def setUp(self):
self.utt1 = assets.Utterance('utt-1', 'file-1')
self.utt1.set_label_list(
assets.LabelList(idx='alpha',
labels=[
assets.Label('music', 0, 5),
assets.Label('speech', 5, 12),
assets.Label('music', 13, 15)
]))
self.utt1.set_label_list(
assets.LabelList(idx='bravo',
labels=[
assets.Label('music', 0, 1),
assets.Label('speech', 2, 6)
]))
self.utt2 = assets.Utterance('utt-2', 'file-2')
self.utt2.set_label_list(
assets.LabelList(idx='alpha',
labels=[
assets.Label('music', 0, 5),
assets.Label('speech', 5, 12),
assets.Label('noise', 13, 15)
]))
self.utt2.set_label_list(
assets.LabelList(idx='bravo',
labels=[
assets.Label('music', 0, 1),
assets.Label('speech', 2, 6)
]))
def test_match(self):
filter = subview.MatchingUtteranceIdxFilter(
utterance_idxs={'a', 'b', 'd'})
self.assertTrue(filter.match(assets.Utterance('a', 'x'), None))
self.assertTrue(filter.match(assets.Utterance('b', 'x'), None))
self.assertTrue(filter.match(assets.Utterance('d', 'x'), None))
self.assertFalse(filter.match(assets.Utterance('c', 'x'), None))
self.assertFalse(filter.match(assets.Utterance('e', 'x'), None))
def test_read_samples(self):
file = assets.File('wav', resources.sample_wav_file('wav_1.wav'))
issuer = assets.Issuer('toni')
utt = assets.Utterance('test',
file,
issuer=issuer,
start=1.0,
end=2.30)
l1 = assets.Label('a', 0.15, 0.448)
l2 = assets.Label('a', 0.5, 0.73)
ll = assets.LabelList(labels=[l1, l2])
utt.set_label_list(ll)
expected, __ = librosa.core.load(file.path,
sr=None,
offset=1.15,
duration=0.298)
assert np.array_equal(l1.read_samples(), expected)
expected, __ = librosa.core.load(file.path,
sr=None,
offset=1.5,
duration=0.23)
assert np.array_equal(l2.read_samples(), expected)
def setUp(self):
file = assets.File('wav', resources.sample_wav_file('wav_1.wav'))
utt = assets.Utterance('utt', file, start=0.3, end=-1)
ll = assets.LabelList()
self.test_label = assets.Label('a', start=0.5, end=-1)
ll.append(self.test_label)
utt.set_label_list(ll)
def test_import_utterance_no_issuer(self):
importing_utterances = [
assets.Utterance('a', self.ex_file, assets.Issuer('notexist'), 0,
10)
]
with pytest.raises(ValueError):
self.corpus.import_utterances(importing_utterances)
def test_import_utterances(self):
importing_utterances = [
assets.Utterance('a', self.ex_file, self.ex_issuer, 0, 10),
assets.Utterance('b', self.ex_file, self.ex_issuer, 10, 20),
assets.Utterance('existing_utt', self.ex_file, self.ex_issuer, 20,
30)
]
mapping = self.corpus.import_utterances(importing_utterances)
assert self.corpus.num_utterances == 4
assert 'a' in self.corpus.utterances.keys()
assert 'b' in self.corpus.utterances.keys()
assert 'existing_utt_1' in self.corpus.utterances.keys()
assert len(mapping) == 3
assert mapping['a'].idx == 'a'
assert mapping['b'].idx == 'b'
assert mapping['existing_utt'].idx == 'existing_utt_1'
def new_utterance(self,
utterance_idx,
file_idx,
issuer_idx=None,
start=0,
end=-1):
"""
Add a new utterance to the corpus with the given data.
Parameters:
file_idx (str): The file id the utterance is in.
utterance_idx (str): The id to associate with the utterance. If None or already exists,
one is generated.
issuer_idx (str): The issuer id to associate with the utterance.
start (float): Start of the utterance within the file [seconds].
end (float): End of the utterance within the file [seconds]. -1 equals the end of the
file.
Returns:
Utterance: The newly added utterance.
