def test_python_pca(self):
'''Make sure that python implementation matches sklearn's.'''
X = lfw_subset()[:10, :, :].reshape((10, -1)).transpose((1, 0))
n_components = 2
Y_py = python_pca(X, n_components)
Y_ski = PCA(n_components=n_components, whiten=False).fit_transform(X)
# Same within scale factor of -1
self.assertTrue(np.allclose(np.abs(Y_py), np.abs(Y_ski)))
def test_haar_features():
print("\nloading lfw subset\n")
images = lfw_subset()
feature_types = ['type-2-x', 'type-2-y']
X, feature_coords, feature_types = compute_haar_features(
feature_types, images)
save_to_disk([X, feature_coords, feature_types],
"sav_files/lfw_subset_t2_features.sav")
def test_python_pca(self):
dim = 4
# X = np.random.normal(1,.1,(dim,dim))
X = lfw_subset()[:10,:,:].reshape((10,-1)).transpose((1,0))
n_components = 2
Y_py = python_pca(X,n_components)
Y_ski = PCA(n_components=n_components,whiten=False).fit_transform(X)
# print(Y_py)
# print(Y_ski)
# Same within scale factor of -1
self.assertTrue(np.allclose(np.abs(Y_py),np.abs(Y_ski)))
######################################################################
#
# Stereo images
# =============
fig, axes = plt.subplots(1, 2, figsize=(8, 4))
ax = axes.ravel()
images = data.stereo_motorcycle()
ax[0].imshow(images[0])
ax[1].imshow(images[1])
fig.tight_layout()
plt.show()
######################################################################
#
# Faces and non-faces dataset
# ===========================
#
# A sample of 20 over 200 images is displayed.
fig, axes = plt.subplots(4, 5, figsize=(20, 20))
ax = axes.ravel()
images = data.lfw_subset()
for i in range(20):
ax[i].imshow(images[90 + i], cmap=plt.cm.gray)
ax[i].axis('off')
fig.tight_layout()
plt.show()
0,
0,
ii.shape[0],
ii.shape[1],
feature_type=feature_type,
feature_coord=feature_coord)
###############################################################################
# We will use a subset of the CBCL 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
# will keep 75 images from each group to train a classifier and check which
# extracted features are the most salient, and use the remaining 25 from each
# class to check the performance of the classifier.
images = lfw_subset()
# For speed, only extract the two first types of features
feature_types = ['type-2-x', 'type-2-y']
# Build a computation graph using dask. This allows using multiple CPUs for
# the computation step
X = delayed(extract_feature_image(img, feature_types) for img in images)
# Compute the result using the "processes" dask backend
t_start = time()
X = np.array(X.compute(scheduler='processes'))
time_full_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,
random_state=0,
def test_lfw_subset():
""" Test that "lfw_subset" can be loaded."""
data.lfw_subset()
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],
feature_type=feature_type,
feature_coord=feature_coord)
###############################################################################
# We will use a subset of the CBCL 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
# will keep 75 images from each group to train a classifier and check which
# extracted features are the most salient, and use the remaining 25 from each
# class to check the performance of the classifier.
images = lfw_subset()
# For speed, only extract the two first types of features
feature_types = ['type-2-x', 'type-2-y']
# Build a computation graph using dask. This allows using multiple CPUs for
# the computation step
X = delayed(extract_feature_image(img, feature_types)
for img in images)
# Compute the result using the "processes" dask backend
t_start = time()
X = np.array(X.compute(scheduler='processes'))
time_full_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,
random_state=0,
stratify=y)
##### from example to see how to do it #type(result) #images = lfw_subset() #type(images) #print(np.shape(images)) #images #images = lfw_subset() #print(type(images)) #print(np.shape(images)) print(np.shape(fullset)) print(lfw_subset()[1].max()) #print( lfw_subset()[1]) #print(fullset[1]/255.9) images = fullset/255.9 # rescaling to 1 so we avoid getting Nan in "X" #images = lfw_subset() #### new to avoid error perhaps, doesn't work #images= fullset/(255.9*100000) images[1] plt.imshow(images[200]) images[5]
