PaddySnap: A CNN-Powered Mobile Chatbot Application for Real-Time Paddy Disease Diagnosis and Support
Loading...
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Malaysia Sarawak
Abstract
Description
Paddy farming is vital to Southeast Asia’s economy and food security but remains vulnerable to diseases that significantly affect crop yield and farmer livelihood. Smallholder farmers often lack timely and affordable diagnostic tools, leading to delayed intervention and greater losses. This study introduces PaddySnap, a mobile application that uses Convolutional Neural Networks (CNNs) and chatbot technology for real-time paddy disease detection and support. Users can upload or capture images of paddy leaves or grains to receive instant predictions and conversational guidance. The development process included dataset collection, image preprocessing, CNN model training, Rasa chatbot integration, and app deployment using Flutter and Firebase. Results show that PaddySnap achieved 85% classification accuracy and a macro F1-score of 0.85 across ten paddy diseases. The chatbot reached 87% intent recognition accuracy and successfully responded to all valid queries. Usability testing produced a SUS score of 85.2, reflecting high user satisfaction. These outcomes demonstrate the system’s ability to support early intervention and informed decision-making. Overall, the integration of AI diagnostics and conversational interfaces helps democratize plant healthcare and encourages sustainable agriculture in underserved communities.