Assistant Research Professor Cornell AgriTech, Cornell University
Disclosure: Disclosure information not submitted.
Advances in precision grape disease management have been severely lacking. Downy mildew, caused by the oomycete pathogen Plasmopara viticola, is one of the most damaging and difficult to control diseases in eastern grape production. Chemical fungicides are the primary management method used to reduce disease incidence and severity. Currently, fungicide efficacy evaluation relies on visual assessment, which is laborious, and can be imprecise and subjective when performed by undertrained individuals. The goal of this study was to develop a deep learning-based approach to evaluate downy mildew severity to quantify fungicide efficacy and compare to human assessment. In 2019, six fungicide treatments were applied biweekly from May 29 through August 20 on four replicate blocks consisting of 8 vines spanning 14.6 m/row. Whole-vine images were collected on August 29, and visual ratings were obtained for 20 leaves per replicate on September 9. A total of 600 images were randomly selected for labeling and split into training (70%), validation (20%), and testing (10%) subsets. A deep learning-based object detection model was trained to identify areas infested by downy mildew in a single image. An image-processing pipeline was developed to calculate the disease severity and distribution for each treatment. Images in each treatment panel were stitched to form a panoramic photo that was then divided into 10 equal sub-images corresponding to 1.46-meter panel. Infected areas were identified by the trained model in individual segments of that panel. The total infected area was used to represent disease severity, and the histogram of the number of infected areas per panel segment was used to describe disease distribution. Correlation analysis was conducted between image-derived measures and human assessment to evaluate the efficacy of the developed approach. The findings of this study have positive implications for data-driven disease management and will ultimately enhance vineyard productivity and profitability.