diff --git a/bob/bio/face/config/database/gbu_bad.py b/bob/bio/face/config/database/gbu_bad.py
index 06354486d2e4336c7cdba99b257d0b02e514db1d..0b4454ef44fca458abcffb8b2adb8637f3845135 100644
--- a/bob/bio/face/config/database/gbu_bad.py
+++ b/bob/bio/face/config/database/gbu_bad.py
@@ -1,11 +1,7 @@
#!/usr/bin/env python
-from bob.extension import rc
from bob.bio.face.database import GBUDatabase
-
-mbgc_v1_directory = rc["bob.bio.face.gbu.directory"]
-
database = GBUDatabase(protocol="Bad")
diff --git a/bob/bio/face/config/database/gbu_good.py b/bob/bio/face/config/database/gbu_good.py
index 521bb9e9c453238fca3223cbe34cd63210b6a746..8d6c5d97a9ed739a2e81ec70844e52677a32079d 100644
--- a/bob/bio/face/config/database/gbu_good.py
+++ b/bob/bio/face/config/database/gbu_good.py
@@ -1,11 +1,7 @@
#!/usr/bin/env python
-from bob.extension import rc
from bob.bio.face.database import GBUDatabase
-
-mbgc_v1_directory = rc["bob.bio.face.gbu.directory"]
-
database = GBUDatabase(protocol="Good")
diff --git a/bob/bio/face/config/database/gbu_ugly.py b/bob/bio/face/config/database/gbu_ugly.py
index 8c271788e53b8098a1065d2019a42dc95ea084c3..5da93feb6a87ea5b5f788779a6e320c407854a99 100644
--- a/bob/bio/face/config/database/gbu_ugly.py
+++ b/bob/bio/face/config/database/gbu_ugly.py
@@ -1,11 +1,7 @@
#!/usr/bin/env python
-from bob.extension import rc
from bob.bio.face.database import GBUDatabase
-
-mbgc_v1_directory = rc["bob.bio.face.gbu.directory"]
-
database = GBUDatabase(protocol="Ugly")
diff --git a/bob/bio/face/config/database/lfw_restricted.py b/bob/bio/face/config/database/lfw_restricted.py
deleted file mode 100644
index 0684999626b9c46327fcd9edbc8b445a56623621..0000000000000000000000000000000000000000
--- a/bob/bio/face/config/database/lfw_restricted.py
+++ /dev/null
@@ -1,29 +0,0 @@
-#!/usr/bin/env python
-
-from bob.bio.face.database import LFWBioDatabase
-from bob.bio.base.pipelines.vanilla_biometrics import DatabaseConnector
-from bob.extension import rc
-
-
-lfw_directory = rc["bob.db.lfw.directory"]
-
-database = DatabaseConnector(
- LFWBioDatabase(
- original_directory=lfw_directory,
- annotation_type="funneled",
- protocol="view1",
- training_depends_on_protocol=True,
- models_depend_on_protocol=True,
- all_files_options={"world_type": "restricted"},
- extractor_training_options={
- "world_type": "restricted"
- }, # 'subworld' : 'twofolds'
- projector_training_options={
- "world_type": "restricted"
- }, # 'subworld' : 'twofolds'
- enroller_training_options={
- "world_type": "restricted"
- }, # 'subworld' : 'twofolds'
- ),
- allow_scoring_with_all_biometric_references=False,
-)
diff --git a/bob/bio/face/config/database/lfw_unrestricted.py b/bob/bio/face/config/database/lfw_unrestricted.py
deleted file mode 100644
index 39ebafb4ee14cbfb47f226ba4f0e07a9c95fc09f..0000000000000000000000000000000000000000
--- a/bob/bio/face/config/database/lfw_unrestricted.py
+++ /dev/null
@@ -1,29 +0,0 @@
-#!/usr/bin/env python
-
-from bob.bio.face.database import LFWBioDatabase
-from bob.bio.base.pipelines.vanilla_biometrics import DatabaseConnector
-from bob.extension import rc
-
-
-lfw_directory = rc["bob.db.lfw.directory"]
-
-database = DatabaseConnector(
- LFWBioDatabase(
- original_directory=lfw_directory,
- annotation_type="funneled",
- protocol="view1",
- training_depends_on_protocol=True,
- models_depend_on_protocol=True,
- all_files_options={"world_type": "unrestricted"},
- extractor_training_options={
- "world_type": "unrestricted"
- }, # 'subworld' : 'twofolds'
