileNet v۱, YOLO v۳, and YOLO v۳-tiny that perform well on embedded computers. The algorithm is implemented on ۹۰۰ images of the GTSDB dataset for learning traffic signs and ۱۱۵۳۰ images of the in a real environment using two Raspberry Pi by the Raspberry Pi camera module v۲. The speed of the proposed system is ۱.۴ frames per second at a high accuracy level

Keywords:

Self-driving car , Traffic signs detection , Vehicle detection , YOLO , Embedded