Disease Detection on Rice Leaves through Deep Learning with InceptionV3 Method

Abstract

The rate of growth in the agricultural sector in Indonesia puts demands on people who work as farmers to maintain and improve the quality of agriculture. Rice, which is one of the basic needs of the community, is currently the most in demand. Therefore, the need for rice continues to increase from year to year with the increase in the population of Indonesia. To maintain the quality and quantity of rice, it is necessary to monitor continuously which for developing countries, there are limited tools and costs to develop technology to deal with problems of maintaining rice quality, especially diseases in rice. Rice disease is influenced by various factors, some of which are season, weather, temperature, media, availability of water sources, and others. The purpose of this research is to prevent diseases in rice from spreading and spreading by making disease detectors in rice through a deep learning approach using the InceptionV3 method. There are 4 classes of rice diseases diagnosed, namely bacterial blight, blast, brown spot, and tungro. The total loaded dataset is 5932 images used in this study. The InceptionV3 model used can learn hidden patterns in the image thanks to the CNN transfer learning method technology with an accuracy of 97.47%. The results show that InceptionV3 can be one of the choices of various existing CNN methods because of its accuracy.