def test_load_fake_lfw_pairs():
lfw_pairs_train = fetch_lfw_pairs(data_home=SCIKIT_LEARN_DATA,
download_if_missing=False)
# The data is croped around the center as a rectangular bounding box
# around the face. Colors are converted to gray levels:
assert_equal(lfw_pairs_train.pairs.shape, (10, 2, 62, 47))
# the target is whether the person is the same or not
assert_array_equal(lfw_pairs_train.target, [1, 1, 1, 1, 1, 0, 0, 0, 0, 0])
# names of the persons can be found using the target_names array
expected_classes = ['Different persons', 'Same person']
assert_array_equal(lfw_pairs_train.target_names, expected_classes)
# It is possible to ask for the original data without any croping or color
# conversion
lfw_pairs_train = fetch_lfw_pairs(data_home=SCIKIT_LEARN_DATA, resize=None,
slice_=None, color=True,
download_if_missing=False)
assert_equal(lfw_pairs_train.pairs.shape, (10, 2, 250, 250, 3))
# the ids and class names are the same as previously
assert_array_equal(lfw_pairs_train.target, [1, 1, 1, 1, 1, 0, 0, 0, 0, 0])
assert_array_equal(lfw_pairs_train.target_names, expected_classes)
def test_load_fake_lfw_pairs():
lfw_pairs_train = fetch_lfw_pairs(data_home=SCIKIT_LEARN_DATA,
download_if_missing=False)
# The data is croped around the center as a rectangular bounding box
# around the face. Colors are converted to gray levels:
assert lfw_pairs_train.pairs.shape == (10, 2, 62, 47)
# the target is whether the person is the same or not
assert_array_equal(lfw_pairs_train.target, [1, 1, 1, 1, 1, 0, 0, 0, 0, 0])
# names of the persons can be found using the target_names array
expected_classes = ["Different persons", "Same person"]
assert_array_equal(lfw_pairs_train.target_names, expected_classes)
# It is possible to ask for the original data without any croping or color
# conversion
lfw_pairs_train = fetch_lfw_pairs(
data_home=SCIKIT_LEARN_DATA,
resize=None,
slice_=None,
color=True,
download_if_missing=False,
)
assert lfw_pairs_train.pairs.shape == (10, 2, 250, 250, 3)
# the ids and class names are the same as previously
assert_array_equal(lfw_pairs_train.target, [1, 1, 1, 1, 1, 0, 0, 0, 0, 0])
assert_array_equal(lfw_pairs_train.target_names, expected_classes)
assert lfw_pairs_train.DESCR.startswith(
".. _labeled_faces_in_the_wild_dataset:")
def LoadData(im_dim=150, im_ch=3, method='train'):
from sklearn.datasets import fetch_lfw_pairs
import warnings
start = 0
stop = 250
color = False
if im_dim < 250:
start = (250 - im_dim) // 2
stop = start + im_dim
if im_ch > 1:
color = True
warnings.filterwarnings("ignore")
lfw_people = fetch_lfw_pairs(subset=method,
slice_=(slice(start, stop,
None), slice(start, stop,
None)),
resize=1,
color=color)
lfw_people.pairs = lfw_people.pairs / 255.
if method == 'train':
lfw_people.pairs = np.concatenate(
(lfw_people.pairs,
np.swapaxes(
np.array([
lfw_people.pairs[:, 1, ...], lfw_people.pairs[:, 0, ...]