- projector_training_options={
- "world_type": "unrestricted"
- }, # 'subworld' : 'twofolds'
- enroller_training_options={
- "world_type": "unrestricted"
- }, # 'subworld' : 'twofolds'
- ),
- allow_scoring_with_all_biometric_references=False,
-)
diff --git a/bob/bio/face/config/database/lfw_view2.py b/bob/bio/face/config/database/lfw_view2.py
new file mode 100644
index 0000000000000000000000000000000000000000..719d763e1f6538250b711e22714472c481ee3aa0
--- /dev/null
+++ b/bob/bio/face/config/database/lfw_view2.py
@@ -0,0 +1,3 @@
+from bob.bio.face.database import LFWDatabase
+
+database = LFWDatabase(protocol="view2")
diff --git a/bob/bio/face/database/__init__.py b/bob/bio/face/database/__init__.py
index 84568b99302305b2ac8a76413a5709e641220220..90c77f948269ca42352489ebba0d5e94eafdfc7d 100644
--- a/bob/bio/face/database/__init__.py
+++ b/bob/bio/face/database/__init__.py
@@ -6,7 +6,8 @@ from .mobio import MobioDatabase
from .replay import ReplayBioDatabase
from .atnt import AtntBioDatabase
from .gbu import GBUDatabase
-from .lfw import LFWBioDatabase
+from .arface import ARFaceDatabase
+from .lfw import LFWDatabase
from .multipie import MultipieDatabase
from .ijbc import IJBCDatabase
from .replaymobile import ReplayMobileBioDatabase
@@ -46,7 +47,8 @@ __appropriate__(
ReplayBioDatabase,
AtntBioDatabase,
GBUDatabase,
- LFWBioDatabase,
+ ARFaceDatabase,
+ LFWDatabase,
MultipieDatabase,
ReplayMobileBioDatabase,
FargoBioDatabase,
diff --git a/bob/bio/face/database/lfw.py b/bob/bio/face/database/lfw.py
index 1b8927307fe814dc4c743845da953e7496df7847..37985b03abc4f19c7781c4c33a51e2df38ac5fd4 100644
--- a/bob/bio/face/database/lfw.py
+++ b/bob/bio/face/database/lfw.py
@@ -3,76 +3,359 @@
# Tiago de Freitas Pereira <tiago.pereira@idiap.ch>
# Sat 20 Aug 15:43:10 CEST 2016
-"""
- LFW database implementation of bob.bio.base.database.Database interface.
- It is an extension of an SQL-based database interface, which directly talks to LFW database, for
- verification experiments (good to use in bob.bio.base framework).
-"""
+from bob.bio.base.pipelines.vanilla_biometrics.abstract_classes import Database
+from bob.pipelines import DelayedSample, SampleSet
+import os
+from functools import partial
+from bob.extension import rc
+import bob.io.image
+import copy
+import logging
+import numpy as np
-from .database import FaceBioFile
-from bob.bio.base.database import BioDatabase
+logger = logging.getLogger(__name__)
-class LFWBioFile(FaceBioFile):
+class LFWDatabase(Database):
+ """
+ This package contains the access API and descriptions for the `Labeled Faced in the Wild <http://vis-www.cs.umass.edu/lfw>`_ (LFW) database.
+ It only contains the Bob_ accessor methods to use the DB directly from python, with our certified protocols.
+ The actual raw data for the database should be downloaded from the original URL (though we were not able to contact the corresponding Professor).
- def __init__(self, f):
- super(LFWBioFile, self).__init__(client_id=f.client_id, path=f.path, file_id=f.id)
- self._f = f
+ The LFW database provides two different sets (called "views").
+ The first one, called ``view1`` is used for optimizing meta-parameters of your algorithm.
+ The second one, called ``view2`` is used for benchmarking.
+ This interface supports only the ``view2`` protocol.
+ Please note that in ``view2`` there is only a ``'dev'`` group, but no ``'eval'``.
+
+
+ .. warning::
+
+ To use this dataset protocol, you need to have the original files of the LFW datasets.
