Automated Detection and Classification of Rice Leaf Diseases Using Hybrid Deep Learning CNN

Abstract

Rice (Oryza sativa) is one of the most widely cultivated crops in the Philippines. However, several diseases affect this crop, significantly reducing its quality and production. It is crucial to detect them early to stop the spread of infections. “This study proposed a deep hybrid learning CNN with transfer learning for detecting and classifying rice leaf diseases, namely bacterial leaf blight, brown spot, and leaf smut, with 40 images each”. The dataset was trained after the data pre-processing stage. A new model obtained 88.58% and 91.67% accuracy in the training and validation sets, respectively. Its performance metrics were analysed and evaluated based on the confusion matrix and classification report, such as “accuracy, recall (specificity), precision (Positive Predictive Value), and F-1 score”.