Diagnosis of pneumonia is of great importance in healthcare, particularly when applying the best treatment approaches and time early in the stage of the disease. This study provides an evaluation of DL technologies for pneumonia detection. This paper offers a comparative study of CNN architectures utilized diagnosis for pneumonia from chest X-ray images. We examined the achievements of famous CNN models, including AlexNet, GoogleNet, InceptionV3, ResNet18, and a custom CNN architecture, on a testing dataset. This was possible through extensive testing sessions in which we examined the precision with which each model detected pneumonia in comparison with its design constraints. According to our results, ResNet18 and our tailored CNN model achieved accuracies of 94.23% and 95%, respectively, and were the most successful. In contrast, we emphasize the use of existing methods of transference learning and data increase to maximize model performance and ease of generalization. This work provides important clues for medical picture detection and treatment with the opportunity to better develop future pneumonia diagnostic methods. Finally, the exploration of such network architectures and optimization strategies can advance better diagnosis and care for patients in medical institutions.