Convolutional neural network for maize leaf disease image classification

Abstract

This article discusses the maize leaf disease image classification. The experimental images consist of 200 images with 4 classes: healthy, cercospora, common rust and northern leaf blight. There are 2 steps: feature extraction and classification. Feature extraction obtains features automatically using Convolutional Neural Network (CNN). Seven CNN models were tested i.e AlexNet, Virtual Geometry Group (VGG) 16, VGG19, GoogleNet, Inception-V3, Residual Network 50 (ResNet50) and ResNet101. While the classification using machine learning methods include k-Nearest Neighbor, Decision Tree and Support Vector Machine. Based on the testing results, the best classification was AlexNet and Support Vector Machine with accuracy, sensitivity, specificity of 93.5%, 95.08%, and 93%, respectively.