A Comparative Analysis of SVM and CNN for Accurate Diagnosis of Plant Diseases

Plants constitute vital components of nature, helping to keep the ecosystem balanced. Potatoes and peppers are among the most popular plants available all year. Potato is a popular carbohydrate-containing meal that comes in various colors and flavors. Pepper is also an essential ingredient in many meals, adding a distinct taste and color. The main motive for doing this work is to take care of these plants and reveal the diseases they infect using two different algorithms and compare them. The first algorithm, the support vector machine, was employed for classification after multiple processing steps on plant leaves. The second algorithm is the convolution neural network, which consists of several layers built to detect diseases in plant leaves. Five classes have been classified in the two mentioned algorithms. Three are categories for potatoes and two categories for peppers. The comparison result was that the convolution neural network algorithm outperformed in accuracy and performance. In the convolution neural network algorithm, we obtained an overall accuracy rate in training (99.90) and testing (96.10). In contrast, in the support vector machine algorithm, we received (81.03) in training and (71.61) in testing, and even in terms of time, the best thing about the convolution neural network algorithm was that it does not require additional time for feature extraction.