+ Once you have it downloaded, please run the following command to set the path for Bob
+
+ .. code-block:: sh
+
+ bob config set bob.bio.face.lfw.directory [LFW PATH]
+ bob config set bob.bio.face.lfw.annotation_directory [LFW ANNOTATION_PATH] # for the annotations
+
+
+
+ .. code-block:: python
+
+ >>> from bob.bio.face.database import LFWDatabase
+ >>> lfw = LFWDatabase(protocol="view2")
+ >>>
+ >>> # Fetching the gallery
+ >>> references = lfw.references()
+ >>> # Fetching the probes
+ >>> probes = lfw.probes()
+
+
+
+ Parameters
+ ----------
+
+ protocol: str
+ One of the database protocols. Options are `view2`
+
+ annotation_type: str
+ Type of the annotations used for face crop. Default to `eyes-center`
+
+ image_relative_path: str
+ LFW provides several types image crops. Some with the full image, some with with specific
+ face crop. Use this variable to set which image crop you want. Default to `all_images`, which means
+ no crop.
+
+ annotation_directory: str
+ LFW annotations path. Default to what is set in the variable `bob.bio.face.lfw.directory`
+
+ original_directory: str
+ LFW phisical path. Default to what is set in the variable `bob.bio.face.lfw.directory`
+
+ annotation_issuer: str
+ Type of the annotations. Default to `funneled`. Possible types `funneled` or `idiap`
-class LFWBioDatabase(BioDatabase):
- """
- LFW database implementation of bob.bio.base.database.Database interface.
- It is an extension of an SQL-based database interface, which directly talks to LFW database, for
- verification experiments (good to use in bob.bio.base framework).
"""
def __init__(
- self,
- original_directory=None,
- original_extension='.jpg',
- annotation_type=None,
- **kwargs
+ self,
+ protocol,
+ annotation_type="eyes-center",
+ image_relative_path="all_images",
+ fixed_positions=None,
+ original_directory=rc.get("bob.bio.face.lfw.directory"),
+ extension=".jpg",
+ annotation_directory=rc.get("bob.bio.face.lfw.annotation_directory"),
+ annotation_issuer="funneled",
):
- from bob.db.lfw.query import Database as LowLevelDatabase
- self._db = LowLevelDatabase(original_directory, original_extension, annotation_type)
-
- # call base class constructors to open a session to the database
- super(LFWBioDatabase, self).__init__(
- name='lfw',
- original_directory=original_directory,
- original_extension=original_extension,
+
+ if original_directory is None or not os.path.exists(original_directory):
+ raise ValueError(
+ "Invalid or non existant `original_directory`: f{original_directory}."
+ "Please, do `bob config set bob.bio.face.lfw.directory PATH` to set the LFW data directory."
+ )
+
+ if annotation_directory is None or not os.path.exists(annotation_directory):
+ logger.warning(
+ "Invalid or non existant `annotation_directory`: f{annotation_directory}."
+ "As a result, `SampleSet` will not contain annotations"
+ "Please, do `bob config set bob.bio.face.lfw.annotation_directory PATH` to set the LFW annotation directory."
+ )
+
+ if annotation_issuer != "funneled" and annotation_issuer != "idiap":
+ raise ValueError(
+ f"Invalid annotation issuer: {annotation_issuer}. Possible values are `idiap` or `funneled`"
+ )
+ self.annotation_issuer = annotation_issuer
+ # Hard-coding the extension of the annotations
+ # I don't think we need this exposed
+ # Please, open an issue if otherwise
+ self.annotation_extension = (
+ ".pos" if annotation_issuer == "idiap" else ".jpg.pts"
+ )
+
+ self._check_protocol(protocol)
+
+ self.references_dict = {}
+ self.probes_dict = {}
+ self.pairs = {}
+ self.probe_reference_keys = {} # Inverted pairs
+
+ self.annotations = None
+ self.original_directory = original_directory
+ self.annotation_directory = annotation_directory
+ self.extension = extension
+ self.image_relative_path = image_relative_path
+
+ # Some path manipulation lambdas
+ self.subject_id_from_filename = lambda x: "_".join(x.split("_")[0:-1])
+
+ self.make_path_from_filename = lambda x: os.path.join(
+ self.subject_id_from_filename(x), x
+ )
+
+ self.load_pairs()
+
+ super().__init__(
+ name="lfw",
+ protocol=protocol,
+ allow_scoring_with_all_biometric_references=False,
annotation_type=annotation_type,
- **kwargs)
-
- def groups(self):
- return self._db.groups()
+ fixed_positions=None,
+ memory_demanding=False,
+ )
+
+ def _extract_funneled(self, annotation_path):
+ """Interprets the annotation string as if it came from the funneled images.
