def internal_comparisons_test(self,
n_stolen,
n_sighted,
n_matches,
quiet=False,
seed=DEF_SEED):
""" Test that the student has correctly counted the code against what
we have counted. This does not mean that the count is correct, just
that it was correctly counted.
setting quiet = True means the feedback summary won't be printed,
which is useful if using along with standard comparisons in
a single test case.
"""
base_file = self.base_filename(n_stolen, n_sighted, n_matches, seed)
(stolen, sighted, _) = utilities.read_dataset(TEST_FOLDER + base_file)
start = time.perf_counter()
_, student_count = self.matching_function(stolen, sighted)
end = time.perf_counter()
delta = end - start
# prints student comparisons and time taken
template = '{}, c={}, {:.4f}s'
feedback = template.format(base_file, student_count, delta)
if not quiet:
print(feedback, end=' ... ')
self.assertEqual(student_count, real_comparisons())
def run_simple_tests():
""" A nice place for your testing code """
filename = './test_data/10s-10000-10-a.txt'
db_list, sighted_list, matches_list = utilities.read_dataset(filename)
table_db_make = make_db_hash_table(db_list, 10, LinearHashTable)
table_process = process_camera_stream(db_list, sighted_list, 20, 5)
def run_tests():
""" Use this function to run some simple tests
to help with developing your awesome answer code.
You should leave this out of your submission """
from utilities import read_dataset
file_name = './test_data/0s-5-0-a.txt'
stolen_list, sighted_list, matches = read_dataset(file_name)
result_list = []
total_comparisons = 0
for plate in sighted_list:
comparisons = search(stolen_list, plate, result_list)
total_comparisons += comparisons
return result_list, total_comparisons
def plates_test(self, n_stolen, n_sighted, n_matches, seed=DEF_SEED):
""" Test that the given matching_function returns the correct
result for the file specified by test_file_name.
"""
base_file = self.base_filename(n_stolen, n_sighted, n_matches, seed)
stolen, sightings, expected_list = utilities.read_dataset(TEST_FOLDER +
base_file)
start = time.perf_counter()
student_answer, comps = self.matching_function(stolen, sightings)
end = time.perf_counter()
delta = end - start
print('{}, c={}, {:.4f}s'.format(base_file, comps, delta), end=' ... ')
self.assertEqual(student_answer, expected_list)
if len(student_answer) > 0:
self.assertTypesEqual(student_answer[0], expected_list[0])
def process_stream_test(self, n_items, n_sighted, n_matches, db_table_size, results_table_size):
""" Note: The condensed_str of results is used when results_table_size > 100 """
self.setUp()
# get expected db table
base_file = self.base_filename(n_items, n_sighted, n_matches)
test_file_name = TEST_FOLDER + base_file
database_list, sightings, matches = utilities.read_dataset(
test_file_name)
expected_db_part = ''
template = '{}expected_db_linear_{}-{}.txt'
txtless_basefile = base_file[:-4] # strips the .txt
expected_db_file_name = template.format(TEST_FOLDER,
txtless_basefile,
db_table_size)
expected_db_table_str = utilities.read_expected(expected_db_file_name)
# get expected results table
template = '{}expected_results_table_{}-{}-{}.txt'
txtless_basefile = base_file[:-4] # strips the .txt
expected_results_file_name = template.format(TEST_FOLDER,
txtless_basefile,
db_table_size,
results_table_size)
expected_results_table_str = utilities.read_expected(
expected_results_file_name)
# get db table and results table
database, results = process_camera_stream(database_list,
sightings,
db_table_size,
results_table_size)
self.assertEqual(str(database), expected_db_table_str)
