Pneumonia is a highly contagious disease that can be life-threatening, impacting either or both of the lungs, usually because of a bacterial infection known as Streptococcus pneumonia. As stated by the World Health Organization (WHO), pneumonia is responsible for one in three fatalities in India. Diagnosing pneumonia requires X-rays of the chest, which need to be evaluated by expert radiologists. Therefore, it would be beneficial to build a computerized mechanism for identifying pneumonia so that the sickness can be promptly treated, particularly in remote areas. CNNs, or convolutional neural networks, have gained significant attention for disease diagnostic because of deep learning technologies’ success in analyzing medical images. Large-scale datasets with pre-trained CNN models are helpful for image diagnosis jobs. In this research, we assessed how well-trained CNN models functioned as different classifiers and feature extractors to distinguish between normal and abnormal chest X-rays. We used statistical analysis to choose the most effective CNN architecture for the task. The outcomes show that pneumonia can be correctly identified using pre-trained CNN models.