"""
new_utt_idx = utterance_idx
# Check if there is a file with the given idx
if file_idx not in self._files.keys():
raise ValueError(
'File with id {} does not exist!'.format(file_idx))
# Check if issuer exists
issuer = None
if issuer_idx is not None:
if issuer_idx not in self._issuers.keys():
raise ValueError(
'Issuer with id {} does not exist!'.format(issuer_idx))
else:
issuer = self._issuers[issuer_idx]
# Add index to idx if already existing
if new_utt_idx in self._utterances.keys():
new_utt_idx = naming.index_name_if_in_list(new_utt_idx,
self._utterances.keys())
new_utt = assets.Utterance(new_utt_idx,
self.files[file_idx],
issuer=issuer,
start=start,
end=end)
self._utterances[new_utt_idx] = new_utt
return new_utt
def setUp(self):
self.tempdir = tempfile.mkdtemp()
self.corpus = audiomate.Corpus(self.tempdir)
self.ex_file = assets.File('existing_file', '../any/path.wav')
self.ex_issuer = assets.Issuer('existing_issuer')
self.ex_utterance = assets.Utterance('existing_utt',
self.ex_file,
issuer=self.ex_issuer)
self.corpus.files['existing_file'] = self.ex_file
self.corpus.issuers['existing_issuer'] = self.ex_issuer
self.corpus.utterances['existing_utt'] = self.ex_utterance
def test_encode_utterance_takes_lower_index_first(self):
file = assets.File('file-idx', resources.sample_wav_file('wav_1.wav'))
utt = assets.Utterance('utt-idx', file, start=0, end=5)
ll = assets.LabelList(
labels=[assets.Label('music', 0, 3),
assets.Label('speech', 3, 5)])
utt.set_label_list(ll)
enc = label_encoding.FrameOrdinalEncoder(
['speech', 'music', 'noise'],
frame_settings=units.FrameSettings(32000, 16000),
sr=16000)
actual = enc.encode(utt)
expected = np.array([1, 1, 0, 0]).astype(np.int)
assert np.array_equal(expected, actual)
def setUp(self):
self.ll_1 = assets.LabelList(idx='alpha',
labels=[
assets.Label('a', 3.2, 4.5),
assets.Label('b', 5.1, 8.9),
assets.Label('c', 7.2, 10.5),
assets.Label('d', 10.5, 14),
assets.Label('d', 15, 18)
])
self.ll_2 = assets.LabelList(idx='bravo',
labels=[
assets.Label('a', 1.0, 4.2),
assets.Label('e', 4.2, 7.9),
assets.Label('c', 7.2, 10.5),
assets.Label('f', 10.5, 14),
assets.Label('d', 15, 17.3)
])
self.ll_duplicate_idx = assets.LabelList(
idx='charlie',
labels=[assets.Label('t', 1.0, 4.2),
assets.Label('h', 4.2, 7.9)])
self.ll_3 = assets.LabelList(
idx='charlie',
labels=[assets.Label('a', 1.0, 4.2),
assets.Label('g', 4.2, 7.9)])
self.file = assets.File('wav', resources.sample_wav_file('wav_1.wav'))
self.issuer = assets.Issuer('toni')
self.utt = assets.Utterance('test',
self.file,
issuer=self.issuer,
start=1.25,
end=1.30,
label_lists=[
self.ll_1, self.ll_2,
self.ll_duplicate_idx, self.ll_3
])
def test_duration_end_of_file(self):
utt = assets.Utterance('utt', self.file, start=0.3, end=-1)
assert utt.duration == pytest.approx(2.2951875)
def test_end_abs_end_of_file(self):
utt = assets.Utterance('utt', self.file, start=0.3, end=-1)
assert utt.end_abs == pytest.approx(2.5951875)
def test_end_abs(self):
label = assets.Label('a', 2, 5)
ll = assets.LabelList(labels=[label])
assets.Utterance('utt-1', None, start=1, end=19, label_lists=[ll])
assert label.end_abs == 6
def sample_utterance():
file = assets.File('test_file', resources.sample_wav_file('wav_1.wav'))
utterance = assets.Utterance('test', file)
return utterance
