Developing an application to detect groundnut diseases to enhance decision-making

Abstract

Agricultural development faces significant threats from plant diseases, particularly in regions where climate change amplifies the spread of pests and pathogens. Traditional methods of detecting these diseases, such as visual observation of symptoms or sending samples to laboratories, can be time-consuming, costly, and difficult for small farmers to access. To help farmers to address these issues, this study aims to develop a mobile application that uses deep learning to identify and sort groundnut leaf diseases by analyzing pictures. The application uses three different deep learning models, CNN, VGG16, and VGG19, that were trained on a dataset from Mendeley with over 1,700 labeled images of healthy and sick groundnut leaves. Out of the learning models, VGG19 performed the best with 96% accuracy results for four main diseases and healthy categories: Alternaria, leaf spot, rosette, rust, and healthy leaves. The system is built with Flutter and offers features like logging in securely, working offline with SQLite for diagnosis, and a chatbot that helps farmers. It is designed to be simple to use, even in rural areas where the internet might be spotty. Images were pre-processed and augmented to improve the AI’s performance, making the model better at handling different situations. Its accuracy was tested using metrics like precision, recall, F1-score, and confusion matrices, which all showed that the VGG19 model is dependable. This innovative tool gives farmers quick, real-time insights to help them make smarter decisions, cut down on crop losses, and support sustainable farming. By combining artificial intelligence with an intuitive mobile interface, the study shows how smart systems can boost climate-resilient and precise agriculture, helping farmers adapt to changing conditions more easily.