PubMed · 2026-05-12
Researchers built a smartphone- and web-friendly AI tool that can identify diseases in eight common food crops from a single photo, then instantly tells the farmer what the disease is and how to treat it — no expert needed on-site.
MobileNetV3, a lightweight AI model, outperformed heavier alternatives for on-device deployment while maintaining high classification accuracy across 8 crops and multiple disease classes.
A super-resolution preprocessing step using Real-ESRGAN improved disease detection on low-quality real-world photos, addressing a key gap between lab datasets and field conditions.
The system integrates Grad-CAM visual explanations and LLM-generated treatment advice into a multilingual Progressive Web App, making AI diagnostics accessible without specialized hardware or expertise.
