Implementation of the YOLOv8 object detection algorithm for enhancing traffic surveillance through accurate identification of multiple road entities, including cars, motorcycles, trucks, and pedestrians. Using a 41-second CCTV video as the primary dataset, the research adopts a deep learning-based training approach via Google Colab to evaluate YOLOv8's performance under real-world urban conditions. The detection model was assessed using key evaluation metrics such as accuracy, precision, recall, and Mean Average Precision (mAP). The experimental results demonstrate that YOLOv8 achieves an overall detection accuracy of 80%, showing reliable performance in identifying vehicles and people despite challenges such as occlusions, varied lighting, and complex backgrounds. However, accuracy variations were observed in cases involving partial visibility and non-optimal camera angles. The findings highlight the potential of YOLOv8 as a robust and scalable solution for real-time traffic object detection, with implications for smart city development and automated traffic management systems. Further improvements are recommended in dataset diversity and model fine-tuning to enhance detection robustness across dynamic traffic scenarios
Saragih et al. (Mon,) studied this question.