Smart Road Planning

Effective vehicle detection has become a progressively significant tasks in modern life. In this project, we propose a system to detect & track vehicles with speed estimation using deep learning and OpenCV. We will be using MobileNet V2 SSD (Single Shot Multibox Detector) Caffe model, which was pre-trained on Coco dataset. Our system detects vehicles using deep learning Caffe model from video sources, and tracks them. Using OpenCV, the system counts the number of cars who are heading “Left” and “Right” in the roads, and calculates speed estimation to detect the KPH (Kilometers Per Hour) of the moving vehicle. VASCAR method is the base of our speed estimation. Moreover, we have developed a system to simulate the traffic light based on the data provided by different vehicle on the road. We have saved the trained model and implement of Dropbox API & Google Co-lab for storing the summarized results. Finally, we have gained train accuracy of 81% from the latest CNN model of MobileNet V2 SSD.
Department Name
Electrical and Computer Engineering
North South University
Printed Thesis