Comparison of Plant Leaf Classification Using Modified AlexNet and Support Vector Machine

Abstract

Automatic identification methods for the early detection of disease in plants play a significant role in precision crop protection. Various methods have been employed in the task of plant disease recognition. This work benefits in actual identification of a plant and further detection of disease in them. In this paper, the leaf images of 9 different plants with 32 different classes of the PlantVillage database are analyzed for the process. The main contribution of this work is to classify the plant leaf disease with the proposed network-based on AlexNet and comparing with the traditional support vector machine. The convolutional neural network is used to detect the plant leaf and identify the healthy and diseased plant through this network. The mixed combination of healthy and diseased plant leaf data is used for training the convolutional neural network. Transfer learning is used for the pre-trained AlexNet network for a different amount of data for training of the network, and results are validated with a support vector machine and deep learning classifier. AlexNet performed well with an accuracy of 91.15% as compared to SVM giving 88.96% and 89.69% for radial basis function kernel and linear kernel respectively.