Optimized Fruit Detection in Complex Environments Using YOLOv11n for Smart Agricultural Applications

Abstract

This study explores the application of the YOLOv11 object detection algorithm for precise and efficient fruit detection in complex image environments. The research focuses on leveraging YOLOv11’s advanced single-stage detection architecture and Transfer Learning to enable real-time fruit identification while maintaining high detection accuracy, particularly for edge computing scenarios where computational efficiency is essential. To evaluate the effectiveness of the proposed system, the model was trained on a specialized fruit dataset consisting of images taken from real-world agricultural settings. The dataset comprises eleven distinct fruit classes captured under diverse conditions, including varying lighting, occlusions, and different angles. Experimental results revealed that the model achieved a mean average precision (mAP) of 90%, demonstrating its ability to localize and classify fruits with a high degree of reliability. The findings of this study have significant implications for the agricultural sector, particularly in automation-driven applications such as fruit harvesting, post-harvest processing, quality assessment, and inventory management. By integrating YOLOv11-based detection systems in the process, farms and agribusinesses can improve efficiency, reduce labor costs, and enhance productivity.