Development of Paddy Leaf Disease Detection System by CNN Using Thermal Camera

This project aims to leverage the advanced artificial intelligence capabilities of the Mobilenetv2 architecture to introduce an innovative approach to identify diseases in rice crops. Growing rice presents numerous challenges, including the impact of diverse factors such as shifting ecosystems, weather conditions, pests, and diseases on its productivity. To achieve widespread food production, a combination of traditional practices, modern agricultural techniques, innovation, and technological tools is essential. In this effort, we have created a model utilizing a convolutional neural network, MobileNetV2 architecture, designed to detect diseases in rice leaves with higher accuracy. Our proposed model for identifying rice diseases promises to deliver more accurate outcomes. This model could be adapted into TensorFlow Lite (TFLite), enabling its use in mobile devices and drone technology, among other applications. The model developed in this study achieved a training accuracy close to 96% to 97% and a test accuracy of 97% when trained on 2200 well-structured datasets. A loss function of 0.014 over 100 epochs indicated that the model was successfully trained using ResNet50.