]), 0, 1)), 0)
lfw_people.target = np.concatenate(
(lfw_people.target, lfw_people.target), 0)
return lfw_people
def prepare_data():
lfw_pairs_train = fetch_lfw_pairs(subset='train')
images = lfw_pairs_train.pairs
targets = lfw_pairs_train.target
# print "lfw_pairs_train.pairs.shape: ", lfw_pairs_train.pairs.shape
# print "images.shape: ", images.shape
# print "targets.shape: ", targets.shape
# for i in range(2):
# subplot(1, 2, i + 1), imshow(images[202,i,:,:], cmap=cm.gray)
# show()
return images, targets
def __init__(self,
train=True,
color=False,
funneled=True,
resize=1.0,
transform=None,
should_invert=True):
splitName = 'train' if train else 'test'
self.data = fetch_lfw_pairs(subset=splitName,
funneled=funneled,
resize=resize,
color=color)
print("Wild Faces")
print(self.data.pairs.shape)
self.image_len = len(self.data.target)
self.train = train
self.transform = transform
self.should_invert = should_invert
self.color = color
def test_load_empty_lfw_pairs():
fetch_lfw_pairs(data_home=SCIKIT_LEARN_EMPTY_DATA, download_if_missing=False)
def test_load_empty_lfw_pairs():
with pytest.raises(IOError):
fetch_lfw_pairs(data_home=SCIKIT_LEARN_EMPTY_DATA,
download_if_missing=False)
def test_load_empty_lfw_pairs():
fetch_lfw_pairs(data_home=SCIKIT_LEARN_EMPTY_DATA,
download_if_missing=False)
# -*- encoding: utf-8 -*- """ 8.2.1 Scikit-learn自带的数据集 """ from sklearn import datasets as dss import matplotlib.pyplot as plt lfwp = dss.fetch_lfw_pairs() # 加载数据集 print(lfwp.keys()) # 数据集带有若干子集 print(lfwp.data.shape, lfwp.data.dtype) # data子集有2200个样本 print(lfwp.pairs.shape, lfwp.pairs.dtype) # pairs子集有2200个样本，每个样本2张图片 print(lfwp.target_names) # 有两个标签：不是同一个人、是同一个人 print(lfwp.target.shape, lfwp.target.dtype) # 2200个样本的标签，表示样本是否是一个人 plt.subplot(121) plt.imshow(lfwp.pairs[0,0], cmap=plt.cm.gray) plt.subplot(122) plt.imshow(lfwp.pairs[0,1], cmap=plt.cm.gray) plt.show()
# python 3.7 # Scikit-learn ver. 0.23.2 import sklearn from sklearn import datasets from sklearn import preprocessing from sklearn.preprocessing import scale from sklearn.linear_model import LogisticRegression from sklearn.model_selection import train_test_split from sklearn import metrics # matplotlib 3.3.1 from matplotlib import pyplot irises = datasets.load_iris() wine = datasets.load_wine() facePairs = datasets.fetch_lfw_pairs() breastCancer = datasets.load_breast_cancer() covertypes = datasets.fetch_covtype() diabetes = datasets.load_diabetes() olvettiFaces = datasets.fetch_olivetti_faces() # change this to test different datasets dataset = wine dimensionality = 13 X = dataset.data.reshape(len(dataset.data), dimensionality) Y = dataset.target scaled = preprocessing.MinMaxScaler() scaled_data = scaled.fit_transform(X) trainX, testX, trainY, testY = train_test_split(X,
def __init__(self,
batch_size=64,
resize=0.3,
validation_per=0.1,
epochs=10,
iterations=5):
"""
Class that handles creating new model, retreiving dataset (train and test),
training and testing the model.
:param batch_size: choose the size of batch for the training process
:param resize: resize the images (float between 0 and 1)
:param validation_per: percent of the validation set
:param epochs: number of epoch for each iteration
:param iterations: number of iterations
"""
self.lfw_pairs_train = fetch_lfw_pairs(subset='train',
funneled=False,
resize=resize,
slice_=None,
color=False)
self.lfw_pairs_train.pairs = self.lfw_pairs_train.pairs / 255
shape = self.lfw_pairs_train.pairs.shape[2]
K.clear_session()
learning_rate = 10e-4
# l2-regularization penalization for each layer
self.l2_penalization = self.def_penalization()
self.input_shape = (shape, shape, 1)
convolutional_net = self.create_conv_net()
# Now the pairs of images
input_image_1 = Input(self.input_shape)
input_image_2 = Input(self.input_shape)
encoded_image_1 = convolutional_net(input_image_1)
encoded_image_2 = convolutional_net(input_image_2)
# L1 distance layer between the two encoded outputs
# One could use Subtract from Keras, but we want the absolute value
# l1_distance_layer = Lambda(lambda tensors: K.abs(tensors[0] - tensors[1]))
distance = Lambda(euclidean_distance,
output_shape=eucl_dist_output_shape)(
[encoded_image_1, encoded_image_2])
self.model = Model([input_image_1, input_image_2], distance)
# l1_distance = l1_distance_layer([encoded_image_1, encoded_image_2])
rms = RMSprop()
opt = Adam(lr=learning_rate)
self.model.compile(loss=contrastive_loss,
optimizer=rms,
metrics=['binary_accuracy'])
self.train_network(batch_size, epochs, iterations, validation_per)
lfw_pairs_test = fetch_lfw_pairs(subset='test',
funneled=False,
resize=shape / 250,
slice_=None,
color=False)
lfw_pairs_test.pairs = lfw_pairs_test.pairs / 255
x_test, y_test, _, _ = train_validation_pairs(lfw_pairs_test, valid=0)
self.model.evaluate(x_test, y_test)
def load_data():
train_set = fetch_lfw_pairs(subset='train')
test_set = fetch_lfw_pairs(subset='test')
return train_set.data, test_set.data, train_set.target, test_set.target
def get_dataset(self):
return fetch_lfw_pairs()
def __init__(self):
self.flp = fetch_lfw_pairs(
subset='test',
color=True,
resize=1 # this transform inputs to (125, 94) from (62, 47)
)
