Deep Learning-Based Hybrid Model For Severity Prediction of Leaf Smut Rice Infection

Abstract

Conventional rice crop disease prediction models show some drawbacks, such as the expensive cost of acquiring the input data necessary to run the model, the absence of spatial information, or the shortage of high-quality datasets. These problems are discussed in this work, which also develops a yield prediction fusion model. Convolutional neural networks (CNN) and support vector machines make up the prediction model (SVM). In this work, Leaf smut infection of rice health is discussed. The infected plant's pictures are first collected through secondary sources. The deep learning method's best characteristic is the feature extraction and classification of the different levels of blight infection severity is done using CNN and SVM. Mild, Average, Severe, and Profound are the four severity projection levels used in the study. Kaggle etc. are the data repositories that were utilized, and the total size of the dataset was 272. The suggested approach produces four severity-level predictions with 98% accuracy.