Tomato mosaic virus diagnosis system

Abstract

The following compilation presents an in-depth overview of our final year project, which includes a meticulous software design document, an exhaustive report, and a comprehensive user manual. The central objective of this project was to create a Tomato Mosaic Virus Diagnosis System—an Android mobile application that is not only user-friendly and cost-effective but also extraordinarily accurate. The application leverages a Convolutional Neural Network (CNN)—a Deep Learning algorithm—to determine the presence of the tomato mosaic virus through leaf images uploaded by the user. Our software design document is methodically segmented into sections encompassing the project's scope, architectural design, decomposition description, design rationale, data design, and human interface design. This detailed framework provides a deep insight into the system's design and the intricacies of its implementation. In the report section, we elucidate vital aspects of the system like specifications, design output, inspection and testing, installation, and system acceptance tests. Developed using a three-tier layered architecture, the system's front-end user interface was designed with React Native and the CNN was built using a dataset of over 3000 images from Kaggle. The application also integrates Firebase for user authentication and FastAPI for image data processing and diagnosis. The Tomato Mosaic Virus Diagnosis System targets three primary user groups: farmers, agricultural researchers, and agricultural educators. Farmers can upload images and promptly identify the presence of the virus. Researchers can leverage the system to access real-world virus data and make meaningful contributions to improved management strategies. Agricultural educators can utilize the system as a teaching tool, showcasing the application's features and the role of technology in contemporary agriculture. The project offers a number of advantages. Its high accuracy and efficiency in virus diagnosis facilitate early detection and treatment, improving plant health and yield. By reducing the need for labor-intensive and time-consuming manual diagnosis methods, it enhances productivity and cost-effectiveness. The user-friendly interface improves user experience, and the system’s adaptability allows for potential diagnosis of other plant diseases. Ultimately, this project exemplifies how deep learning and mobile computing can drive innovation in the agriculture sector. With potential for scalability to other crops, this project has set the stage for more intelligent, efficient and sustainable food production systems.