An Automatic Plants Leaf Disease Identification Method Based on A New Proposed Optimized Convolutional Neural Networks Using Symbiotic Organism Search Algorithm

Abstract

This study presents a novel method for accurately and efficiently classifying plant leaves based on different diseases and health conditions. The approach combines deep learning (DL) and optimization techniques. The model's hyperparameters are optimized using the Symbiotic Organism Search (SOS) algorithm, while a convolutional neural network (CNN) is employed for disease classification. CNNs are effective in identifying distinctive features that aid in distinguishing various classes. The Plant Village dataset is utilized, and data augmentation is performed to enhance the model's performance. The optimized CNN, which does not require segmentation, achieves a classification accuracy of 97.29% for the test set (8,777 images) and 99.89% for the training set (70,295 images). Precision and recall values of 97.33% and 97.32% are attained, respectively. The proposed method is compared to primary CNN models on the Leaf disease database, as well as two well-known pre-trained networks (VGG-16 and VGG-19), using five performance measures. The results demonstrate the superior performance and effectiveness of the developed method. This research contributes to a fast and accurate approach for leaf disease classification, with potential applications in other image classification tasks and benefits for agriculture, orchards, and plant disease identification. Furthermore, a comparison with three similar studies reveals that the proposed network achieves higher accuracy with fewer trainable parameters and computational requirements.