Plant Disease Detection System Using Convolutional Neural Networks and TensorFlow Lite

This research paper presents a comprehensive, intelligent system for detecting plant diseases using deep learning models. The system's novelty lies in its ability to automatically retrain itself when sufficient new data is gathered, allowing for continuous model improvement without manual intervention. Unclassified plant images are routed to an “unknown” category, where domain experts can manually label them. The retraining process is triggered once a threshold number of images are labelled, ensuring the model can learn from real-world data continuously. The system utilizes a backend developed with Flask and a frontend using Next.js 14 and Tailwind CSS for an intuitive user experience. Additionally, secure APIs are implemented to ensure that only authenticated administrators have access to manage the system and the unclassified images. Over several retraining cycles, the system demonstrated improved accuracy and adaptability, addressing critical challenges in plant disease detection and classification in real-world applications.