Deep Learning-Based Classification of Plant Leaf Diseases Using Advanced Particle Swarm Optimization

Diseases impacting plant leaves significantly challenge global crop production, affecting farmers worldwide. Early diagnosis, treatment, and resolution of plant leaf diseases are essential to promoting healthy plant growth. Artificial intelligence (AI) has rapidly developed in recent years, and several applications leveraging AI have been presented. Deep learning (DL) research has excellent potential to improve performance in precision agriculture. Significant advancements and developments in the industry have created opportunities to improve the system’s accuracy and coordination in recognizing and understanding plant leaf diseases. Deep learning neural networks are employed due to their ability to address the technical challenges associated with classification problems. However, CNN faces the complexity of identifying the optimal hyperparameters for the specific architecture and the application. Hence, this paper employed an advanced particle swarm optimization algorithm to optimize CNN classifiers’ hyperparameter tuning. A novel hybrid approach, named CNN-aPSO, for classifying corn and cotton crops into distinct categories of diseases has been proposed. The work explores optimizing several CNN variants, including ResNet-50, ResNet-101, MobileNetV1, MobileNetV2, DenseNet-121, and DenseNet-169, utilizing advanced PSO to identify their optimal hyperparameters. The research highlights the significance of optimization for effective disease classification for specific pre-trained neural networks. The findings demonstrate that DenseNet-169 integrated with PSO achieves the highest accuracy of 98.04% among all other variants.