A HYBRID VGGNET–VISION TRANSFORMER APPROACH FOR LEAF DISEASE CLASSIFICATION USING PGGAN-AUGMENTATION

- Plant diseases represent a significant issue to agriculture field globally, causing substantial reductions in crop yield and considerable economic losses. Automated plant disease detection from leaf images has gained momentum with advances in deep learning, yet its effectiveness is often hindered by limited and imbalanced datasets. This work presents a hybrid deep learning framework as the integration of the Visual Geometry Group (VGG16) convolutional network with a Vision Transformer (ViT) to capture both local texture information and global contextual dependencies. To overcome data scarcity, Progressive Growing of GANs (PGGAN) is employed to synthesize realistic plant leaf images for augmentation. Experimental results on tomato and potato disease datasets demonstrate that the proposed VGG16–ViT model achieves superior performance compared to conventional CNN and transformer-based baselines. Without augmentation, the hybrid model attains 92.67% accuracy on the tomato dataset and potato dataset. With PGGAN augmentation, classification accuracy improves substantially to 98.49%. These findings highlight the potential of combining hybrid deep learning architectures with GAN-based data augmentation to enhance accuracy, robustness, and generalization in automated plant disease diagnosis.