Explainable AI for Multi-Granular Paddy Disease Identification and classification via a Superb Fairy-Wren Optimized Shuffle Transformer

Abstract

Significant obstacles are presented by the recent increase in paddy leaf diseases, highlighting the necessity of targeted study and quick adoption of an AI method for crop leaf disease detection. Paddy, a staple meal for more than half of the world's population and a major component of many different cuisines, has many health advantages but is hampered by conditions like brown spot and blast disease. Accurate classification is necessary for managing paddy leaf disease effectively. Therefore, in this research we introduced a novel research framework for accurate predictions and classification of multiple plant leaf disease. In this proposed methodology there four phases, preprocessing, segmentation, feature selection and optimized detection and classification. Initially the raw input images are fed into the preprocessing phase to perform some initial phases. After the completion of preprocessing the, the affected portions are segmented. With the help of TransU-Net approach the segmentation is performed. The essential features are selected using Sine-Cosine Harris Hawks Optimization (SCHHO) algorithm. Finally the multiple paddy leafs are classified using a novel shuffle transformer. To enhance the prediction performance even more the classification approach parameters are fine-tuned using Superb Fairy-wren Optimization Algorithm (SFOA). These findings demonstrate the model's ability for real-time implementation in agricultural applications, offering small-scale farmers a dependable and effective option. This study provides a framework for tackling different crop diseases and emphasizes the importance of combining thermal imaging and deep learning to improve crop disease management.