Revitalizing Plant Disease Detection using Optimized DNNs for Enhanced Leaf Images

Leaf diseases can significantly impact plant health and crop yields. They can spread rapidly, leading to defoliation and reduced photosynthesis, affecting overall plant growth and productivity. Effective disease management strategies are crucial to minimize the economic and environmental consequences of leaf diseases. The productivity of agriculture greatly benefits from the automated identification and categorization of leaf diseases. Deep Neural Networks (DNNs) has shown great potential in detecting leaf diseases. The effectiveness of EfficientNet, ResNet, and DenseNet three well-known Convolutional Neural Network (CNN) architectures, for classifying leaf diseases is examined in this study and the comparative effectiveness is demonstrated. This study makes use of a large collection of annotated leaf pictures, enhancing the training data's quality. The study used large collection of annotated leaf pictures which have been pre-processed by utilizing the adaptive median filter to improve the quality of training data. Additionally, image enhancing methods like resizing, contrast adjusting are used to enhance the appearance of disease-related patterns. The classification of the leaf disease achieved a remarkable accuracy of 99.80 percent utilizing the EfficientNet architecture. Performance evaluation includes the use of confusion matrix analysis to gain insights into classification errors and potential misclassifications.