+ Inspired by: https://gitlab.idiap.ch/bob/bob.db.lfw/-/blob/5ac22c5b77aae971de6b73cbe23f26d6a5632072/bob/db/lfw/models.py#L69
+ """
+ with open(annotation_path) as f:
+ splits = np.array(f.readlines()[0].split(" "), "float")
+
+ assert len(splits) == 18
+ locations = [
+ "reyeo",
+ "reyei",
+ "leyei",
+ "leyeo",
+ "noser",
+ "noset",
+ "nosel",
+ "mouthr",
+ "mouthl",
+ ]
+ annotations = dict(
+ [
+ (locations[i], (float(splits[2 * i + 1]), float(splits[2 * i])))
+ for i in range(9)
+ ]
+ )
+ # add eye center annotations as the center between the eye corners
+ annotations["leye"] = (
+ (annotations["leyei"][0] + annotations["leyeo"][0]) / 2.0,
+ (annotations["leyei"][1] + annotations["leyeo"][1]) / 2.0,
+ )
+ annotations["reye"] = (
+ (annotations["reyei"][0] + annotations["reyeo"][0]) / 2.0,
+ (annotations["reyei"][1] + annotations["reyeo"][1]) / 2.0,
+ )
- @property
- def original_directory(self):
- return self._db.original_directory
+ return annotations
- @original_directory.setter
- def original_directory(self, value):
- self._db.original_directory = value
+ def _extract_idiap(self, annotation_file):
+ """Interprets the annotation string as if it came from the Idiap annotations.
+ Inspired by: https://gitlab.idiap.ch/bob/bob.db.lfw/-/blob/5ac22c5b77aae971de6b73cbe23f26d6a5632072/bob/db/lfw/models.py#L81"""
- def model_ids_with_protocol(self, groups=None, protocol=None, **kwargs):
- return self._db.model_ids(groups=groups, protocol=protocol)
+ annotations = {}
+ splits = {}
+ with open(annotation_path) as f:
+ for line in f.readlines():
+ line = line.split(" ")
+ if len(line) == 3:
+ # splits.append([float(line[2]), float(line[1])])
+ splits[int(line[0])] = (float(line[1]), float(line[2]))
- def objects(self, groups=None, protocol=None, purposes=None, model_ids=None, **kwargs):
- retval = self._db.objects(groups=groups, protocol=protocol, purposes=purposes, model_ids=model_ids, **kwargs)
- return [LFWBioFile(f) for f in retval]
+ if 3 in splits:
+ annotations["reye"] = splits[3]
+
+ if 8 in splits:
+ annotations["leye"] = splits[8]
+
+ return annotations
+
+ def load_pairs(self):
+ pairs_path = os.path.join(self.original_directory, "view2", "pairs.txt")
+ self.pairs = {}
+
+ make_filename = lambda name, index: f"{name}_{index.zfill(4)}"
+
+ with open(pairs_path) as f:
+ for i, line in enumerate(f.readlines()):
+ # Skip the first line
+ if i == 0:
+ continue
+
+ line = line.split("\t")
+
+ # Three lines, genuine pairs otherwise impostor
+ if len(line) == 3:
+ # self.subject_id_from_filename()
+ key_filename = make_filename(line[0], line[1].rstrip("\n"))
+ value_filename = make_filename(line[0], line[2].rstrip("\n"))
+
+ else:
+ key_filename = make_filename(line[0], line[1].rstrip("\n"))
+ value_filename = make_filename(line[2], line[3].rstrip("\n"))
+
+ key = self.make_path_from_filename(key_filename)
+ value = self.make_path_from_filename(value_filename)
+
+ if key not in self.pairs:
+ self.pairs[key] = []
+ self.pairs[key].append(value)
+
+ self._create_probe_reference_dict()
+
+ @staticmethod
+ def protocols():
+ return ["view2"]
+
+ def background_model_samples(self):
+ return []
+
+ def _create_probe_reference_dict(self):
+ """
+ Returns a dictionary whose each key (probe key) holds the list of biometric references
+ where that probe should be compared with.
