Chapter nine - Crop disease classification using deep learning approach: an overview and a case study

Abstract

The application of deep learning for classification of various disease in crops are evolving at a rapid phase with the advent of breakthrough technologies in graphics processing units (GPUs). Existing research indicates the use of pretrained convolutional neural networks (CNN) with an ImageNet data set known as the transfer learning approach yields the highest accuracy for the disease classification. The important shortcomings of the deep learning approaches are the exploitation of the computational resources which limits its implementation in a commercially available smartphone. SqueezeNet is one of the architectures that has been exclusively developed with reduced parameters and provides a performance equivalent to AlexNet. This study has utilized this least explored architecture for evaluating the accuracy in disease classification using the PlantVillage data set. In addition, we provide an overview of the deep learning architecture with basic building layers. The use of SqueezeNet resulted in the best classification accuracy of 98.49% with original color images. We discuss the misclassification of a few classes and possible factors influencing the learning of features. The possible approach for future studies has been highlighted for the development of feasible real-time disease classification.