Elisabeth Clover Artemis Kitchin, David S. McCall, Caleb A. Henderson
Abstract
Dollar spot (Clarireedia jacksonii) is one of the most economically significant diseases to amenity turfgrasses, with frequent fungicides often applied for acceptable suppression. The disease is one of the most widely studied diseases among academic and industry researchers. Accurate and efficient quantification of dollar spot incidence is vital for research but requires extensive time requirements for routine infection center counts. This project presents the development of a machine learning model for objective quantification of dollar spot incidence and estimated disease coverage from digital images. Through extensive training on a diverse and comprehensive dataset, our machine learning model has achieved an accuracy and precision rate of 93% and 76%, respectively, in identifying and quantifying dollar spot on creeping bentgrass. These metrics underscore the model's efficacy in recognizing and quantifying dollar spot manifestation across various conditions. Leveraging deep learning techniques, the model identifies and delineates the spatial extent of dollar spots in turfgrass images, thus eliminating the inherent subjectivity in manual assessments. This model aids researchers by reducing the reliance on labor-intensive, manual assessments susceptible to observer bias. The automation of this machine-learning model fosters the standardization of dollar spot quantification. It offers a tool for turf pathologists to make well-informed decisions regarding disease suppression in various chemical, cultural, and cultivar evaluations.
Comments