A robust approach for olive leaf disease detection in uncontrolled environments

Detecting diseases in olive leaves is crucial for maintaining tree health and ensuring stable olive production. Early signs of infection often appear on the leaves, making them a key indicator for timely disease detection and intervention. Traditionally, farmers rely on visual inspection or laboratory tests to diagnose plant diseases. However, recent advancements in deep learning (DL) have significantly improved the accuracy and efficiency of olive leaf disease diagnosis. Numerous studies in the literature have explored this task using CNN-based architectures and, more recently, Vision Transformers. While these models have shown promising performance on benchmark datasets, they are often trained and validated on controlled data curated by researchers, which limits their generalizability to real-world agricultural conditions. To address this gap, this paper introduces a hybrid approach that integrates leaf segmentation using SAM, feature extraction with EfficientNet, and disease classification via the Swin Transformer, enhanced with visual explanations using LIME, SHAP, GradCAM, and GradCAM++. Unlike previous studies, which have been restricted to controlled environments, the proposed method is designed for real-world applications. It was tested on multiple datasets of olive trees from Tunisia, Lebanon, and Turkey, achieving an accuracy of 98.2%.