Analysis of Rice Leaf Diseases with Deep Learning

Abstract

– In middle-income countries, access to food is gaining importance. The inability to obtain the desired harvest from rice products, which are widely produced in countries where access to food is difficult, affects the farmer, the consumer and the country's economy. In order to reduce this impact, technologybased solutions should be integrated into agricultural production. Rice, which feeds a large part of the world's population, especially in Asian countries, can be grown throughout the world. However, rice production has been struggling with many difficulties for centuries due to various reasons such as bacterial, viral and virus. Diseases in rice product can occur in very different structures such as stem, root, leaf and stem. In this study, a new 13-layer Convolution Neural Network (CNN) model is proposed to classify the diseases in rice leaves. In order to compare the effectiveness of the proposed model, a comparison was made using a basic model consisting of two different transfer learning-based architectures named Inception and ResNet. As a result of the analyzes made, the transfer learning-based InceptionResNetV2 model gave more successful results than the proposed CNN model during the training phase. However, when the validation results were examined, the performance results of the proposed CNN model and the InceptionResNetV2 model were similar. In fact, the proposed CNN model gave slightly less losses and provided a better result than a model with a high amount of parameters such as InceptionResNetV2.