An Explainable Support Vector Machine (E-SVM) for plant disease detection using transfer learning

Plant diseases greatly affect global agricultural yield and quality, leading to economic losses and food insecurity. Manual diagnosis is slow and often inaccurate, motivating the need for automated solutions. This paper presents an Explainable Support Vector Machine (E-SVM) model integrated with transfer learning for accurate and interpretable plant disease detection. Deep feature representations are extracted from pre-trained convolutional neural networks (ResNet-50 and VGG-16), and these are used to train a modified SVM classifier. Explainability is introduced through SHAP and LIME frameworks to visualize how individual features influence decisions. Experiments conducted on the PlantVillage dataset demonstrate an accuracy of 99.12%, outperforming CNN-only and handcrafted SVM models. The proposed approach combines strong classification performance with transparency, enabling greater trust and real-world usability in smart agriculture.