Segmenting the picture of a raccoon face in regions

This example uses Spectral clustering on a graph created from voxel-to-voxel difference on an image to break this image into multiple partly-homogeneous regions.

This procedure (spectral clustering on an image) is an efficient approximate solution for finding normalized graph cuts.

There are two options to assign labels:

with ?kmeans? spectral clustering will cluster samples in the embedding space using a kmeans algorithm
whereas ?discrete? will iteratively search for the closest partition space to the embedding space.

print(__doc__)
 
# Author: Gael Varoquaux <gael.varoquaux@normalesup.org>, Brian Cheung
# License: BSD 3 clause
 
import time
 
import numpy as np
import scipy as sp
import matplotlib.pyplot as plt
 
from sklearn.feature_extraction import image
from sklearn.cluster import spectral_clustering
from sklearn.utils.testing import SkipTest
from sklearn.utils.fixes import sp_version
 
if sp_version < (0, 12):
    raise SkipTest("Skipping because SciPy version earlier than 0.12.0 and "
                   "thus does not include the scipy.misc.face() image.")
 
 
# load the raccoon face as a numpy array
try:
    face = sp.face(gray=True)
except AttributeError:
    # Newer versions of scipy have face in misc
    from scipy import misc
    face = misc.face(gray=True)
 
# Resize it to 10% of the original size to speed up the processing
face = sp.misc.imresize(face, 0.10) / 255.
 
# Convert the image into a graph with the value of the gradient on the
# edges.
graph = image.img_to_graph(face)
 
# Take a decreasing function of the gradient: an exponential
# The smaller beta is, the more independent the segmentation is of the
# actual image. For beta=1, the segmentation is close to a voronoi
beta = 5
eps = 1e-6
graph.data = np.exp(-beta * graph.data / graph.data.std()) + eps
 
# Apply spectral clustering (this step goes much faster if you have pyamg
# installed)
N_REGIONS = 25

Visualize the resulting regions

for assign_labels in ('kmeans', 'discretize'):
    t0 = time.time()
    labels = spectral_clustering(graph, n_clusters=N_REGIONS,
                                 assign_labels=assign_labels, random_state=1)
    t1 = time.time()
    labels = labels.reshape(face.shape)
 
    plt.figure(figsize=(5, 5))
    plt.imshow(face, cmap=plt.cm.gray)
    for l in range(N_REGIONS):
        plt.contour(labels == l, contours=1,
                    colors=[plt.cm.spectral(l / float(N_REGIONS))])
    plt.xticks(())
    plt.yticks(())
    title = 'Spectral clustering: %s, %.2fs' % (assign_labels, (t1 - t0))
    print(title)
    plt.title(title)
plt.show()

Out:

  Spectral clustering: kmeans, 6.25s
Spectral clustering: discretize, 5.26s

Total running time of the script: (0 minutes 12.577 seconds)

Download Python source code: plot_face_segmentation.py

Download IPython notebook: plot_face_segmentation.ipynb

Links:

http://scikit-learn.org/stable/auto_examples/cluster/plot_face_segmentation.html

doc_scikit_learn

2025-01-10 15:47:30

Comments