def compute_catalog_fingerprint(catalog_path: str, verbose=True) -> str:
"""Computes the fingerprint of a given catalog
Args:
catalog_path (str): The path of the input catalog.
verbose (bool, optional): [description]. Defaults to True.
Returns:
str: the catalog fingerprint.
"""
# Init. catalog fingerprint
catalog_fingerprint = ""
# Get all paths
image_paths = get_all_images_from_folder(catalog_path)
# Iterator
iterator = get_iterator(image_paths,
verbose=verbose,
description="Computing fingerprint...")
# Loop over image_paths
for image_path in iterator:
# Hash image_path
path_hash = sha224(str.encode(image_path)).hexdigest()
# Update catalog_fingerprint
catalog_fingerprint = f"{catalog_fingerprint}{path_hash}"
# Compute final fingerprint
catalog_fingerprint = sha224(str.encode(catalog_fingerprint)).hexdigest()
return catalog_fingerprint
def predict_batch(self, query_paths: List[str]) -> np.array:
"""Method used to predict a class for a batch of queries.
Args:
query_paths (List[str]): The list of all query paths.
Returns:
np.array: The scores per class for each query.
"""
# Init scores
scores = []
# Get iterator
iterator = utils.get_iterator(query_paths,
verbose=self.config.verbose,
description="Prediction of all queries")
# Loop over all queries
for query_path in iterator:
# Predict score of query
query_scores = self.predict(query_path)
# Update scores
scores.append(query_scores)
# To numpy
scores = np.array(scores)
return scores
def _get_catalog_descriptors(self):
# Init descriptors list
catalog_descriptors = []
# Init iterator
iterator = utils.get_iterator(
utils.get_all_images_from_folder(self.catalog_path),
verbose=self.config.verbose,
description="Computing catalog descriptors")
# Compute all descriptors
for path in iterator:
# Read image
img = utils.read_image(path, size=self.config.image_size)
# Compute keypoints
keypoints = utils.compute_keypoints(img,
self.config.keypoint_stride,
self.config.keypoint_sizes)
# Compute descriptors
descriptors = utils.compute_descriptors(
img, keypoints, self.config.feature_descriptor)
# Update descriptors list
catalog_descriptors.append(descriptors)
# Reshape descriptors list
catalog_descriptors = np.array(catalog_descriptors)
catalog_descriptors = catalog_descriptors.reshape(
-1, catalog_descriptors.shape[-1])
return catalog_descriptors
def test_verbose():
"""[summary]
"""
# List
test_list = range(10)
# Get iterator
iterator = get_iterator(test_list, verbose=True)
# Assert
assert isinstance(iterator, tqdm)
def test_no_verbose():
"""[summary]
"""
# List
test_list = range(10)
# Get iterator
iterator = get_iterator(test_list, verbose=False)
# Assert
assert isinstance(iterator, type(test_list))
def _create_matcher_index(self):
# Get descriptors
catalog_descriptors = self._get_catalog_descriptors()
# Get iterator
descriptors_iterator = utils.get_iterator(
catalog_descriptors,
verbose=self.config.verbose,
description="Creating Index...")
# Config matcher
for k, descriptor in enumerate(descriptors_iterator):
self.matcher.add_item(k, descriptor)
self.matcher.build(self.config.matcher_n_trees)
def test_with_description():
"""[summary]
"""
# List
test_list = range(10)
# Description
test_description = "Test Description"
# Get iterator
iterator = get_iterator(test_list,
verbose=True,
description=test_description)
# Assert
assert iterator.desc == test_description
def compute_catalog_embeddings(self) -> np.array:
"""[summary]
Returns:
np.array: [description]
"""
# Init. catalog embeddings
self.catalog_embeddings = []
# Loop over catalog images
for catalog_img_path in utils.get_iterator(self.catalog_images,
verbose=self.config.verbose):
# Read catalog image
catalog_image = utils.read_image(catalog_img_path,
size=self.config.image_size)
catalog_image = np.expand_dims(catalog_image, axis=0)
# Compute embedding
catalog_emdding = self.triplet_model.predict(catalog_image)[0]
# Update catalog_emddings
self.catalog_embeddings.append(catalog_emdding)
self.catalog_embeddings = np.array(self.catalog_embeddings)