Identification of Leaf Diseases in Potato Crop Using Deep Convolutional Neural Networks (DCNNs)
Zubair Saeed, Misha Urooj Khan, Ali Raza, Nazish Sajjad, Sana Naz, Ahmad Salal
Abstract
Potatoes are used by a billion people globally as its intake per capita is 31.3 kilograms per month worldwide and its crop production surpasses 300 million metric tons. They have great nutritional potential and offer more protein and iron than other vegetables. Various diseases, such as early blight, late blight, and others, will attack potato plants and manifest their symptoms in the leaves. If these outbreaks are discovered early and appropriate treatment is taken, the farmer will not suffer significant economic losses. In this article, the suggested computer vision deep learning approach would significantly help in early identification and detection of different potato disease. Many plant diseases have different visual characteristics that may be utilized to accurately identify and classify them. This study describes a potato disease classification method based on computer vision-deep learning combination. The system trains deep convolutional neural networks such as ResNet-152 and InceptionV3 on the Kaggle potato dataset with an accuracy of 98.34% and 95.24% respectively at learning rate of 0.0005. This system successfully performs the classification of three categories: healthy, early, and late blight for potato leaves.