def predict(image):

    return 'dummy caption!', ['dummy', 'caption', '!'], [1, 2, 3]