Rice plant disease diagnosing using machine learning techniques: a comprehensive review

Abstract

Article Highlights When identifying rice plant diseases through machine learning models, many of the studies have focused only on fewer number of diseases due to the lack of datasets available in the literature. There are some diseases that cannot be identified by just observing or scanning the external surface of the rice leaf. Those kinds of diseases may occur due to pests who are attacking the rice plant internally. In these kind of situations, it is impossible to identify these diseases in the initial stages of spreading. According to the literature review which was conducted by the authors, they were able to identify some common and popular CNN models which have been used to identify rice plant diseases which are known as VGG 16, VGG 19, MobileNet, LeNet5 and ResNet 50. According to the studies, researchers have highlighted that they was able to reach 77.09%, 76.63%, and 76.92% training accuracies for VGG-19, LeNet5, and MobileNet-V2 respectively. When comparing the machine learning models which has been used for the identification process of rice plant diseases in the past studies, it has been identified that ResNet 50 is the best suited model which has given the highest accuracy for Rice plant disease detection.

Keywords

• Deep learning approaches
• Hybrid systems
• Convolutional neural networks
• Plant disease diagnosis
• Machine learning