Enhancing Plant Pathology with CNNs: A Hierarchical Approach for Accurate Disease Identification

Abstract

The accurate diagnosis and classification of plant diseases are pivotal for enhancing agricultural yield and quality. While this task presents challenges, particularly for those without specialized knowledge, the advent of Deep Learning (DL) has introduced promising solutions. This research presents a state-of-the-art Convolutional Neural Network (CNN) model, specifically designed for precise and efficient plant disease classification. Leveraging the capabilities of DL, our model accurately identifies subtle disease manifestations, ensuring exceptional classification accuracy. Notably, the model achieved a remarkable accuracy of 99.7%, with specific performances of 99.6% on the PlantVillage dataset and 99.8% on the PlantDoc dataset. This success can be attributed to the model's unique architecture, comprehensive hyperparameter optimization, and strategic use of transfer learning and data augmentation techniques. Moreover, our model adeptly handles the complexities of disease symptoms, discerning even minor anomalies on leaves under challenging image conditions. Looking forward, we aim to refine the model's capabilities, potentially extending its diagnostic reach to other plant components, such as stems and flowers, thereby offering a comprehensive and intuitive approach to plant disease classification.