preprocess.py

from scipy import misc
# data preprocess
def preprocess(imgRdd):
    means = np.array([127.5, 127.5, 127.5]) # mean value in BGR order

    sizes = imgRdd.map(lambda img: img.shape)

    data =  imgRdd.map(lambda original_img: misc.imresize(original_img, (300, 300))) \
    .map(lambda resized_img: (resized_img[:, :] - means) * 0.007843) \
    .map(lambda normalized_img: normalized_img.transpose(2, 0, 1)) \
    .map(lambda normalized_img: Sample.from_ndarray(normalized_img, np.array([-1])))
    return (data, sizes)