Plant Disease Detection Using Ensemble Learning

Abstract

Ensemble learning is a machine learning method that combines the predictions of multiple models in order to make a more accurate prediction. In this study, a plant disease detection system was developed using ensemble learning, with six base models Inception V3, MobileNet, MobileNetV2, VGG16, GoogleNet, and ResNet50 trained on a dataset of 87,000 images of healthy and diseased plant leaves from 38 different classes. The base models achieved accuracies ranging from 72.7% to 97.2% on the plant disease dataset. To improve the accuracy of the system, both soft and hard voting classifiers were applied to the base models. Soft voting involves weighting the predictions of each base model according to their accuracy and taking the class with the highest weighted average as the final prediction. Hard voting consists of simply counting the number of votes cast for each class, and the class with the most votes is selected as the final prediction. Both soft and hard voting significantly improved the accuracy of the system, with the soft voting ensemble achieving an accuracy of 97.8% and the hard voting ensemble achieving an accuracy of 98.3%. The results of this study demonstrate the effectiveness of using ensemble learning for plant disease detection and the potential for such a system to assist in the accurate and efficient identification of plant diseases.