Accuracy Analysis of Various CNN Models with Synthetic and Original Dataset for Cotton leaf

In this day and age, almost in every field of work Artificial Intelligence has its part to play. Likewise in the field of agriculture, for the early detection of diseases Artificial Intelligence algorithms were used. These algorithms use image classification models to perform this operation. So, to get better result in the early disease detection these models need to be trained with a balance dataset with huge of number of training samples. But finding a balanced dataset is quite impossible so in this manuscript we created a fresh dataset for cotton leaf by combining the images from the original and synthetic dataset. For that first we generated the synthetic dataset for cotton leaf using Deep Convolutional Generative Adversarial Network(DCGAN)algorithm which generates the artificial images like the images present in the original dataset. Then Convolutional Neural Network models like: Inception V3, Visual Geometry Group(VGG) 16, Residual Network(ResNet) 50 and MobileNet V2 are used to implement the accuracy analysis for the original dataset and newly created dataset to show that newly created dataset with synthetic images provides better accuracy rate when compared with the original dataset.