from scipy.io.matlab import loadmat from sklearn.datasets._base import _pkl_filepath import joblib import os data_home = os.environ.get('SCIKIT_LEARN_DATA', os.path.join('~', 'scikit_learn_data')) data_home = os.path.expanduser(data_home) mfile = loadmat(file_name=os.path.join(data_home, "olivettifaces.mat")) faces = mfile['faces'].T.copy() filepath = _pkl_filepath(data_home, 'olivetti.pkz') joblib.dump(faces, filepath, compress=6)