Performance Comparison of Image Classification Models for Corn Leaf Disease

Abstract

This study was to performance comparison of image classification model for corn leaf disease using Microsoft Azure tool to corn leaf disease images classification. The objective of this study was to compare the accuracy of models, size of models, and processing time of models for corn leaf disease image classification : InceptionV3, VGG16, Xception and Custom Vision which used 4 classes of corn leaf disease datasets : Leaf Blight, Corn Rust, Gray Spot Leaf, and Normal for use as a dataset of training datasets in machine learning methods to create the model and use as testing datasets to test the model of corn leaf disease image classification. From the experimental results, were found that the most performance model for corn leaf disease image classification is Custom Vision with the accuracy of model equal 98.1 percent and size of model equal 21.2 megabytes, and processing time of model equal 6.89 seconds.