A Proposed Deep Learning Approach For Detecting Rice Leaf Disease

Abstract

Rice leaf diseases are a type of plant diseases affecting the leaves of rice crops, causing various kinds of crop damage. These diseases may have major financial effects, hurting both rice quality and production. Using a dataset taken from Kaggle, this paper gives an in-depth review of various deep learning methods for the identification of rice leaf diseases. The dataset contains four target attributes: Brown Spot, Bacterial Blight, Blast, and Tungro, using image counts of 2368, 1820, 1584, and 1440, respectively. InceptionV3, ResNet101, ResNet50, VGG19, CNN01, and CNN02 are among the algorithms being tested. With a result of 99.10% accuracy, our proposed CNN01 comes out as the highest performer, showing its ability in capturing difficult illness patterns. InceptionV3 and CNN02 perform effectively, with 99.06% and 98.96%, respectively, showing the efficiency of deep residual networks. VGG19, ResNet50, and ResNet101 had lower accuracies, indicating that they may be limited in their capacity to identify complex characteristics. The findings aid in making informed decisions about the algorithm to use according to the differences between accuracy, clarity, and processing resources.