Classification of Rice Leaf using RCNN with Transfer Learning

Bangladesh, as one of the 10 leading rice suppliers and users in the globe, heavily relies on grain to power its economic system and meet its food demand. Rice is one of the world’s most common foods. However, rice production is impeded by a variety of crop illnesses. One of most prevalent paddy diseases is leaf disease. Recognizing leaf illnesses is time-consuming as well as difficult for farmers in remote areas due to the lack of expertise. Despite the presence of experts in some areas, illness detection is achieved by the recognition through human eye, which may result in incorrect diagnosis and requires a lot of effort. The use of an automated process can help to solve these problems and hence having an Artificial Intelligence (AI) system is mandatory. In this study, an AI approach to detect four prevalent rice leaf diseases such as, Blast, Sheath Blight, Tungro, and Brownspot is provided. The input was clear photos of affected rice leaves on a white backdrop. The datasets were getting trained on using a variety of Machine Learning methods after appropriate preprocessing. When applied to the test datasets, the proposed approach attained an accuracy of over 97% after 10-fold cross validation. Moreover, to preserve the rice plants’ healthy and suitable growth, it is vital to detect sickness and administer the needed therapy to the injured plants. Therefore, pesticides and/or fertilizers have been recommended based on the severity of the illness detected.