According to the latest reports of the World Health Organization (WHO), pneumonia was and unfortunately remains the biggest killer of children under the age of five in the world, and thus kills more than a million children every year, in addition to infecting adults, especially the elderly, who are over the age of 60, which is fatal for them in this regard. Pneumonia is an infectious disease that affects the respiratory system due to a viral, bacterial, or other infection. It causes difficulty breathing, in addition to being fatal if treated in its early stages. It is usually detected by chest X-ray, but X-rays sometimes reveal other similar diseases. As for pneumonia, this in itself is a diagnostic problem. To solve this problem, convolutional neural networks (CNN), or so-called deep learning, have been used in medical imaging for early detection of pneumonia without suspicion of other diseases. The experiments whose results were presented in this article were conducted in two ways. The first method relied on transfer learning with training adjustment. The second method relied on training adjustment without freezing any of the layers of neural networks. Both methods used chest X-ray images on the Kaggle website, which we relied on from the beginning after preparing the data. The data was augmented and trained with the traditional convolutional neural network model, and then we implemented both methods on six deep learning models, namely VGG 16, InceptionV3, Xception, ResNet50, MobileNetV2, and DenseNet169. The VGG 16 model showed the highest accuracy among the models.