DCGAN-Leaf: A Deep Convolutional GAN Approach for Synthetic Plant Disease Image Generation and Detection

Plant disease occurs due to pests that negatively impact agricultural productivity. Classical methods for detecting disease are time-consuming and prone to human error. Such challenges to detect plant disease have been efficiently addressed by various Deep Learning (DL) and Machine Learning (ML) models. This research presents a feature augmentation method that utilizes Deep Convolutional Generative Adversarial Networks (DCGANs) to generate synthetic images combination with the Convolutional Neural Network (CNN) model for improved classification. The PlantVillage is primary dataset in this research, and DCGAN is used to broaden the dataset by generating synthetic images. By fine-tuning hyperparameters and optimizing the CNN and DCGAN architectures, the implemented model has reached an identification accuracy of 98%. The combination of generated images not only increases size of the datasets but also improves model's performance in adapting to different categories of plant disease accurately.