Corn Disease Identification from Leaf Images Using Convolutional Neural Networks

Abstract

Plant diseases are one of the major factors that can negatively impact crop yields. Therefore, we want to create an automated system that can identify plant diseases as easily as possible. Our research goal is to help crop growers who might not be experts in plant pathology be able to identify diseases as soon as they occur in order to minimize the spread of disease and the damage to the crops. We propose a very light-weight convolutional neural network architecture that is trained to classify images of corn leaves into one of the four categories: common rust disease, northern leaf blight disease, gray leaf spot disease, or healthy. The dataset used is the PlantVillage dataset taken from plantvillage.org. The dataset consists of 8506 images of corn leaves from 4 categories. We evaluated our model using 5-fold cross-validation, and we achieves an accuracy of 97.09 % on average. The result of our model is slightly worse than the performance of other well-known CNNs architectures, but the complexity of our architecture is much lower, making it more suitable for real-time inferences on resource-constrained devices such as smartphones.