Face classification using Haar-like feature descriptor

Haar-like feature descriptors were successfully used to implement the first real-time face detector 1. Inspired by this application, we propose an example illustrating the extraction, selection, and classification of Haar-like features to detect faces vs. non-faces.


This example relies on scikit-learn for feature selection and classification.



Viola, Paul, and Michael J. Jones. “Robust real-time face detection.” International journal of computer vision 57.2 (2004): 137-154. https://www.merl.com/publications/docs/TR2004-043.pdf DOI:10.1109/CVPR.2001.990517

import sys
from time import time

import numpy as np
import matplotlib.pyplot as plt

from dask import delayed

from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import train_test_split
from sklearn.metrics import roc_auc_score

from skimage.data import lfw_subset
from skimage.transform import integral_image
from skimage.feature import haar_like_feature
from skimage.feature import haar_like_feature_coord
from skimage.feature import draw_haar_like_feature

The procedure to extract the Haar-like features from an image is relatively simple. Firstly, a region of interest (ROI) is defined. Secondly, the integral image within this ROI is computed. Finally, the integral image is used to extract the features.

def extract_feature_image(img, feature_type, feature_coord=None):
    """Extract the haar feature for the current image"""
    ii = integral_image(img)
    return haar_like_feature(ii, 0, 0, ii.shape[0], ii.shape[1],

We use a subset of CBCL dataset which is composed of 100 face images and 100 non-face images. Each image has been resized to a ROI of 19 by 19 pixels. We select 75 images from each group to train a classifier and determine the most salient features. The remaining 25 images from each class are used to assess the performance of the classifier.

images = lfw_subset()
# To speed up the example, extract the two types of features only
feature_types = ['type-2-x', 'type-2-y']

# Build a computation graph using Dask. This allows the use of multiple
# CPU cores later during the actual computation
X = delayed(extract_feature_image(img, feature_types) for img in images)
# Compute the result
t_start = time()
X = np.array(X.compute(scheduler='threads'))
time_full_feature_comp = time() - t_start

# Label images (100 faces and 100 non-faces)
y = np.array([1] * 100 + [0] * 100)

X_train, X_test, y_train, y_test = train_test_split(X, y, train_size=150,

# Extract all possible features
feature_coord, feature_type = \
    haar_like_feature_coord(width=images.shape[2], height=images.shape[1],

A random forest classifier can be trained in order to select the most salient features, specifically for face classification. The idea is to determine which features are most often used by the ensemble of trees. By using only the most salient features in subsequent steps, we can drastically speed up the computation while retaining accuracy.

# Train a random forest classifier and assess its performance
clf = RandomForestClassifier(n_estimators=1000, max_depth=None,
                             max_features=100, n_jobs=-1, random_state=0)
t_start = time()
clf.fit(X_train, y_train)
time_full_train = time() - t_start
auc_full_features = roc_auc_score(y_test, clf.predict_proba(X_test)[:, 1])

# Sort features in order of importance and plot the six most significant
idx_sorted = np.argsort(clf.feature_importances_)[::-1]

fig, axes = plt.subplots(3, 2)
for idx, ax in enumerate(axes.ravel()):
    image = images[0]
    image = draw_haar_like_feature(image, 0, 0,

_ = fig.suptitle('The most important features')
The most important features

We can select the most important features by checking the cumulative sum of the feature importance. In this example, we keep the features representing 70% of the cumulative value (which corresponds to using only 3% of the total number of features).

cdf_feature_importances = np.cumsum(clf.feature_importances_[idx_sorted])
cdf_feature_importances /= cdf_feature_importances[-1]  # divide by max value
sig_feature_count = np.count_nonzero(cdf_feature_importances < 0.7)
sig_feature_percent = round(sig_feature_count /
                            len(cdf_feature_importances) * 100, 1)
print(('{} features, or {}%, account for 70% of branch points in the '
       'random forest.').format(sig_feature_count, sig_feature_percent))

# Select the determined number of most informative features
feature_coord_sel = feature_coord[idx_sorted[:sig_feature_count]]
feature_type_sel = feature_type[idx_sorted[:sig_feature_count]]
# Note: it is also possible to select the features directly from the matrix X,
# but we would like to emphasize the usage of `feature_coord` and `feature_type`
# to recompute a subset of desired features.

# Build the computational graph using Dask
X = delayed(extract_feature_image(img, feature_type_sel, feature_coord_sel)
            for img in images)
# Compute the result
t_start = time()
X = np.array(X.compute(scheduler='threads'))
time_subs_feature_comp = time() - t_start

y = np.array([1] * 100 + [0] * 100)
X_train, X_test, y_train, y_test = train_test_split(X, y, train_size=150,


724 features, or 0.7%, account for 70% of branch points in the random forest.

Once the features are extracted, we can train and test a new classifier.

t_start = time()
clf.fit(X_train, y_train)
time_subs_train = time() - t_start

auc_subs_features = roc_auc_score(y_test, clf.predict_proba(X_test)[:, 1])

summary = (('Computing the full feature set took {:.3f}s, plus {:.3f}s '
            'training, for an AUC of {:.2f}. Computing the restricted '
            'feature set took {:.3f}s, plus {:.3f}s training, '
            'for an AUC of {:.2f}.')
           .format(time_full_feature_comp, time_full_train,
                   auc_full_features, time_subs_feature_comp,
                   time_subs_train, auc_subs_features))



Computing the full feature set took 70.821s, plus 2.550s training, for an AUC of 1.00. Computing the restricted feature set took 0.204s, plus 2.052s training, for an AUC of 1.00.

Total running time of the script: ( 1 minutes 20.743 seconds)

Gallery generated by Sphinx-Gallery