Oryza Sativa Disease Identification Using Image Processing and Pattern Recognition by Deep Learning Algorithms-

Abstract

Rice (Oryza sativa) is the major food of over 50% of the global population and serves for sustenance security as well as economic stability in many parts of Asia and Africa. However, rice cultivation faces serious risks from a number of diseases, including rice blast, brown spot, and bacterial leaf blight, which can cause yield losses of 20-30% annually. Traditional manual based visual disease detection is time consuming, subjective and not suitable for large scale farming. This study proposes an intelligent visual disease detection framework based on deep learning to automatically identify rice plant diseases. The research presents an extensive comparative study of Convolutional Neural Networks (CNN) vs. Swin Transformer architectures by utilizing advanced image preprocessing, data augmentation methods, and transfer learning approaches. We experimentally verify that the Swin Transformer model achieves an impressive performance (97.0%), surpassing the popular CNN approach (94.0%) by a large margin of 3.2%. Healthy leaves, brown leaf spot, blast, and bacterial leaf blight are the four major diseases that the system correctly distinguishes. With a useful and precise decision-support system for early disease detection and prompt treatment, this study will advance precision agriculture and have the potential to transform areas with scarce agricultural resources.