Classification of Plant Diseases with Machine Learning

Abstract

Deep learning applications are also of great importance in the field of agriculture. It is an issue that should be examined in terms of what the diagnosis of the disease is and its treatment in the disease of plants. A study on the detection of plant diseases was desired. This study is intended to be done using CNN and Alexnet methods. It was aimed to observe the success rates of two different methods. It is aimed to examine the effect of classification with and without feature extraction. In other words, it is aimed to examine the effect of deep layers that alexnet has. In this study, the solution of a 38-class problem was dealt with. 38-class test and 38-class train data are available. Each of the 38 classes contains approximately 1500 to 2000 pictures. And as a result of the study, 68500 of approximately 70000 pictures consist of healthy leaves and 1500 of them are pictures of sick leaves. this whole data set is separated and classified as 80% and 20%. While the accuracy obtained with CNN was 98.27%, the accuracy obtained with Alexnet was 98.49%. In conclusion it was seen that the use of Alexnet increased the accuracy rate.