Early detection of pneumonia, caused by various reasons like COVID-19 and viral infections, is crucial. Prompt action is the need of the hour to prevent disease progression, reduce hospitalization, and limit long-term damage to the lungs. In the healthcare domain, computer-aided diagnosis is increasingly popular to bring down the vigorous pressure on the healthcare system. The proposed work presents a multiclass pulmonary disease classification system. Two different models, namely VGG16 and DenseNet169, are applied to classify normal lungs, lungs infected with COVID-19, lung opacities, and viral pneumonia from human chest radiography images. This study utilizes COVID-19 radiography dataset which has class imbalance. Conditional generative adversarial network (cGAN) is used to execute data augmentation to overcome the problem of the limited samples in dataset. Accuracy, precision, recall, and loss are used to check the effectiveness of the proposed scheme. The accuracy obtained by VGG16 and DenseNet169 are 92.39% and 90.45%, respectively. The presented work effectively and efficiently classifies the lung diseases that can used for the early detection and treatment of pulmonary disease.