Abstract Pomegranate is a fruit of substantial economic and nutritional importance worldwide. However, its production and quality are significantly threatened by a range of diseases, whose early and accurate detection remains a challenging task. Traditional disease identification relies primarily on manual inspection by agronomists, a process that is time-consuming, labor-intensive, and often leads to delayed intervention, resulting in considerable crop losses. To address these limitations, this study proposes a dual-model deep learning framework that integrates advanced deep learning techniques with iterative refinement for both disease classification and detection, leveraging the YOLOv11 architecture. A key contribution of this work is the development of a comprehensive Pomegranate Fruit Diseases Dataset, consisting of 5,099 carefully annotated images spanning five classes: Alternaria fruit spot, Anthracnose, Bacterial blight, Cercospora fruit spot, and healthy fruits. In the classification module, a deep learning–based model enhanced through iterative refinement is employed to achieve robust disease recognition. The model is extensively optimized using advanced hyperparameter tuning and progressive learning strategies. Experimental results demonstrate strong classification performance, achieving an F1-score of 0.988 for Cercospora fruit spot and 0.950 for Anthracnose. In the detection module, YOLOv11 is adopted to accurately localize and identify disease regions within pomegranate images. Model performance is further improved through systematic data augmentation, anchor box optimization, and fine-tuned hyperparameters. Quantitative evaluation confirms the effectiveness of the proposed detection framework, yielding a precision of 0.9253, recall of 0.8558, AP@50 of 0.9453, mAP@50–95 of 0.8456, and an overall fitness score of 0.8556. These results highlight the capability of the proposed YOLOv11-based system to accurately detect and classify both common and complex pomegranate disease patterns, offering a reliable and efficient solution for precision agriculture and early disease management.
Shams et al. (Wed,) studied this question.