Classification of Soybean Diseases Using Pre-trained Deep Convolutional Neural Networks

Abstract

In this work, a novel soybean leaf disease classification technique related to pre-trained GoogleNet deep convolutional neural networks (CNN) architecture proposed. The proposed GoogleNet architecture trained on a database of 550 image samples of unhealthy and healthy soybean leaflets with 3 symptoms of an unhealthy class particularly, septoria brown spot, bacterial blight, frogeye leaf spot, and 1 healthy class using a deep transfer learning approach. As specified 3 unhealthy class and 1 healthy class identification, we have used the 5-fold cross-validation approach, the intended pre-trained GoogleNet-CNN architecture attains an accuracy of 96.25%. It was found that the accuracy of our proposed CNN architecture is enormously more precise than the formal machine learning models. The results of performance analysis to the recognition of soybean diseases exhibit the expediency and highest success rate using the proposed GoogleNet CNN model.