I have built a CNN model to classify images on either two categories, “Pneumonia” and “Normal”. The data i fed are from chest X-ray images converted to a hdf5 file then loaded into a numpy array. The original size of images varied quite a lot so i resized them as 1024 x 1024 and its in greyscale. the training set has 460 images per class for training and 230 images per class for testing. The overall accuracy seems to be just better than a random guess. I also added data augmentation to improve the accuracy but it doesn’t seem to work.