def test_feature_extraction_with_bow_with_export_and_load(
self, detector: str):
k = 3
dst = test_utils.get_test_output_csv_route()
vocabulary_file = test_utils.get_test_output_ndarray_route()
runner = CliRunner()
result: Result = runner.invoke(cli, [
'extract', '--bow', '--size', 64, '--detector', detector, '--src',
test_utils.get_test_base_route(), '--dst', dst, '--k', k,
'--export', '--vocabulary-route', vocabulary_file
])
test_utils.assert_validation_test(result)
vocabulary = np.load(vocabulary_file)
assert (vocabulary.shape[0] == k)
test_utils.clean_test_output_csv_route()
result: Result = runner.invoke(cli, [
'extract', '--bow', '--size', 64, '--detector', detector, '--src',
test_utils.get_test_base_route(), '--dst', dst, '--k', k, '--load',
'--vocabulary-route', vocabulary_file
])
test_utils.assert_validation_test(result)
def test_features_are_extracted_using_a_loaded_model(self, method: str):
output_file = test_utils.get_test_output_csv_route()
vocabulary_file = test_utils.get_test_output_ndarray_route()
extractor = BoWExtractor(base_route=test_utils.get_test_base_route(),
k=3,
method=method,
size=64,
cluster_mode='manual')
extractor.setup()
extractor.fit()
extractor.export(vocabulary_file)
new_extractor = BoWExtractor(
base_route=test_utils.get_test_base_route(),
k=3,
method=method,
size=64,
cluster_mode='manual')
new_extractor.setup()
new_extractor.load(vocabulary_file)
new_extractor.extract_and_save(output_file)
features = test_utils.load_csv_from_route(output_file)
expected_width = extractor.k + 1
expected_height = test_utils.count_test_images()
assert (features.shape == (expected_height, expected_width))
def test_image_is_converted_into_matrix_of_selected_size(self):
extractor = LBPExtractor(test_utils.get_test_base_route(),
size=64,
points=3,
radius=1,
grid_x=2,
grid_y=2)
image = extractor._read_image(test_utils.get_test_image_route())
assert (image.shape == (extractor.width, extractor.height))
def test_feature_extraction_with_bow(self, detector: str):
dst = test_utils.get_test_output_csv_route()
runner = CliRunner()
result: Result = runner.invoke(cli, [
'extract', '--bow', '--size', 64, '--detector', detector, '--src',
test_utils.get_test_base_route(), '--dst', dst, '--k', 2
])
test_utils.assert_validation_test(result)
def test_feature_extraction_with_deep_models(self, cnn: str):
dst = test_utils.get_test_output_csv_route()
runner = CliRunner()
result: Result = runner.invoke(cli, [
'extract', '--deep', '--cnn', 'vgg19', '--size', 64, '--src',
test_utils.get_test_base_route(), '--dst', dst
])
test_utils.assert_validation_test(result)
def test_local_binary_patterns_are_extracted_from_image(self):
extractor = LBPExtractor(test_utils.get_test_base_route(),
size=64,
points=3,
radius=1,
grid_x=2,
grid_y=2)
image = extractor._read_image(test_utils.get_test_image_route())
lbp = extractor._extract_lbp(image)
assert (lbp.shape == (extractor.width, extractor.height))
def test_single_image_features_extraction(self, x, y):
extractor = LBPExtractor(test_utils.get_test_base_route(),
size=64,
points=3,
radius=1,
grid_x=x,
grid_y=y)
features = extractor.extract(test_utils.get_test_image_route())
expected_size = x * y * extractor.bins
assert (features.shape[0] == expected_size)
def test_features_are_extracted_from_a_given_image_route(self, model: str):
extractor = DeepExtractor(base_route=test_utils.get_test_base_route(),
model_name=model,
size=75)
features = extractor.extract(
image_route=test_utils.get_test_image_route())
output_shape = extractor.model.layers[-1].output_shape
expected_shape = output_shape[1] * output_shape[2] * output_shape[3]
assert (features.shape[0] == expected_shape)
def test_extraction_setup(self, method: str):
extractor = BoWExtractor(base_route=test_utils.get_test_base_route(),
k=3,
method=method,
size=64)
extractor.setup()
descriptors = extractor.bow_kmeans.getDescriptors()
assert (len(descriptors) > 0)
assert (len(descriptors) <= len(extractor.image_keypoints.keys()))
def test_keypoints_and_descriptors_detection(self, method: str):
extractor = BoWExtractor(base_route=test_utils.get_test_base_route(),
k=3,
method=method,
size=64)
descriptors, keypoints = extractor._process_image(
image_route=test_utils.get_test_image_route())
assert (len(descriptors) > 0)
assert (keypoints.shape[0] == len(descriptors))
assert (keypoints.shape[1] == extractor.detector.descriptorSize())
def test_feature_extraction_with_lbp(self):
dst = test_utils.get_test_output_csv_route()
runner = CliRunner()
