Optimizing Plant Health Monitoring: A CNN Model Based Rice and Cotton Disease Prediction

Rice and cotton are crucial cash crops in India that contribute significantly to the country’s economy and provide employment to millions of people. To ensure high crop yields, it is essential to effectively manage crop diseases. However, traditional monitoring methods relying on human observation can be time consuming and inefficient. Hence, a system that can accurately and effectively detect plant diseases at the beginning is needed. In this paper, we present our study into the training and evaluation of different CNN models using a shared dataset that focuses on five diseases, including blast, bacterial leaf blight, brown spot, and diseased or healthy cotton leaf in cotton plants. The ResNet, VGG19, and Inception models obtained accuracy percentages of 63%, 66%, and 89%, respectively. On the same dataset, the general CNN architecture model had a 95% accuracy rate. Our findings suggest that the basic CNN architecture can be used effectively to predict diseases in rice and cotton plants. The general CNN model outperformed the ResNet and Inception models, indicating that a simpler architecture can yield better results for this specific task. In the context of Indian agriculture, this work offers important insights for the creation of a precise and effective system for the early diagnosis of plant illnesses. Finally, based on the input given to the model the disease is predicted and a cure for that predicted disease is given both in organic and chemical methods.