Benchmarking YOLO-based models for palm oil tree detection: an in-depth analysis from YOLOv5X to YOLOv11X

This study evaluates the performance of different YOLO (You Only Look Once) models in detecting palm oil trees, a critical task for sustainable plantation management. Among the tested models, YOLOv5x demonstrated the best balance between speed and accuracy, achieving an F1-score of 93.8% and a detection time of 34 seconds, making it well-suited for real-time applications in large-scale plantations. In comparison, YOLOv10x, with a slightly higher accuracy (F1-score of 95.5% and mAP50 of 98.7%), was found less practical for time-sensitive scenarios due to its slower detection speed. The study highlights the potential of YOLOv5x to streamline plantation management tasks, such as yield estimation, health monitoring, and tree counting, significantly reducing reliance on manual labor and enabling timely responses to pest and disease outbreaks. Additionally, the versatility of YOLOv5x extends beyond palm oil plantations to broader forestry applications, including growth detection, carbon stock assessment, and ecosystem monitoring. By providing robust, efficient, and scalable solutions, this study underscores the role of advanced AI-driven detection models in promoting sustainable practices in forestry and agriculture, addressing critical challenges in resource management and environmental conservation.