Abstract:
Lemon leaf canker, a bacterial disease affecting lemon cultivation globally, requires efficient detection methods to prevent economic losses andcropdamage. This study evaluates the performance of four advanced YOLOmodels—YOLOv5, YOLOv8, YOLOv10, and YOLOv11—for automateddetection of lemon leaf canker using a dataset of 884 annotated imagesrepresenting healthy and diseased classes. The dataset was preprocessedandaugmented to improve robustness under diverse conditions. The models weretrained and evaluated on metrics such as precision, recall, mean AveragePrecision (mAP), and inference time. YOLOv11 emerged as the best-performingmodel, achieving a precision of 94%, a recall of 90.4%, and a mAP50-95of
73.5%, indicating superior accuracy and computational efficiency. These resultsunderscore YOLOv11's potential for real-time application in precisionagriculture, enabling early disease detection and timely interventions. Byreducing reliance on labor-intensive methods and excessive chemical
treatments, the proposed approach supports sustainable farming practices, improves crop management, and mitigates environmental impact. Thisresearch demonstrates the capability of deep learning-based solutions likeYOLOv11 to advance agricultural productivity and contribute to economicstability for farmers, while promoting environmentally conscious agricultural
practices.
