Evolution of the mobile app for early diagnosis and sustainable management of rice diseases

• Roving survey • Field data collection • Machine learning algorithms for rice disease detection and comparison • CNN based model for disease detection • VLM-based model for mobile app development for disease detection and management • Precision, accuracy, F1-score, recall, confusion matrix, ROC-AUC scores are the performance parameters Rice is an important cereal crop in India that meets the food demands of most of the people in India. To ensure qualitative and quantitative rice production, early disease diagnosis followed by sustainable management approaches are highly essential. Detection in early stage is necessary for achieving qualitative and quantitative rice production. In most of the cases, the disease is identified by the farmer in the later stage, by that time the entire crop would have been devastated by the pathogen. A machine learning algorithm provides a quick solution in the identification of diseases in turn minimizing the cost and time. In this research, the major fungal diseases in rice namely, brown spot and sheath rot and bacterial disease such as bacterial leaf blight are taken into consideration for disease detection with DenseNet121 convolution neural network (CNN). The image samples were collected from paddy fields in Virudhunagar and Madurai districts, Tamil Nadu, India during the survey. The samples were trained and tested using CNN algorithm to detect the disease in the early stage. The training and validation accuracy was 97%. The performance parameters precision, recall, f1-score, ROC-AUC scores, confusion matrix details were presented to validate the algorithm. For easy handling, a mobile application was developed for disease detection and management using large language model (LLM) and visual language model (VLM) approach. The mobile app development was supported by GPT4 and PaLM2. The GPT4 accepts images and identify the disease type. The PaLM2 was used for diagnosis and disease management. The diseases detection with integrated disease management system was presented. The real-world testing of the mobile application was carried out at Madurai agricultural college and research station, Tamil Nadu, India.