DETECTION OF MAJOR RICE AND POTATO DISEASES USING TENSORFLOW AND MACHINE LEARNING

Abstract

Rice and potato are the staple food for over half the world's population. Early and quick detection of rice and Potato diseases are crucial important for our agricultural industry. Several studies focused this issue, and their findings varied depending on their methods. The approach used in this piece of research to identify the four common diseases of rice and two potato diseases including Rice leaf blast, Rice leaf blight, Rice brown spot, Rice leaf smut, Potato early blight and Potato late blight using TensorFlow machine learning technique. The disease samples were collected and sample pictures were captured while visiting the crops field. The causal organisms of rice blast and Bacterial leaf blight of rice were isolated and identified as Magnaporthe oryzae and Xanthomonus oryzae pv. oryzae. The rest of the selected diseases were identified as per the typical symptoms. In this piece of research, the prediction model is built using TensorFlow’s Keras API and the AlexNet CNN. The machine learning model was created using the open-source TensorFlow platform. Following the creation of the TensorFlow Tflite model, this is transformed into the Core ML model, which is then used in the android app to predict diseases. TensorFlow functions by using thousands of plant disease leaf images by converting the input data to Core ML model through Adam optimizer. The model was developed based on the label dataset collected from farmer’s field, research field and online domain. TensorFlow machine learning techniques found to be effective showing 99% accuracy by image augmentation. This concept could be used in the creation of mobile applications that aid farmers in identifying rice and potato diseases and suggesting the suitable solution to the farmers. Thus, to prevent the production losses of rice and potato crops due to the diseases mentioned, the model are suggested to practice by the concerned growers.