Bacterial Leaf Blight Identification of Rice Fields Using Tiny YOLOv3

Abstract

Rice plantations are frequently affected by various rice diseases, one of which being bacterial leaf blight. Although there are scientific methods for determining bacterial blight using various molecular techniques, these tests are frequently more suitable for specific reasons such as genome identification rather than broad applications due to the same effect of bacterial blight. As a result, image processing techniques such as Convolutional Neural Network (CNN) are commonly utilized for general rice disease identification due to their reliability. The purpose of this research is to identify bacterial leaf blight using the Tiny YOLOv3 algorithm. With a total of 20 test photos, 10 of which were bacterial leaf blight and the other 10 were healthy, the prototype was able to predict bacterial blight infected leaves, with 19 correct predictions and one wrong prediction. During its evaluation, the model used to detect the diseases generated acceptable mean average precision and a precision and accuracy of detecting the disease of 90.91 % and 95%, respectively.