def inspect_number_of_occurrences():
# Get image paths in the training and testing datasets
_, training_image_index_list = prepare_data.get_image_paths_in_training_dataset()
repeated_num = 20
seed_array = np.random.choice(range(repeated_num), size=repeated_num, replace=False)
records_list = Parallel(n_jobs=-1)(
delayed(inspect_final_data_set_without_labels)(training_image_index_list, seed) for seed in seed_array
)
# repeated_num = 100
# seed_array = np.random.choice(range(repeated_num), size=repeated_num, replace=False)
# records_list = (Parallel(n_jobs=-1)(delayed(inspect_final_data_set_with_labels)(training_image_index_list, seed) for seed in seed_array))
true_records_num_list = []
false_records_num_list = []
for single_true_records_num_list, single_false_records_num_list in records_list:
for value in single_true_records_num_list:
true_records_num_list.append(value)
for value in single_false_records_num_list:
false_records_num_list.append(value)
for single_list in [true_records_num_list, false_records_num_list]:
repeated_times_list = []
min_value_list = []
max_value_list = []
mean_value_list = []
for end_index in range(len(single_list)):
current_list = single_list[0 : end_index + 1]
repeated_times_list.append(len(current_list))
min_value_list.append(np.min(current_list))
max_value_list.append(np.max(current_list))
mean_value_list.append(np.mean(current_list))
pylab.figure()
pylab.plot(repeated_times_list, min_value_list, color="yellowgreen", label="Minimum")
pylab.plot(repeated_times_list, max_value_list, color="lightskyblue", label="Maximum")
pylab.plot(repeated_times_list, mean_value_list, color="darkorange", label="Mean")
pylab.legend(bbox_to_anchor=(0.0, 1.02, 1.0, 0.102), loc=3, ncol=3, mode="expand", borderaxespad=0.0)
pylab.xlabel("Repeated Times", fontsize="large")
pylab.ylabel("Number of Occurrences", fontsize="large")
pylab.grid()
pylab.show()
def load_feature(facial_image_extension, feature_extension):
"""Load feature.
:param facial_image_extension: the extension of the facial images
:type facial_image_extension: string
:param feature_extension: the extension of the feature files
:type feature_extension: string
:return: valid_training_image_feature_list refers to the features of the training images,
valid_training_image_index_list refers to the indexes of the training images,
testing_image_feature_dict refers to the features of the testing images which is saved in a dict.
:rtype: tuple
"""
print("\nLoading feature ...")
# Get image paths in the training and testing datasets
image_paths_in_training_dataset, training_image_index_list = prepare_data.get_image_paths_in_training_dataset(
)
image_paths_in_testing_dataset = prepare_data.get_image_paths_in_testing_dataset(
)
# Load feature from file
training_image_feature_list = load_feature_from_file(\
image_paths_in_training_dataset, facial_image_extension, feature_extension)
testing_image_feature_list = load_feature_from_file(\
image_paths_in_testing_dataset, facial_image_extension, feature_extension)
# Omit possible None element in training image feature list
valid_training_image_feature_list = []
valid_training_image_index_list = []
for training_image_feature, training_image_index in zip(
training_image_feature_list, training_image_index_list):
if training_image_feature is not None:
valid_training_image_feature_list.append(training_image_feature)
valid_training_image_index_list.append(training_image_index)
# Generate a dictionary to save the testing image feature
testing_image_feature_dict = {}
for testing_image_feature, testing_image_path in zip(
testing_image_feature_list, image_paths_in_testing_dataset):
testing_image_name = os.path.basename(testing_image_path)
testing_image_feature_dict[testing_image_name] = testing_image_feature
print("Feature loaded successfully.\n")
return (valid_training_image_feature_list, valid_training_image_index_list,
testing_image_feature_dict)
def inspect_number_of_images():
# Get image paths in the training and testing datasets
_, training_image_index_list = prepare_data.get_image_paths_in_training_dataset()
images_number_list = []
for current_image_index in np.unique(training_image_index_list):
images_number_list.append(np.sum(np.array(training_image_index_list) == current_image_index))
# the histogram of the data with histtype="step"
bins = np.arange(np.min(images_number_list), np.max(images_number_list) + 2) - 0.5
_, _, patches = pylab.hist(images_number_list, bins=bins)
pylab.setp(patches, "facecolor", "yellowgreen", "alpha", 0.75)
pylab.xlim([bins[0], bins[-1]])
pylab.xticks(np.arange(np.min(images_number_list), np.max(images_number_list) + 1))
pylab.xlabel("Number of Images from the Same Person", fontsize="large")
pylab.ylabel("Number of Occurrences", fontsize="large")
pylab.title("Histogram of Number of Images from the Same Person")
pylab.show()
def load_feature(facial_image_extension, feature_extension):
"""Load feature.
:param facial_image_extension: the extension of the facial images
:type facial_image_extension: string
:param feature_extension: the extension of the feature files
:type feature_extension: string
:return: valid_training_image_feature_list refers to the features of the training images,
valid_training_image_index_list refers to the indexes of the training images,
testing_image_feature_dict refers to the features of the testing images which is saved in a dict.
:rtype: tuple
"""
print("\nLoading feature ...")
