Factors Affecting Accuracy and Deployment of Wheat Rust Detection Models

Abstract

Wheat rusts, a group of fungal diseases including stem rust, leaf rust, and stripe rust, continue to be one of the most devastating threats to global wheat production. The disease can cause yield losses of up to 70% in epidemic years, posing serious challenges for food security and agricultural sustainability. With the global demand for wheat projected to rise by 60% by 2050, developing effective strategies for early detection and control of wheat rust has become an urgent research priority. Recent advances in artificial intelligence, particularly deep learning, have provided promising tools for the detection of plant diseases through image analysis. However, the accuracy and real-world deployment of wheat rust detection models face several critical barriers. This paper discusses key influencing factors affecting both the performance and adoption of these models. Drawing on experimental data and comparative model analysis, the study identifies the most important determinants of detection accuracy, such as dataset quality, model complexity, environmental variability and image resolution. Additionally, deployment-related factors such as computational efficiency, hardware availability, user trust and socio-economic barriers are examined. Recommendations are made for improving datasets, optimizing architectures, developing resource-efficient models and strengthening farmer engagement. The findings aim to bridge the gap between algorithmic research and practical implementation, contributing to the broader goal of leveraging AI for sustainable agriculture.