def test_csv_file_list_dev_only_metadata():
dataset = CSVDataset(
name="example_csv_filelist",
dataset_protocol_path=example_dir,
protocol="protocol_only_dev_metadata",
)
assert len(dataset.background_model_samples()) == 8
assert check_all_true(dataset.background_model_samples(), DelayedSample)
assert np.alltrue(
["metadata_1" in s.__dict__ for s in dataset.background_model_samples()]
)
assert np.alltrue(
["metadata_2" in s.__dict__ for s in dataset.background_model_samples()]
)
assert len(dataset.references()) == 2
assert check_all_true(dataset.references(), SampleSet)
assert np.alltrue(
["metadata_1" in s.__dict__ for s in dataset.references()]
)
assert np.alltrue(
["metadata_2" in s.__dict__ for s in dataset.references()]
)
assert len(dataset.probes()) == 10
assert check_all_true(dataset.probes(), SampleSet)
assert np.alltrue(["metadata_1" in s.__dict__ for s in dataset.probes()])
assert np.alltrue(["metadata_2" in s.__dict__ for s in dataset.probes()])
assert np.alltrue(["references" in s.__dict__ for s in dataset.probes()])
def test_csv_file_list_dev_only():
dataset = CSVDataset(
name="example_csv_filelist",
dataset_protocol_path=example_dir,
protocol="protocol_only_dev",
)
assert len(dataset.background_model_samples()) == 8
assert check_all_true(dataset.background_model_samples(), DelayedSample)
assert len(dataset.references()) == 2
assert check_all_true(dataset.references(), SampleSet)
assert len(dataset.probes()) == 10
assert check_all_true(dataset.probes(), SampleSet)
def run(filename):
dataset = CSVDataset(
name="example_csv_filelist",
dataset_protocol_path=filename,
protocol="protocol_dev_eval",
csv_to_sample_loader=make_pipeline(
CSVToSampleLoaderBiometrics(
data_loader=bob.io.base.load,
dataset_original_directory="",
extension="",
),
AnnotationsLoader(
annotation_directory=annotation_directory,
annotation_extension=".json",
annotation_type="json",
),
),
)
assert len(dataset.background_model_samples()) == 8
assert check_all_true(dataset.background_model_samples(), DelayedSample)
assert len(dataset.references()) == 2
assert check_all_true(dataset.references(), SampleSet)
assert len(dataset.probes()) == 8
assert check_all_true(dataset.references(), SampleSet)
assert len(dataset.references(group="eval")) == 6
assert check_all_true(dataset.references(group="eval"), SampleSet)
assert len(dataset.probes(group="eval")) == 13
assert check_all_true(dataset.probes(group="eval"), SampleSet)
assert len(dataset.all_samples(groups=None)) == 47
assert check_all_true(dataset.all_samples(groups=None), DelayedSample)
# Check the annotations
for s in dataset.all_samples(groups=None):
assert isinstance(s.annotations, dict)
assert len(dataset.reference_ids(group="dev")) == 2
assert len(dataset.reference_ids(group="eval")) == 6
assert len(dataset.groups()) == 3
def test_csv_file_list_atnt():
dataset = CSVDataset(
name="atnt",
dataset_protocol_path=atnt_protocol_path,
protocol="idiap_protocol",
)
assert len(dataset.background_model_samples()) == 200
assert len(dataset.references()) == 20
assert len(dataset.probes()) == 100
assert len(dataset.all_samples(groups=["train"])) == 200
assert len(dataset.all_samples(groups=["dev"])) == 200
assert len(dataset.all_samples(groups=None)) == 400
def test_lst_file_list_dev_eval_sparse():
dataset = CSVDataset(
name="example_filelist",
dataset_protocol_path=legacy_example_dir,
protocol="",
csv_to_sample_loader=LSTToSampleLoader(
data_loader=bob.io.base.load,
dataset_original_directory="",
extension="",
),
is_sparse=True,
)
assert len(dataset.background_model_samples()) == 8
assert check_all_true(dataset.background_model_samples(), DelayedSample)
assert len(dataset.references()) == 2
assert check_all_true(dataset.references(), SampleSet)
assert len(dataset.probes()) == 8
assert check_all_true(dataset.references(), SampleSet)
assert len(dataset.references(group="eval")) == 2
assert check_all_true(dataset.references(group="eval"), SampleSet)
assert len(dataset.probes(group="eval")) == 8
assert check_all_true(dataset.probes(group="eval"), SampleSet)
assert len(dataset.all_samples(groups=None)) == 44
assert check_all_true(dataset.all_samples(groups=None), DelayedSample)
assert len(dataset.reference_ids(group="dev")) == 2
assert len(dataset.reference_ids(group="eval")) == 2
assert len(dataset.groups()) == 3
def test_lst_file_list_dev_sparse_filelist2():
dataset = CSVDataset(
name="example_filelist2",
dataset_protocol_path=legacy2_example_dir,
protocol="",
csv_to_sample_loader=LSTToSampleLoader(
data_loader=bob.io.base.load,
dataset_original_directory="",
extension="",
),
is_sparse=True,
)
assert len(dataset.references()) == 3
assert check_all_true(dataset.references(), SampleSet)
assert len(dataset.probes()) == 9
assert check_all_true(dataset.references(), SampleSet)
def test_csv_file_list_dev_eval_sparse():
annotation_directory = os.path.realpath(
bob.io.base.test_utils.datafile(
".", __name__, "data/example_csv_filelist/annotations"
)
)
dataset = CSVDataset(
name="example_csv_filelist",
dataset_protocol_path=example_dir,
protocol="protocol_dev_eval_sparse",
csv_to_sample_loader=make_pipeline(
CSVToSampleLoaderBiometrics(
data_loader=bob.io.base.load,
dataset_original_directory="",
extension="",
),
AnnotationsLoader(
annotation_directory=annotation_directory,
annotation_extension=".json",
annotation_type="json",
),
),
is_sparse=True,
)
assert len(dataset.background_model_samples()) == 8
assert check_all_true(dataset.background_model_samples(), DelayedSample)
assert len(dataset.references()) == 2
assert check_all_true(dataset.references(), SampleSet)
probes = dataset.probes()
assert len(probes) == 8
# here, 1 comparisons comparison per probe
for p in probes:
assert len(p.references) == 1
assert check_all_true(dataset.references(), SampleSet)
assert len(dataset.references(group="eval")) == 6
assert check_all_true(dataset.references(group="eval"), SampleSet)
probes = dataset.probes(group="eval")
assert len(probes) == 13
assert check_all_true(probes, SampleSet)
# Here, 1 comparison per probe, EXPECT THE FIRST ONE
for i, p in enumerate(probes):
if i == 0:
assert len(p.references) == 2
else:
assert len(p.references) == 1
assert len(dataset.all_samples(groups=None)) == 48
assert check_all_true(dataset.all_samples(groups=None), DelayedSample)
# Check the annotations
for s in dataset.all_samples(groups=None):
assert isinstance(s.annotations, dict)
assert len(dataset.reference_ids(group="dev")) == 2
assert len(dataset.reference_ids(group="eval")) == 6
assert len(dataset.groups()) == 3