AI-Driven Wheat Rust Detection

Abstract

Wheat is a staple crop in Pakistan, contributing significantly to national food security and the agricultural economy. It provides nearly 60% of the daily caloric intake for the population and supports the livelihoods of millions of farmers. However, wheat production faces persistent threats from rust diseases, particularly stripe rust, leaf rust, and stem rust which can severely affect crop yields if not diagnosed and addressed promptly. Conventional rust intensity rating relies on manual visual examination, which is generally less effective, more time-consuming, and susceptible to human error. This project provides an AI-based solution for automating wheat rust severity grading. The system is based on a computer vision solution involving a Siamese Neural Network comparing infected wheat leaf images with a reference set and giving a severity grade on an 8-level scale from Immune to Susceptible. With the help of a deep feature comparison mechanism and a visual similarity determined using a distance metric, the system provides efficient and accurate results. In order to enhance usability and accessibility, the model has been deployed through a stand-alone web application where leaf images are input by users, and an instant severity rating is obtained. The tool will be employed to enable early detection and knowledge-based decision-making for disease management in wheat. The system can be scaled and will serve as the foundation of a complete digital agriculture platform. Future efforts will include expanding the tool to additional plants and diseases, providing treatment recommendations, and making the platform affordable and available to farmers in Pakistan and other similar agricultural regions. Index Terms— Wheat rust. Severity, Deep learning, Siamese network, Feature similarity, Agricultural decision support, Image-based classification, Plant pathology, AI for agriculture.