Rice Sheath Rot Disease Detection and Severity Classification: A Novel Framework Leveraging CNN-LSTM Models for Multi-Classification

The rice sheath rot disease presents a substantial risk to the cultivation of rice, necessitating the prompt and precise identification of the illness and assessment of its severity. In this study, we propose a novel framework that combines Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks, thereby creating a hybrid deep learning (DL) model. The presented model has exceptional proficiency in capturing spatial as well as temporal information, thereby efficiently addressing the complex dynamics associated with the advancement of rice sheath rot disease. By utilizing an extensive dataset obtained from various places and growth stages, we diligently annotated and enhanced the data. The model exhibits outstanding performance, as evidenced by its overall accuracy of 94.84%, which showcases its proficiency in the identification of diseases and classification of their severity. This study not only emphasizes the revolutionary capacity of our hybrid methodology in precision agriculture but also provides farmers and stakeholders with a powerful tool for optimizing resource allocation and reducing crop losses. In addition to addressing rice sheath rot, our framework presents an adaptable solution that can be used for a broader spectrum of plant diseases, facilitating the advent of a new age in sustainable agriculture driven by data analysis.