Real-Time Vehicle Detection using YOLOv8 and Data Augmentation Approach

摘要

There are several uses for object detection in computer vision, including surveillance, robotics, and autonomous vehicles. You Only Look Once (YOLO) is one of the most widely used object identification methods, and YOLOv8 is becoming increasingly popular for computer vision applications. This study focuses on improving vehicle detection accuracy using YOLOv8 by customizing and optimizing the dataset. The dataset was initially small, but data augmentation principles and an increase in the number of training images improved its robustness. The YOLOv8s model was used, and the training of each dataset was conducted for 75 epochs, resulting in a 23% increase in [email protected] with the final dataset. The experiment evaluated the accuracy improvement achieved by training the modified dataset using the bag of freebies technique (data augmentation principles). The work aims to enhance the accuracy of vehicle detection, particularly for small and distant vehicles, contributing to efforts towards safer and more efficient transportation systems.