Agriculture is being a substantial contributor to the world’s economy. Plant diseases may cause several issues like crop losses, economic impact, biodiversity loss, reduces food production and its quality. Convolutional Neural network(CNN) is an artificial nueral network that is most popularly used for analyzing images. Since the CNN model is more suitable for identification and classification of images, the researchers are continuously exploring the use of CNN in agriculture domain. Three CNN pre-trained models namely, VGG16, Inception v3, and Xception are considered in this paper for the comparative study. Analizing these CNN models, it is possible to extract the fine-grained picture characteristics. It helps to improve the deep learning capabilities and prevent overfitting. The inception model extract information from numerous transformations, then concatenate along the output channel. In Xception model, before the 1x1 channel correlations, deep separable convolutional layers are stacked and complete the spatial mapping using residual connections. In VGG16, the image is passed throgh a stack of convolutional layers. Each convolutional filters has very small receptive fields of size 3x3, and the convolution is carried out with stride 1. Here, a comparison model is created where each model is first taken and its accuracy is determined. After combining VGG-Incp, Incp-Xcep, and Xcep- VGG, the model needs to be trained to determine its accuracy. The results of Incp-Xcep-VGG are merged into single model to determine its efficiency. Based on the results of this study, the Incp-Xcep-VGG achieved a greater accuracy of 85%.

Convolutional Nueral Networks ,Inception V3, Xception, Vgg16.