Özet:

The process of cultivation of plants is a difficult and long-term process. One of the most important problems of people who deal with plant cultivation is plant disease. The first thing to do in the fight against the disease is to recognize the disease. It is very important to be able to quickly detect the disease and take the necessary measures. In the study, deep learning methods were used to identify the disease in the tomato leaves. The study contains a total of 18,160 images of tomato leaves consisting of 10 classes as a disease. In the image disease classification, GoogleNet, AlexNet, SqueezeNet, ShuffleNet and ResNet-18 models, which are pre-trained networks from deep evolutionary nervous networks (ESA) models, have been used. The training data set in models is allocated to 70% training, 15% verification and 15% test. With the test data of the trained networks, performance standards are calculated with accuracy, accuracy, peculiarity and f-score values. The accuracy rates for the models AlexNet, GoogleNet, ShuffleNet, SqueezeNet and ResNet-18 were 93.93, 95.18, 94.82, 94.29 and 81.79 respectively. According to the analyses made, it has been found that the best performance model in the study of tomato leaf disease classification of pre-trained networks is GoogleNet.

Anahtar Kelimeler: