Vineyard Leaf Disease Prediction: Bridging the Gap between Predictive Accuracy and Interpretability

Balancing the accuracy and interpretability of predictive models has been a persistent challenge in traditional approaches. In this study, we advance this field by integrating cutting-edge artificial intelligence (AI) techniques with Explainable AI (XAI) methodologies to significantly enhance both the accuracy and interpretability of vineyard leaf disease predictions. We employ state-of-the-art convolutional neural networks (CNNs) and introduce a fine-grained model architecture featuring, adept at discerning subtle disease indicators in vineyard leaves. This innovative approach not only boosts the diagnostic performance of the models but also provides clear visualizations of the decision-making processes. This study utilizes a focused dataset strategy, incorporating one specialized grape disease dataset (Esca) and a subset of the general PlantVillage dataset, specifically selecting categories relevant to Apple and Grape diseases. The obtained results have demonstrated our model’s exceptional capability in accurately identifying and classifying various leaf diseases, showcasing its practical applicability in real-world vineyard management. Furthermore, our approach addresses the vital need for transparency and trust in AI applications within agriculture, particularly in viticulture.