A CNN-Driven Framework for Early Detection of Strawberry Plant Disease

Strawberry (Fragaria × ananassa) cultivation is increasingly challenged by plant diseases that significantly reduce yield and threaten food security. Among these, Strawberry Leaf Scorch—caused by Diplocarpon earlianum—is particularly destructive, leading to leaf necrosis, reduced photosynthesis, and eventual crop failure if left unmanaged. Traditional disease detection methods largely depend on manual inspection, which is not only timeintensive but also infeasible for large-scale farms. In this study, an artificial intelligence-driven approach is proposed using Convolutional Neural Networks (CNNs) to enable automated and highly accurate detection of strawberry leaf scorch disease. A comprehensive dataset of more than 3,400 high-resolution images was assembled, comprising both healthy and infected leaf samples. Images were sourced from open-access agricultural datasets and real-world strawberry farms across diverse geographical zones, supplemented with synthetic augmentation techniques to ensure environmental robustness. A custom CNN architecture was developed, enhanced through transfer learning using pre-trained ResNet50 and VGG16 models. The training process leveraged advanced strategies, including adaptive learning rate tuning, dropout regularization, and aggressive data augmentation, to enhance generalization. The model achieved an impressive classification accuracy of 98.10%, outperforming classical classifiers which encompasses Support Vector Machines (86.2%) and Random Forests (82.7%). To enhance transparency and trust in the system, explainable AI methods such as Grad-CAM, feature visualization, and LIME were utilized, highlighting the regions of the leaf influencing the model's predictions. This work presents a scalable, cost-effective, and user-friendly solution for early disease detection in strawberry farming, with the potential to reduce crop loss by up to 25%. Future extensions of this system include a user-friendly web interface that enables farmers to upload leaf images for real-time diagnosis, drone-based image capture, mobile app deployment, and IoT integration for continuous, automated monitoring in smart technology-driven agricultural environments.