# Get image paths in the training and testing datasets
image_paths_in_training_dataset, training_image_index_list = prepare_data.get_image_paths_in_training_dataset()
image_paths_in_testing_dataset = prepare_data.get_image_paths_in_testing_dataset()
# Load feature from file
training_image_feature_list = load_feature_from_file(
image_paths_in_training_dataset, facial_image_extension, feature_extension
)
testing_image_feature_list = load_feature_from_file(
image_paths_in_testing_dataset, facial_image_extension, feature_extension
)
# Omit possible None element in training image feature list
valid_training_image_feature_list = []
valid_training_image_index_list = []
for training_image_feature, training_image_index in zip(training_image_feature_list, training_image_index_list):
if training_image_feature is not None:
valid_training_image_feature_list.append(training_image_feature)
valid_training_image_index_list.append(training_image_index)
# Generate a dictionary to save the testing image feature
testing_image_feature_dict = {}
for testing_image_feature, testing_image_path in zip(testing_image_feature_list, image_paths_in_testing_dataset):
testing_image_name = os.path.basename(testing_image_path)
testing_image_feature_dict[testing_image_name] = testing_image_feature
print("Feature loaded successfully.\n")
return (valid_training_image_feature_list, valid_training_image_index_list, testing_image_feature_dict)
def inspect_number_of_images():
# Get image paths in the training and testing datasets
_, training_image_index_list = prepare_data.get_image_paths_in_training_dataset(
)
images_number_list = []
for current_image_index in np.unique(training_image_index_list):
images_number_list.append(
np.sum(np.array(training_image_index_list) == current_image_index))
# the histogram of the data with histtype="step"
bins = np.arange(np.min(images_number_list),
np.max(images_number_list) + 2) - 0.5
_, _, patches = pylab.hist(images_number_list, bins=bins)
pylab.setp(patches, "facecolor", "yellowgreen", "alpha", 0.75)
pylab.xlim([bins[0], bins[-1]])
pylab.xticks(
np.arange(np.min(images_number_list),
np.max(images_number_list) + 1))
pylab.xlabel("Number of Images from the Same Person", fontsize="large")
pylab.ylabel("Number of Occurrences", fontsize="large")
pylab.title("Histogram of Number of Images from the Same Person")
pylab.show()
import os
import prepare_data
import solution_basic
import cv2
facial_image_extension = "_open_face.jpg"
feature_extension = "_open_face.csv"
# Get image paths in the training and testing datasets
image_paths_in_training_dataset, training_image_index_list = prepare_data.get_image_paths_in_training_dataset(
)
# Load feature from file
training_image_feature_list = solution_basic.load_feature_from_file(image_paths_in_training_dataset, \
facial_image_extension, feature_extension)
feature_file_paths = [image_path + facial_image_extension + feature_extension \
for image_path in image_paths_in_training_dataset]
# Omit possible None element in training image feature list
invalid_feature_file_path_list = []
for training_image_feature, feature_file_path in zip(
training_image_feature_list, feature_file_paths):
if training_image_feature is None:
invalid_feature_file_path_list.append(feature_file_path)
for invalid_feature_file_path in sorted(invalid_feature_file_path_list):
found_index = invalid_feature_file_path.find(facial_image_extension)
invalid_image_file_path = invalid_feature_file_path[0:found_index]
invalid_image_file_name = os.path.basename(invalid_image_file_path)
def inspect_number_of_occurrences():
# Get image paths in the training and testing datasets
_, training_image_index_list = prepare_data.get_image_paths_in_training_dataset(
)
repeated_num = 20
seed_array = np.random.choice(range(repeated_num),
size=repeated_num,
replace=False)
records_list = (Parallel(n_jobs=-1)(delayed(
inspect_final_data_set_without_labels)(training_image_index_list, seed)
for seed in seed_array))
# repeated_num = 100
# seed_array = np.random.choice(range(repeated_num), size=repeated_num, replace=False)
# records_list = (Parallel(n_jobs=-1)(delayed(inspect_final_data_set_with_labels)(training_image_index_list, seed) for seed in seed_array))
true_records_num_list = []
false_records_num_list = []
for single_true_records_num_list, single_false_records_num_list in records_list:
for value in single_true_records_num_list:
true_records_num_list.append(value)
for value in single_false_records_num_list:
false_records_num_list.append(value)
for single_list in [true_records_num_list, false_records_num_list]:
repeated_times_list = []
min_value_list = []
max_value_list = []
mean_value_list = []
for end_index in range(len(single_list)):
current_list = single_list[0:end_index + 1]
repeated_times_list.append(len(current_list))
min_value_list.append(np.min(current_list))
max_value_list.append(np.max(current_list))
mean_value_list.append(np.mean(current_list))
pylab.figure()
pylab.plot(repeated_times_list,
min_value_list,
color="yellowgreen",
label="Minimum")
pylab.plot(repeated_times_list,
max_value_list,
color="lightskyblue",
label="Maximum")
pylab.plot(repeated_times_list,
mean_value_list,
color="darkorange",
label="Mean")
pylab.legend(bbox_to_anchor=(0., 1.02, 1., .102),
loc=3,
ncol=3,
mode="expand",
borderaxespad=0.)
pylab.xlabel("Repeated Times", fontsize="large")
pylab.ylabel("Number of Occurrences", fontsize="large")
pylab.grid()
pylab.show()
