Potato Disease Diagnosis Using YOLOv5 and Web-Based Deployment
Loading...
Date
2025-05-06
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Uganda Christian University
Abstract
Agriculture plays a central role in Uganda’s economy, and potatoes are among the
most important staple crops. However, potato yields are significantly threatened by
diseases such as early blight and late blight. Timely detection of these diseases is
critical to reduce losses, minimize pesticide misuse, and enhance food security.
This project presents a web-based potato leaf disease diagnosis system using YOLOv5,
a state-of-the-art deep learning object detection model. The system classifies potato
leaves as healthy, early blight, or late blight. The backend is implemented using
FastAPI and deployed to Render, while the frontend is built in React and hosted on
Vercel, ensuring accessibility via modern web browsers. The model was trained on
the PlantVillage dataset.
Evaluation results show that the system achieved high accuracy and fast inference
times, making it suitable for use by farmers and agricultural officers. This report
details the system design, methodology, model training, deployment, and perfor-
mance evaluation. Limitations such as environmental noise, internet dependency,
and limited disease coverage are acknowledged, and future work includes expanding
the disease scope, offline deployment, and integrating treatment recommendations.
This work contributes to the growing field of AI-powered agriculture in Uganda and
aligns with the Sustainable Development Goals for food security and smart farming.