A Hybrid EfficientNetV2S–ResNet50 Model for the Accurate Classification of Plant Leaf Diseases

Abstract

Accurate classification of plant leaf diseases is significant for precision agriculture. This study details the development and evaluation of a hybrid Deep Learning (DL) architecture integrating EfficientNetV2S and ResNet50 to balance accuracy and computational efficiency. The original 8,000-image PlantVillage dataset was augmented to 28,000 images to address class imbalance and enable robust training. The dataset was partitioned into 20,000 images for training, 4,000 for validation, and 4,000 for testing. The hybrid model successfully leveraged the complementary strengths of both components, achieving a final classification accuracy of 97.16%, thereby demonstrating its suitability for real-time agricultural monitoring. This work confirms the potential of feature-fusion strategies for practical, early-stage disease screening (healthy versus diseased).