# expected results for tables larger than 100 will be in condensed form.
if results_table_size <= 100:
self.assertEqual(str(results), expected_results_table_str)
else:
self.assertEqual(results.condensed_str(),
expected_results_table_str)
return True
def run_main(method, fs_functions, score_name, n_clfs=5, dataset_name="PC4"):
print("\nDATASET: %s\nMETHOD: %s\n" % (dataset_name, method))
np.random.seed(1)
##### 1. ------ GET DATASET
X, y, ft_names = ut.read_dataset("datasets/", dataset_name=dataset_name)
##### 2. ------- RUN TRANING METHOD
methods.run_method(method,
X,
y,
n_clfs=n_clfs,
fs_functions=fs_functions,
score_name=score_name)
pl.title(dataset_name)
pl.ylabel(score_name)
pl.legend(loc="best")
img = BytesIO()
pl.savefig(img)
img.seek(0)
return img
def comparisons_test(self,
n_stolen,
n_sighted,
n_matches,
expected=None,
seed=DEF_SEED):
""" Test that the number of comparisons that the student made is
within the expected bounds (provided by self.get_bounds, or expected)
"""
base_file = self.base_filename(n_stolen, n_sighted, n_matches, seed)
stolen, sighted, _ = utilities.read_dataset(TEST_FOLDER + base_file)
start = time.perf_counter()
_, student_count = self.matching_function(stolen, sighted)
end = time.perf_counter()
delta = end - start
print('{}, c={}, {:.4f}s'.format(base_file, student_count, delta),
end=' ... ')
if expected is not None:
self.assertEqual(student_count, expected)
else:
self.check_comparisons_within_bounds(student_count, len(stolen),
len(sighted), n_matches)
def make_db_hash_table_test(self, n_items, n_sighted, n_matches, n_slots):
base_file = self.base_filename(n_items, n_sighted, n_matches)
test_file_name = TEST_FOLDER + base_file
database_list, sightings, matches = utilities.read_dataset(
test_file_name)
if self.table_class == LinearHashTable:
expected_part = 'expected_db_linear_'
elif self.table_class == ChainingHashTable:
expected_part = 'expected_db_chaining_'
else:
expected_part = 'expected_db_table_list_'
template = '{}{}{}-{}.txt'
txtless_basefile = base_file[:-4] # strips the .txt
expected_file_name = template.format(TEST_FOLDER,
expected_part,
txtless_basefile,
n_slots)
with open(expected_file_name) as expected_file:
expected_table_str = expected_file.read()
database = make_db_hash_table(database_list,
n_slots,
table_class=self.table_class)
self.assertEqual(str(database), expected_table_str)
return True
parser.add_argument('-d', '--dataset_name', default="ant")
parser.add_argument('-n', '--n_clfs', default=5, type=int)
parser.add_argument('-s', '--score_name', required=True,
choices=["auc","gmeans"])
args = parser.parse_args()
method = args.method
dataset_name = args.dataset_name
fs_functions = args.fs_functions
n_clfs = args.n_clfs
score_name = args.score_name
print("\nDATASET: %s\nMETHOD: %s\n" % (dataset_name, method))
np.random.seed(1)
##### 1. ------ GET DATASET
X, y, ft_names = ut.read_dataset("datasets/", dataset_name=dataset_name)
pl.title(dataset_name)
pl.ylabel("AUC")
##### 2. ------- RUN TRANING METHOD
methods.run_method(method, X, y, n_clfs=n_clfs,
fs_functions=fs_functions,
score_name=score_name)
pl.legend(loc="best")
pl.show()
import utilities
import os
import operator
inverted_index_file = "inverted_index_5000"
frequency_file = "frequency_5000.csv"
#read the dataset into a dictionary
inverted_dict = {}
inverted_dict = utilities.read_dataset(inverted_index_file)
#Approach 1 : padding keywords such that
# they have the same result length as the most frequent keyword and then identify overhead .
# find the most frequency length & find the real amount of ids
longest = 0
for id_list in inverted_dict.values():
if len(id_list) > longest:
longest = len(id_list)
# append pid & count real & padding
count_real = 0
count_padding = 0
for id_list in inverted_dict.values():
count_real += len(id_list)
if len(id_list) < longest:
diff = longest - len(id_list)
for i in range(diff):
# id_list.add(utilities.generate_random_id())
# because the program runs so slow, we replace the generator with i. But in real case, we will use generate_random_id().
id_list.add(i)
count_padding += 1
transform_sqrt=True,
visualize=True,
block_norm="L1")
return hist
if __name__ == '__main__':
# to hold feature vector for all images
features = []
total_images = 0
# save all feature vectors + label in a csv file
with open("..\\features.csv", 'w', newline="") as f:
writer = csv.writer(f)
# loop over the training images
images_paths, labels = utils.read_dataset(args["training"])
for image_path in images_paths:
image = cv2.imread(image_path)
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
lbp_hist = get_lbp(gray)
rot_lbp_hist = get_rot_lbp(gray)
hog_hist = get_hog(gray)
label = image_path.split("\\")[-2]
features.append(
np.concatenate((lbp_hist, rot_lbp_hist, hog_hist, label),
axis=None))
writer.writerows(features)
features = []
total_images = total_images + 1
parser.add_argument('-n', '--n_clfs', default=5, type=int)
parser.add_argument('-p', '--problem', default="multiclassification",
choices=["multiclassification", "ovr"])
parser.add_argument('-u', '--unbalanced', default=False)
args = parser.parse_args()
method = args.method
fs_functions = args.fs_functions
n_clfs = args.n_clfs
problem = args.problem
use_unbalanced_data = args.unbalanced
np.random.seed(1)
# GET DATASET
feature_matrix, facies_vector = ut.read_dataset()
# draw_data_histogram(facies_vector)
ut.get_feature_statistics(feature_matrix)
# Preprocessing
feature_matrix = preprocessing.normalize(feature_matrix, facies_vector)
# convert the dataset to be a binary classification problem for class x over rest
if problem == "ovr":
facies_vector = ut.convert_to_binary_classification(2, facies_vector)
if not use_unbalanced_data:
feature_matrix, facies_vector = preprocessing.balance(feature_matrix, facies_vector)
# RUN TRANING METHOD
print("Method: ", method)
