Vehicle Detection of Bangladesh using YOLOv7 with Hyper-parameter Tuning

In recent years, infrastructure development in Bangladesh has advanced at a remarkable rate. Road transportation has grown the most for passenger and freight transportation in the last 50 years. As a result, there are now numerous land-based transportation options. However, the traffic in Bangladesh is getting worse due to rapid unplanned expansion, diverse land use, inadequate public transport vehicles, and more private car users. In this paper, we sought to determine the application of Al in this challenging context. Object detection and classification can help distinguish different types of vehicles and determine traffic density in a given area at a given time, which can facilitate vehicle flow control. The study trains a Deep Neural Network to detect local traffic objects in Bangladesh in real-time using the object detection algorithm YOLOv7 with hyperparameter tuning. The experimental data exhibits an improved precision of up to 13.52% by tuning the hyperparameters.