Implementation of Campus Pedestrian Detection Using YOLOv5

This study focuses on the development of accurate human detection using the YOLO open-source deep learning framework derived from OpenCV. The aim is to implement campus pedestrian detection. YOLO is a rapidly evolving and popular detection algorithm in recent years. The campus pedestrian detection system can improve students’ safety, contributing to the creation of a beautiful campus. This paper begins with a requirement analysis, which involves categorizing campus image information into pedestrians, traffic signs, fences, cones, fire hydrants, speed bumps, manhole covers, etc. A YOLOv5 dataset is created for training the model. The training process includes adjusting the model based on the training situation and employing data augmentation techniques such as mosaic enhancement, flipping, and grayscale to balance the richness of data labels and improve the recognition performance of the model. The experimental goal is to achieve a recognition rate of 70% to 80%. Overall, this research proposes a project design for campus pedestrian detection using the YOLO framework. By effectively utilizing YOLO’s fast and efficient capabilities, the project aims to enhance pedestrian safety, contributing to the establishment of a beautiful campus environment.