def test_ranges_zero_to_end(self):
labels = [
Label('a', 0, float('inf')),
Label('b', 5.1, 8.9)
]
ll = LabelList(labels=labels)
ranges = ll.ranges()
r = next(ranges)
assert 0 == r[0]
assert 5.1 == r[1]
assert labels[0] in r[2]
r = next(ranges)
assert 5.1 == r[0]
assert 8.9 == r[1]
assert labels[0] in r[2]
assert labels[1] in r[2]
r = next(ranges)
assert 8.9 == r[0]
assert float('inf') == r[1]
assert labels[0] in r[2]
with pytest.raises(StopIteration):
next(ranges)
def test_split_without_cutting_points_raises_error(self):
ll = LabelList(idx='test',
labels=[Label('a', 0.0, 9.0),
Label('c', 9.0, 12.0)])
with pytest.raises(ValueError):
ll.split([])
def test_tokenized(self):
ll = LabelList(idx='test', labels=[
Label('a u t', 0.0, 4.0),
Label('b x', 4.0, 8.0),
Label('c', 9.0, 12.0)
])
assert ll.tokenized() == ['a', 'u', 't', 'b', 'x', 'c']
def test_join_with_custom_delimiter(self):
ll = LabelList(idx='test', labels=[
Label('a', 0.0, 4.0),
Label('b', 4.0, 8.0),
Label('c', 9.0, 12.0)
])
assert ll.join(delimiter=' - ') == 'a - b - c'
def test_join(self):
ll = LabelList(idx='test', labels=[
Label('a', 0.0, 4.0),
Label('b', 4.0, 8.0),
Label('c', 9.0, 12.0)
])
assert ll.join() == 'a b c'
def test_relabel_maps_a_onto_b(self):
label_list = LabelList(labels=[Label('a', 3.2, 4.5)])
actual = relabeling.relabel(label_list, {('a', ): 'b'})
expected = LabelList(labels=[Label('b', 3.2, 4.5)])
assert actual == expected
def test_tokenized_raises_error_if_overlap_is_higher_than_threshold(self):
ll = LabelList(idx='test', labels=[
Label('a u t', 0.0, 4.0),
Label('b x', 3.85, 8.0),
Label('c', 9.0, 12.0)
])
with pytest.raises(ValueError):
ll.tokenized()
def test_join_raises_error_if_overlap_is_higher_than_threshold(self):
ll = LabelList(idx='test', labels=[
Label('a', 0.0, 4.0),
Label('b', 3.8, 8.0),
Label('c', 9.0, 12.0)
])
with pytest.raises(ValueError):
ll.join(overlap_threshold=0.1)
def test_tokenized_raises_error_if_overlap_is_higher_than_threshold_given_an_endless_label(self):
ll = LabelList(idx='test', labels=[
Label('a u t', 0.0, 4.0),
Label('b x', 4.5, float('inf')),
Label('c', 9.0, 12.0)
])
with pytest.raises(ValueError):
ll.tokenized()
def create_sample_dataset(temp_dir):
ds = audiomate.Corpus(str(temp_dir))
file_1_path = resources.sample_wav_file('wav_1.wav')
file_2_path = resources.sample_wav_file('wav_2.wav')
file_3_path = resources.get_resource_path(
['audio_formats', 'flac_1_16k_16b.flac'])
file_1 = ds.new_file(file_1_path, track_idx='wav_1')
file_2 = ds.new_file(file_2_path, track_idx='wav_2')
file_3 = ds.new_file(file_3_path, track_idx='wav_3')
issuer_1 = Speaker('spk-1', gender=Gender.MALE)
issuer_2 = Speaker('spk-2', gender=Gender.FEMALE)
issuer_3 = Issuer('spk-3')
ds.import_issuers([issuer_1, issuer_2, issuer_3])
# 2.5951875
utt_1 = ds.new_utterance('utt-1', file_1.idx, issuer_idx=issuer_1.idx)
utt_2 = ds.new_utterance('utt-2',
file_2.idx,
issuer_idx=issuer_2.idx,
start=0,
end=1.5)
utt_3 = ds.new_utterance('utt-3',
file_2.idx,
issuer_idx=issuer_2.idx,
start=1.5,
