Advances in image identification and classification algorithms combined with the available computing power have enabled computers to identify and classify images with great precision and speed. Modern computers, with well-trained algorithms, are capable of identifying and classifying images in fractions of a second when compared to any human competitor. This work will demonstrate the comparative results of three convolutional neural networks, namely: InceptionV3, ResNet50V2 and Xception. The work will discuss the neural networks mentioned and their respective results regarding accuracy, precision, f1-score and the Confusion Matrix. The algorithms above were trained using chest x-ray images of healthy patients, patients with viral pneumonia and patients with bacterial pneumonia. Although there are three types of patients, we only see two possibilities: normal or pneumonia. As a result of the three networks analyzed, the one that performed best in identifying images with pneumonia was InceptionV3, delivering an average weighted result of 91% of the demonstrated metrics and was also the one that presented the best result in the Confusion Matrix. The other neural networks were behind by approximately four percent in the metrics. Given these results, it can be stated that all networks performed well in classifying images with Pneumonia.