Deployment of a real-time (video) model for diseose detection in agriculture

Abstract

Plant diseases represent a major threat to agriculture and global food security. Traditional diagnostic methods, which rely on human expertise, are often slow, costly, and difficult to deploy on a large scale, thereby limiting the effectiveness of early disease detection. To address this challenge, this master's thesis proposes a real-time automated plant disease detection solution based on computer vision and deep learning techniques. The YOLOv8n model is employed to identify and localize plant diseases from images and video streams, providing a fast, accurate, and accessible detection system. The proposed solution integrates a Flutter mobile application, a FastAPI backend, Docker containerization, and cloud services and Drone. Experimental results demonstrate the effectiveness of the system for smart agriculture and plant health monitoring applications, contributing to early disease detection and the reduction of agricultural losses