Deep learning-based image detection for wheat streak mosaic virus

Overview

Wheat streak mosaic virus (WSMV) is one of the most economically damaging viral diseases affecting winter wheat production. This solution applies deep learning to plant disease detection by training a deep convolutional neural network (DCNN) to recognize WSMV infection directly from leaf photographs. The goal is rapid, non-invasive, image-based identification that can eventually be deployed through smartphone-assisted diagnostics, supporting precision agriculture and enabling earlier intervention by growers, breeders, and crop scouts.

Technical specifications

• Deep convolutional neural network (DCNN) trained to classify WSMV-infected versus healthy wheat leaves from photographic images
• Training dataset built from hundreds of images collected across multiple Kansas winter wheat fields and controlled greenhouse studies, capturing natural variability in lighting, growth stage, and symptom expression
• Ground-truth validation using ELISA and RT-qPCR laboratory results paired with corresponding leaf images, ensuring that model labels reflect confirmed viral presence rather than visual assumption
• Field and greenhouse pipelines for ongoing image collection, including healthy controls and stressed plants that mimic viral symptoms to reduce false positives
• Performance evaluation through specificity and sensitivity testing to characterize model accuracy and reliability
• Target deployment pathway toward smartphone-based image capture for in-field disease diagnosis

Technology readiness level

This technology is at an early-to-mid stage of development (TRL 3–4). The research team has established a foundational image library with laboratory-confirmed labels and is actively expanding the dataset through additional greenhouse and field collections. Future validation will include controlled healthy plant imaging, stressed-plant controls tested with ELISA and RT-PCR, and database augmentation to improve model robustness. Specificity and sensitivity benchmarking are planned to confirm diagnostic performance before field deployment.