Synthetic Generation and Augmentation of Rice Disease Images Using Conditional GANs

Rice disease identification suffers from limited labeled datasets, class imbalance, and poor field generalization. This paper proposes a conditional, GAN (cGAN)-driven synthetic image generation and augmentation framework for rice leaf disease diagnosis. The methodology integrates dataset preprocessing, class-conditioned image synthesis, hybrid augmentation, and a CTLCN-based classifier, followed by expert system deployment for decision support. Experimental design targets improved diversity, reduced overfitting, and robust recognition of blast, bacterial blight, brown spot, and sheath blight. Expected findings include improved classification accuracy, F1-score, and robustness over non-augmented baselines. The framework applies to scalable smart agriculture, mobile advisory systems, and precision crop disease surveillance.