A Deep Learning and Android-Based System for Real-Time Wheat Disease Detection in the Field

As a crop vital to global food security, wheat productivity is significantly impacted by pervasive fungal and bacterial diseases. Effective disease management hinges on early and precise identification, a task traditionally reliant on agricultural experts who are often unavailable to farmers in remote regions. This paper introduces a comprehensive system that empowers farmers by integrating advanced diagnostic tools into a mobile application. We propose a compact Convolutional Neural Network (CNN) model, optimized for identifying common wheat diseases from leaf imagery. This model is deployed within a user-friendly Android application, enabling on-device inference. Farmers can capture an image of a wheat leaf and receive an immediate diagnosis directly on their device. Through the application of transfer learning and model quantization, our system balances high diagnostic accuracy with the minimal latency and small memory footprint essential for mobile deployment. This research offers a tangible advancement in democratizing precision agriculture, providing farmers with a readily accessible, cost-effective, and real-time decision-making aid.