LEVERAGING DEEP LEARNING FOR ACCURATE RICE LEAF DISEASE RECOGNITION

Abstract

: Imagine a Rice Field, seemingly healthy, yet harboring invisible foes. This paper delves into the world of crop disease detection for the Tan spot, Leaf blight, Sheath decay, Bacterial leaf rot, and False smut Dataset consists of 3546 images. The image segmentation is done by BIRCH clustering followed by GMM clustering; various texture features have been extracted through this. The classification is done using existing models such as VGG16, VGG19, RESNET50, and INCEPTION V3. After comparing various models and validation, the observation was made that VGG19 performs well compared to other models. The overall accuracy obtained for the VGG19 is 95.88.%. The achieved accuracy surpasses that of conventional backpropagation neural network models, indicating significant advancements in crop disease diagnosis. This study introduces a novel approach that opens avenues for future research in the field of deep learning for crop disease diagnosis.