Classification of vine leaf diseases using deep learning methods

Vine leaves are an important food source for humanity. Besides its fresh consumption, this plant provides a wide variety of products. This plant, which is widely grown throughout the world, has economic value. Diseases affecting vine leaves can cause significant financial losses by damaging both the fruit and the leaves. The increasing demand for pickled vine leaves leads to a new production model in viticulture. In order to obtain abundant and high-quality products, it is very important to apply different cultural methods in the vineyards, as well as to diagnose diseases early and take the necessary precautions. The image processing method that emerged with developing technology enables accurate and rapid identification of plant diseases in agricultural activities. Today, the increasing use of artificial intelligence and image processing applications in agriculture offers solutions to problems in this field and creates alternatives to the methods used to date. In recent years, plant identification systems have been successfully used to solve problems such as species identification, yield and disease. The aim of this study is to classify and identify grape leaves as diseased and healthy leaves using convolutional neural networks. Colomerus vitis (vine scab) and Plasmopara viticola (downy mildew) infected leaves and healthy grapevine leaves were classified using pre-trained deep learning models VGG16, VGG19, InceptionV3, Xception, MobilNet and DenseNet201. Analysis, 95.33% with VGG16 model; 98.00% with VGG19 model; 96.67% with InceptionV3 model; 94.67% with Xception model; It showed 96.00% classification accuracy with the MobileNet model and 96.67% with the DenseNet201 model. The highest classification success was achieved with the VGG19 model. The performance of the models was evaluated through accuracy, sensitivity, precision, and F-1 score metrics. The results obtained reveal that these methods can be used successfully in determining grapevine leaf disease types. In addition, the main contribution of this study to the literature is the creation of a new data set consisting of grape leaves of Narince grape variety for three different classes: vineyard scab and downy mildew diseases and healthy ones.