scikit-learn compatible transformer that extracts haralick features from a given collection of images

haralick_extractor.py

import joblib
from mahotas import features
from sklearn.base import TransformerMixin, BaseEstimator


class HaralickExtractor(BaseEstimator, TransformerMixin):
    def __init__(self, n_jobs=1):
        self.n_jobs = n_jobs

    def fit(self, X, y=None):
        return self

    def transform(self, X):
        extract_features = joblib.delayed(lambda x: features.haralick(x).mean(0))
        execute = joblib.Parallel(n_jobs=self.n_jobs)
        images = execute(extract_features(x) for x in X)
        return np.array(images)