Leveraging MobileNet, InceptionV3, and CropNet to Classify Cassava Plant Disease

Abstract

In sub-Saharan Africa, cassava is widely grown and considered to be a large source of carbohydrates for human food. However, the plant is plagued with diseases which can threaten food supply for millions of people. By using computer vision, researchers attempted to create an image classification model that can tell farmers whether the plant is sick or not by taking pictures of their leaves. In this short paper, the author attempts to train three Convolutional Neural Network: CropNet, MobileNet, and InceptionV3 that can classify cassava plant diseases based on visual data. As a novelty, the author creates an ensemble voting classifier that combines the prediction of CropNet, MobileNet, and InceptionV3 to create a better prediction. Turns out, creating an ensemble voting classifier enables us to achieve an accuracy score which is 6.8% higher than the average individual scores of each model.