+ """
+
+ self.probe_reference_keys = {}
+ for key in self.pairs:
+ for value in self.pairs[key]:
+
+ if value not in self.probe_reference_keys:
+ self.probe_reference_keys[value] = []
+
+ self.probe_reference_keys[value].append(key)
+
+ def probes(self, group="dev"):
+ if self.protocol not in self.probes_dict:
+ self.probes_dict[self.protocol] = []
+
+ for key in self.probe_reference_keys:
+ image_path = os.path.join(
+ self.original_directory,
+ self.image_relative_path,
+ key + self.extension,
+ )
+ if self.annotation_directory is not None:
+ annotation_path = os.path.join(
+ self.annotation_directory, key + self.annotation_extension,
+ )
+ annotations = (
+ self._extract_funneled(annotation_path)
+ if self.annotation_issuer == "funneled"
+ else self._extract_idiap(annotation_path)
+ )
+ else:
+ annotations = None
+
+ sset = SampleSet(
+ key=key,
+ reference_id=key,
+ subject_id=self.subject_id_from_filename(key),
+ references=copy.deepcopy(
+ self.probe_reference_keys[key]
+ ), # deep copying to avoid bizarre issues with dask
+ samples=[
+ DelayedSample(
+ key=key,
+ load=partial(bob.io.image.load, image_path),
+ annotations=annotations,
+ )
+ ],
+ )
+ self.probes_dict[self.protocol].append(sset)
+
+ return self.probes_dict[self.protocol]
+
+ def references(self, group="dev"):
+
+ if self.protocol not in self.references_dict:
+ self.references_dict[self.protocol] = []
+
+ for key in self.pairs:
+
+ image_path = os.path.join(
+ self.original_directory,
+ self.image_relative_path,
+ key + self.extension,
+ )
+ if self.annotation_directory is not None:
+ annotation_path = os.path.join(
+ self.annotation_directory, key + self.annotation_extension,
+ )
+ annotations = (
+ self._extract_funneled(annotation_path)
+ if self.annotation_issuer == "funneled"
+ else self._extract_idiap(annotation_path)
+ )
+ else:
+ annotations = None
+
+ sset = SampleSet(
+ key=key,
+ reference_id=key,
+ subject_id=self.subject_id_from_filename(key),
+ samples=[
+ DelayedSample(
+ key=key,
+ load=partial(bob.io.image.load, image_path),
+ annotations=annotations,
+ )
+ ],
+ )
+ self.references_dict[self.protocol].append(sset)
+
+ return self.references_dict[self.protocol]
+
+ def groups(self):
+ return ["dev"]
- def annotations(self, myfile):
- return self._db.annotations(myfile._f)
+ def all_samples(self, group="dev"):
+ self._check_group(group)
- def client_id_from_model_id(self, model_id, group='dev'):
- """Return the client id associated with the given model id.
- In this base class implementation, it is assumed that only one model is enrolled for each client and, thus, client id and model id are identical.
- All key word arguments are ignored.
- Please override this function in derived class implementations to change this behavior."""
+ return self.references() + self.probes()
- # since there is one model per file, we can re-use the function above.
