Pneumonia is a disease that affects the lungs .It causes 15% of all deaths in children under five years old. Also, it is difficult to diagnose and require an expert radiologist of chest X-ray to avoid the misdiagnosis. Chest X-ray is the most commonly and cheaply way that is used to detect pneumonia. The Lack of experts in poor countries causes a long wait for diagnosis Pneu- monia, which increase the mortality. The project aims to use Convolutional Neural Networks (CNN) to automatically diagnose pneumonia by reading chest x-rays and classify the result to normal case or Pneumonia case, and this will help to quickly and easily diagnose the disease. The proposed work is to improve the accuracy of Vgg-16 model for detecting pneumonia by applying a shuffle technique on our dataset before train the model.