<![CDATA[Oil palm represents one of Indonesia’s principal commodities. Traditionally, farmers manually monitor the ripeness level of palm oil, but this method is neither effective nor efficient for large-scale harvests. Therefore, a system that can automatically detect the ripeness level of fresh fruit bunches (FFB) is needed. In this study, the YOLOv8 algorithm was used which was integrated into a web-based application. The system is designed to improve accuracy and efficiency in the grading process of oil palm fruits, which directly impacts the quality of processed products and palm oil production. The dataset used consists of 6.592 images obtained through the Roboflow platform, covering various ripeness categories. The system development follows the CRISP-DM approach, consisting of business understanding, data understanding, data preparation, modeling, evaluation and deployment. The model training process approximately 3,1 hours, with evaluation results showing a precision of 94,5%, recall of 94,7%, and a mean Average Precision (mAP) of 98%. The model’s performance is further supported by an F1-confidence curve of 95% and a precision-recall curve of 98%, indicating stable and accurate classification capabilities. The model is deployed through a Streamlit-based web interface, allowing users to perform real-time detection from images or videos without requiring additional installations.]]>
Copyrights © 2025
