Enhancing EfficientNetV2 for 12 Wheat Disease Classification Using Particle Swarm Optimization

Abstract

Wheat is an important crop for many people, yet vulnerable to diseases that significantly impact yield and quality. This research proposes an enhanced deep learning model for classifying 12 wheat disease types using EfficientNetV2 optimized with Particle Swarm Optimization (PSO). Unlike previous studies that focus on fewer classes, our approach utilizes a diverse dataset of over 11,167 real-field images. PSO is employed to fine-tune hyperparameters, such as learning rate, batch size, dropout, and hidden units, to improve model generalization. The EfficientNetV2+PSO model achieved 90.58% accuracy. Evaluation metrics, including accuracy, precision, recall, and F1-score, represent the improvement of the model's performance after using PSO. This study shows the effectiveness of combining PSO with EfficientNetV2 for accurate 12 wheat disease classification. It offers an effective method that can be applied to agricultural systems.