<![CDATA[The increasing number of motorcycles in developing countries has intensified parking management challenges, particularly in high-density environments with irregular vehicle arrangements. This study proposes a motorcycle parking availability detection system using the YOLOv5 object detection algorithm to address limitations of conventional parking methods. The research contribution is the development of a context-aware detection framework using a locally collected dataset and the evaluation of its performance under real-world parking conditions.The dataset consists of 1,200 images collected from campus parking areas and is divided into training, validation, and testing sets. The images were annotated into occupied and vacant classes and trained using YOLOv5 with 100 epochs. Model performance was evaluated using precision, recall, F1-score, and mean Average Precision (mAP@0.5) on a held-out test set.The results show that the model achieves an F1-score of 0.57 and mAP@0.5 of 0.566, indicating moderate detection performance in dense and occluded environments. Although a precision of 1.00 is obtained at a confidence threshold of 0.978, this condition significantly reduces recall, highlighting a trade-off between detection accuracy and coverage. The confusion matrix and recall–confidence analysis reveal that errors are primarily caused by occlusion, shadow effects, and background interference. Compared to previous studies focusing on car parking detection, this system demonstrates comparable performance while addressing the unique complexity of motorcycle parking. However, the relatively small dataset size and environmental variability limit generalization.In conclusion, the proposed system provides a feasible initial approach for motorcycle parking detection, but further improvements in dataset diversity, annotation quality, and model robustness are required to achieve reliable large-scale deployment.]]>
Copyrights © 2026
