Feeding Images to be predicted in a CNN

I have a got a CNN, have run it on given training/testing data for a satisfactory amount of epochs, and then saved it as a .hd5 file. Now, I have got fresh images, in jpeg format, and need to run them through my model and obtain predictions. I have just the jpegs and they are not labelled. How do I proceed?I am using this snippet, to run the predictions on the test/validation dataset.

y_pred = mod(test_examples[:50], training = False) y_pred_argmax = np.argmax(y_pred, axis=3) 

Test and Train images are of 512x512x1 shape. Number of classes(features) in the model is 8.