Real-Time Implementation of Mini Autonomous Car Based on MobileNet - Single Shot Detector
Buse Pehlivan, Ceren Kahraman, Deniz Kurtel, Mert Nakıp, Cuneyt Guzelis · 2020 Innovations in Intelligent Systems and Applications Conference (ASYU) · 2020
In this paper, in order to realize a prototype of an autonomous vehicle, we present a framework that consists of convolutional neural networks and image processing methods. The study is comprised of two main parts as software and hardware. In the hardware part, a small-sized smart video car kit is used as the prototype of the autonomous car. This programmable tool consists of Raspberry Pi, servo motors and a USB webcam whose angle of vision is equal to 120°. In the software part, we propose an algorithm in which we use Convolutional Neural Networks to detect the objects (vehicles, pedestrians, and traffic signs) and Hough transformation to detect the road lanes. Based on the outputs of the object and lane detections, the system decides the speed and the direction of the car in real-time. In our results, the vehicle performs autonomous driving in the scaled real-world application.