Classification of level of powdery mildew disease on cherry leaves using deep attributes

Cherry fruit is a fleshy and stone fruit variety belonging to the Prunus avium family. Cherry fruit, which is consumed all over the world and has a high commercial value, is also produced and exported in our country. The cherry plant is one of the fruit types that is difficult to grow. Various pesticide diseases are encountered in the process of growing and maintaining this widely grown plant species. Although expert opinions and farmers' experience are used in the detection of these diseases, it is not enough for the healthy growth of plants. For this reason, computer technology is used to monitor all processes from planting to harvesting. Artificial intelligence applications based on image processing and machine learning used in agriculture enable safer, faster, and more cost-effective operations. Computer-aided agricultural analysis systems are widely used in the detection, identification, and monitoring of plant diseases. These developed systems offer innovative solutions to increase agricultural productivity and grow healthy crops. It helps to make the right decisions, especially during the detection and monitoring of plant diseases. In the study, it was aimed to determine the bacterial diseases seen on the leaves of the cherry plant. In the detection of these diseases, the features of the disease were determined by using pre-trained convolutional neural networks (Convolutional Neural Network - CNN). These obtained features were used to detect healthy, more or less diseased plants through LDA (Linear Discriminant Analysis), KNN (K-Nearest Neighbor), SVM (Support Vector Machine) classifiers instead of the default classifier in the last layer of the CNN network. According to the general test results, the best performance in this multi-class problem has been achieved with the Linear SVM classifier with 88.1%.