Leveraging Artificial Intelligence to Develop a Decision Support Tool for Visual Symptom Assessment of Grapevine Leafroll and Red Blotch Diseases

Abstract

The timely detection of viral pathogens in vineyards is a critical aspect of management. Diagnostic methods can be labor-intensive and may require specialized training or facilities. The emergence of artificial intelligence (AI) has the potential to provide innovative solutions for disease detection but requires a significant volume of high-quality data as input. With that purpose, we partnered with wine grape growers to collect a robust dataset of verified images. We used those images to train an AI model and develop a handheld application as a decision support tool for grapevine leafroll and red blotch diseases. The tool allows users to scan a grapevine canopy with a mobile device and view a confidence reading describing the likelihood that the imaged vine has visual symptoms consistent with leafroll, red blotch, or a healthy vine. The 86% accuracy under field conditions and generally positive user experience suggest there is potential for the trained use of AI as an investigative tool to quickly assess visual symptoms associated with these grapevine diseases.

Keywords: artificial intelligence; leafroll; red blotch.