A Deep Learning-Based Approach for Potato Disease Classification

The potato disease epidemic can cause a huge threat to food safety. In this chapter, early detection of potato disease is described through deep learning strategies. A dataset is generated using 1574 images of various diseases. This dataset is expanded to 7870 images through the data augmentation technique by utilizing scaling and rotation. Experimentation is performed by dividing the data into training and testing categories at a ratio of 8:2. Three different deep convolutional neural network architectures, such as AlexNet, ResNet, and GoogLeNet are used. The results are compared based on accuracy, precision, recall, and F1 score, and it is found that the ResNet gives the best performance for this particular application.