Vehicle Detection Using a Deep Learning Method

Abstract

The use of deep learning-based object detection methods in computer vision is covered in this thesis. The primary objective is the real-time detection and identification of various types of vehicles found in Bangladesh. In the present One of the key prerequisites for applications like traffic monitoring and autonomous vehicles is vehicle detection. Bangladesh presents extra difficulties because to the unpredictable traffic, the wide variety of vehicles, and the dearth of a robust dataset. On the "Vehicle Detection System Dataset," various neural networks were trained utilizing YOLOV5, YOLOV6, and YOLOV7 for the Detection. The model was trained using Google Colaboratory, a cloud-based platform, using YOLOV5, YOLOV6, and YOLOV7. Written with Python 3.10.4, the codes. The dataset for training was organized using Rob flow, an online-based computer vision technology. Using a smartphone camera, data was collected. The model with the most effective performance and the most promising results was YOLOV7. YOLOV7 has an extremely remarkable recall rate and precision. which is sufficient to find vehicles. In comparison to the other two models, it was also able to detect vehicles with greater accuracy. This study has a lot of potential and can be seen as a development for autonomous vehicles and smart traffic systems in Bangladesh.