A Novel Method of Plant Leaf Disease Detection Based on Deep Learning and Convolutional Neural Network
Xulang Guan
Abstract
This study has developed a new plant disease detection approach by combining four CNN models. The experiment used an open source database of 36258 images classified in 10 plant species and 61 classes of healthy and disease plant leaves. 36258 images were divided into two datasets with 31718 images for the training set and 4540 images for the validation set. Four CNN models including Inception, Resnet, Inception Resnet, and Densenet were deployed and the results of CNN models were processed by a stacking method. The use of the stacking method achieved an 87% accuracy rate, which is a significant improvement compared to the result of using a single CNN model. The relatively high accuracy rate indicates that using a combination of CNN models by a stacking method could be an appropriate approach that can be extended to practical cultivation conditions as an advanced plant disease warning tool.