ABSTRACT
In this paper, we assess the usage of machine learning techniques to predict the infection events of Downy Mildew. Every year, Champagne vineyards are exposed to grapevine diseases that affect the plants and fruits, most caused by fungi. Using data from an agro-meteorological station, we compare machine learning performances against traditional prediction methods for Downy Mildew (Plasmopara viticola) infections. Indeed, depending on the year, we obtain 82 to 97% accuracy for primary infections and 98% for secondary infections. These results may guide the development of Edge AI applications integrated to meteorological stations and agricultural sensors,and help winegrowers to rationalize the vine’s treatment, limiting the damages and the usage of fungicide or chemical products.
