A Hybrid Approach for Cotton Leaf Disease Detection using DCGAN and Diverse CNN Models

Machine learning and deep learning algorithms have recently played a significant role in the development of modern innovations as well as in a number of other fields, including commercial, farming, and educating. However, one of the main problems with the algorithms used in ML and DL is that they require an extensive number of samples in an appropriate dataset in order to train a neural network model. Thus, we used DCGAN (Deep Convolutional Generative Adversarial Network) to create an artificial dataset in order to accomplish data augmentation for cotton leaves in this paper work and also performed accuracy analysis for various CNN models with original and synthetic dataset. Here, an artificial dataset is one that is created by generating a fresh set of data with more samples using a variety of deep learning algorithms. Then CNN models like InceptionV3, VGG16, ResNet 50 and MobileNet V2 are used to implement the accuracy analysis for the original and generated dataset before and after data augmentation respectively, as a proof to show that synthetic dataset improves the training accuracy rate of CNN model when compared with the original dataset. At last, MobileNetV2 was found to be the best model, achieving an accuracy of 99%.