result: Result = runner.invoke(cli, [
'extract', '--lbp', '--size', 64, '--grid', 4, '--points', 8,
'--radius', 1, '--src',
test_utils.get_test_base_route(), '--dst', dst
])
test_utils.assert_validation_test(result)
def test_kmeans_model_is_loaded_from_file(self, method: str):
vocabulary_file = test_utils.get_test_output_ndarray_route()
extractor = BoWExtractor(base_route=test_utils.get_test_base_route(),
k=3,
method=method,
size=64,
cluster_mode='manual')
extractor.setup()
extractor.fit()
extractor.export(vocabulary_file)
new_extractor = BoWExtractor(
base_route=test_utils.get_test_base_route(),
k=3,
method=method,
size=64,
cluster_mode='manual')
new_extractor.setup()
new_extractor.load(vocabulary_file)
assert (np.array_equal(new_extractor.bow_extractor.getVocabulary(),
extractor.bow_extractor.getVocabulary()))
def test_feature_detection_and_addition(self, method: str):
image_route = test_utils.get_test_image_route()
extractor = BoWExtractor(base_route=test_utils.get_test_base_route(),
k=3,
method=method,
size=64)
extractor._extract_and_add_image(image_route)
descriptors = extractor.bow_kmeans.getDescriptors()
expected_keypoints, expected_descriptors = extractor._process_image(
image_route)
assert (len(descriptors[0]) == len(expected_descriptors))
assert (len(descriptors[0]) == len(expected_keypoints))
def test_histogram_is_generated_from_local_binary_patterns(self, x, y):
extractor = LBPExtractor(test_utils.get_test_base_route(),
size=64,
points=3,
radius=1,
grid_x=x,
grid_y=y)
image = extractor._read_image(test_utils.get_test_image_route())
lbp = extractor._extract_lbp(image)
histogram = extractor._convert_lbp_to_histogram(lbp)
assert (len(histogram) == x * y)
for histogram_row in histogram:
assert (histogram_row.shape[0] == extractor.bins)
def test_features_are_extracted_from_a_given_image_route(self, model: str):
output_csv = test_utils.get_test_output_csv_route()
extractor = DeepExtractor(base_route=test_utils.get_test_base_route(),
model_name=model,
size=75)
extractor.extract_and_save(output_csv)
features = test_utils.load_csv_from_route(output_csv)
cnn_output_shape = extractor.model.layers[-1].output_shape
expected_width = cnn_output_shape[1] * cnn_output_shape[
2] * cnn_output_shape[3] + 1
expected_height = test_utils.count_test_images()
assert (features.shape == (expected_height, expected_width))
def test_kmeans_model_is_exported(self, method: str):
vocabulary_file = test_utils.get_test_output_ndarray_route()
extractor = BoWExtractor(base_route=test_utils.get_test_base_route(),
k=3,
method=method,
size=64,
cluster_mode='manual')
extractor.setup()
extractor.fit()
extractor.export(vocabulary_file)
assert (os.path.exists(vocabulary_file))
vocabulary = np.load(vocabulary_file)
assert (vocabulary.shape[0] == extractor.k)
assert (np.array_equal(vocabulary,
extractor.bow_extractor.getVocabulary()))
def test_features_are_extracted_from_a_given_image_route(self, x, y):
output_csv = test_utils.get_test_output_csv_route()
base_route = test_utils.get_test_base_route()
extractor = LBPExtractor(base_route=base_route,
size=64,
points=3,
radius=1,
grid_x=x,
grid_y=y)
extractor.extract_and_save(output_csv)
features = test_utils.load_csv_from_route(output_csv)
expected_width = x * y * extractor.bins + 1 # Number of features + label column
expected_height = test_utils.count_test_images()
assert (features.shape == (expected_height, expected_width))
def test_features_are_extracted_from_a_given_image_route_in_manual_mode(
self, method: str):
output_file = test_utils.get_test_output_csv_route()
extractor = BoWExtractor(base_route=test_utils.get_test_base_route(),
k=3,
method=method,
size=64,
cluster_mode='manual')
extractor.setup()
extractor.fit()
extractor.extract_and_save(output_file)
features = test_utils.load_csv_from_route(output_file)
expected_width = extractor.k + 1
expected_height = test_utils.count_test_images()
assert (features.shape == (expected_height, expected_width))
def test_features_are_extracted_from_a_given_image_route_in_automatic_mode(
self, method: str):
output_file = test_utils.get_test_output_csv_route()
extractor = BoWExtractor(base_route=test_utils.get_test_base_route(),
method=method,
size=64,
cluster_mode='auto',
min_k=2,
max_k=100,
step=1,
threshold=0.95)
extractor.setup()
extractor.fit()
extractor.extract_and_save(output_file)
features = test_utils.load_csv_from_route(output_file)
expected_width = extractor.k + 1
expected_height = test_utils.count_test_images()
assert (features.shape == (expected_height, expected_width))
