Optimizing an Object Detection Algorithm for Detecting Oil Palm Fruit Bunches and Their Ripeness

Abstract

The detection and classification of oil palm fruit bunches are critical for optimizing palm oil yield and quality. Traditional methods relying on human visual inspection are prone to errors and inconsistencies. To address these challenges, this study proposes an optimized You Only Look Once Version 7 (YOLOv7) algorithm for real-time detection and classification of oil palm fruit bunches. By leveraging a comprehensive dataset and implementing strategic fine-tuning techniques, we significantly improved the model’s performance. The optimized model achieved a classification accuracy of 92.55% and a mean average precision (mAP) of 95.08%, demonstrating its effectiveness in real-world applications. The model was integrated into web application, allowing farmers and processing facilities to assess the ripeness of palm fruit in real time, thereby improving decision-making and operational efficiency. This research highlights the potential of advanced deep learning models in agricultural applications, promising a transformative impact on the palm oil industry by ensuring consistent quality assessments and optimizing production processes.