- return self._db.get_client_id_from_file_id(model_id)
+ def _check_protocol(self, protocol):
+ assert protocol in self.protocols(), "Unvalid protocol `{}` not in {}".format(
+ protocol, self.protocols()
+ )
+ def _check_group(self, group):
+ assert group in self.groups(), "Unvalid group `{}` not in {}".format(
+ group, self.groups()
+ )
diff --git a/bob/bio/face/test/test_databases.py b/bob/bio/face/test/test_databases.py
index 13192153594201f9363f26b832fb7c4b48d132dc..6d832d25bbf191166b852f77295057a6f6e5e63f 100644
--- a/bob/bio/face/test/test_databases.py
+++ b/bob/bio/face/test/test_databases.py
@@ -77,41 +77,6 @@ def test_atnt():
)
-@db_available("lfw")
-def test_lfw():
- database = bob.bio.base.load_resource(
- "lfw-restricted", "database", preferred_package="bob.bio.face"
- )
- try:
- check_database(database, training_depends=True, models_depend=True)
- check_database(
- database,
- groups=("dev", "eval"),
- protocol="fold1",
- training_depends=True,
- models_depend=True,
- )
- check_database(
- bob.bio.base.load_resource(
- "lfw-unrestricted", "database", preferred_package="bob.bio.face"
- ),
- training_depends=True,
- models_depend=True,
- )
- except IOError as e:
- pytest.skip(
- "The database could not queried; probably the db.sql3 file is missing. Here is the error: '%s'"
- % e
- )
- try:
- _check_annotations(database, limit_files=1000)
- except IOError as e:
- pytest.skip(
- "The annotations could not be queried; probably the annotation files are missing. Here is the error: '%s'"
- % e
- )
-
-
def test_mobio():
from bob.bio.face.database import MobioDatabase
@@ -682,3 +647,19 @@ def test_arface():
references=("sunglasses", "scarf"),
)
+
+@pytest.mark.skipif(
+ rc.get("bob.bio.face.lfw.directory") is None,
+ reason="LFW original protocols not available. Please do `bob config set bob.bio.face.gbu.directory [GBU PATH]` to set the GBU data path.",
+)
+def test_lfw():
+
+ from bob.bio.face.database import LFWDatabase
+
+ database = LFWDatabase(protocol="view2")
+ references = database.references()
+ probes = database.probes()
+ assert len(references) == 4564
+ assert len(probes) == 4576
+ # We need to have 6000 comparisons
+ assert sum([len(p.references) for p in probes]) == 6000
diff --git a/doc/implemented.rst b/doc/implemented.rst
index 583564ca7646d3ae19b310fa59c6278787f6f115..49a12f402db63cab6a37c848426547814e2f84d9 100644
--- a/doc/implemented.rst
+++ b/doc/implemented.rst
@@ -18,7 +18,7 @@ Databases
bob.bio.face.database.ReplayBioDatabase
bob.bio.face.database.ReplayMobileBioDatabase
bob.bio.face.database.GBUDatabase
- bob.bio.face.database.LFWBioDatabase
+ bob.bio.face.database.LFWDatabase
bob.bio.face.database.MultipieDatabase
bob.bio.face.database.FargoBioDatabase
bob.bio.face.database.MEDSDatabase
diff --git a/setup.py b/setup.py
index 6a1204a6891eea0160d42fba884c6148bb71d56f..0f15ee8b2d08fc8a26d533f5bcebf2b93ae6b072 100644
--- a/setup.py
+++ b/setup.py
@@ -119,8 +119,7 @@ setup(
"ijbc-test2 = bob.bio.face.config.database.ijbc_test2:database",
"ijbc-test4-g1 = bob.bio.face.config.database.ijbc_test4_g1:database",
"ijbc-test4-g2 = bob.bio.face.config.database.ijbc_test4_g2:database",
- "lfw-restricted = bob.bio.face.config.database.lfw_restricted:database",
- "lfw-unrestricted = bob.bio.face.config.database.lfw_unrestricted:database",
+ "lfw-view2 = bob.bio.face.config.database.lfw_view2:database",
"meds = bob.bio.face.config.database.meds:database",
"mobio-all = bob.bio.face.config.database.mobio_all:database",
"mobio-male = bob.bio.face.config.database.mobio_male:database",
@@ -203,8 +202,7 @@ setup(
"ijbc-test2 = bob.bio.face.config.database.ijbc_test2",
"ijbc-test4-g1 = bob.bio.face.config.database.ijbc_test4_g1",
"ijbc-test4-g2 = bob.bio.face.config.database.ijbc_test4_g2",
- "lfw-restricted = bob.bio.face.config.database.lfw_restricted",
- "lfw-unrestricted = bob.bio.face.config.database.lfw_unrestricted",
+ "lfw-view2 = bob.bio.face.config.database.lfw_view2",
"meds = bob.bio.face.config.database.meds",
"mobio-all = bob.bio.face.config.database.mobio_all",
"mobio-male = bob.bio.face.config.database.mobio_male",