end=2.5)
# 5.0416875
utt_4 = ds.new_utterance('utt-4', file_3.idx, issuer_idx=issuer_3.idx)
utt_1.set_label_list(
LabelList(audiomate.corpus.LL_WORD_TRANSCRIPT,
labels=[Label('who am i')]))
utt_2.set_label_list(
LabelList(audiomate.corpus.LL_WORD_TRANSCRIPT,
labels=[Label('who are you')]))
utt_3.set_label_list(
LabelList(audiomate.corpus.LL_WORD_TRANSCRIPT,
labels=[Label('who is he')]))
utt_4.set_label_list(
LabelList(audiomate.corpus.LL_WORD_TRANSCRIPT,
labels=[Label('who are they')]))
train_filter = subview.MatchingUtteranceIdxFilter(
utterance_idxs={'utt-1', 'utt-2', 'utt-3'})
sv_train = subview.Subview(ds, filter_criteria=[train_filter])
dev_filter = subview.MatchingUtteranceIdxFilter(utterance_idxs={'utt-4'})
sv_dev = subview.Subview(ds, filter_criteria=[dev_filter])
ds.import_subview('train', sv_train)
ds.import_subview('dev', sv_dev)
return ds
def test_addl(self):
ll = LabelList()
assert len(ll) == 0
ll.addl('a', 12.3, 19.3)
assert ll == LabelList(labels=[
Label('a', 12.3, 19.3),
])
def test_split_with_cutting_point_after_last_label(self):
ll = LabelList(idx='test',
labels=[Label('a', 0.0, 4.0),
Label('c', 4.0, 8.9)])
res = ll.split([10.5])
assert len(res) == 2
assert len(res[0]) == 2
assert len(res[1]) == 0
def test_label_values(self):
ll = LabelList(labels=[
Label('a', 3.2, 4.5),
Label('b', 5.1, 8.9),
Label('c', 7.2, 10.5),
Label('a', 10.5, 14),
Label('c', 13, 14)
])
assert ll.label_values() == ['a', 'b', 'c']
def test_split_cutting_point_on_boundary_doesnot_split_label(self):
ll = LabelList(idx='test',
labels=[Label('a', 0.0, 9.0),
Label('c', 9.0, 12.0)])
res = ll.split([9.0])
assert len(res) == 2
assert len(res[0]) == 1
assert len(res[1]) == 1
def test_add(self):
ll = LabelList()
assert len(ll) == 0
label = Label('some text')
ll.add(label)
assert len(ll) == 1
assert label.label_list == ll
assert sorted(ll)[0] == label
def test_split_single_label_that_doesnt_start_at_zero(self):
ll = LabelList(idx='test', labels=[Label('c', 8.0, 12.0)])
res = ll.split([11.2], shift_times=True)
assert len(res) == 2
assert len(res[0]) == 1
assert sorted(res[0])[0] == Label('c', 8.0, 11.2)
assert len(res[1]) == 1
assert sorted(res[1])[0] == Label('c', 0.0, pytest.approx(0.8))
def test_with_label_values_sets_correct_idx(self):
ll = LabelList.with_label_values([
'a',
'b',
'c',
], idx='letters')
assert ll == LabelList(idx='letters', labels=[
Label('a'),
Label('b'),
Label('c'),
])
def test_with_label_values(self):
ll = LabelList.with_label_values([
'a',
'b',
'c',
])
assert ll == LabelList(labels=[
Label('a'),
Label('b'),
Label('c'),
])
def test_ranges_include_labels(self):
labels = [Label('a', 3.2, 4.5), Label('b', 5.1, 8.9)]
ll = LabelList(labels=labels)
ranges = ll.ranges(include_labels=['a'])
r = next(ranges)
assert 3.2 == r[0]
assert 4.5 == r[1]
assert labels[0] in r[2]
with pytest.raises(StopIteration):
next(ranges)
def test_merge_overlaps_with_threshold(self):
ll = LabelList(idx='test', labels=[
Label('c', 9.0, 12.0),
Label('c', 12.05, 14.0),
])
ll.merge_overlaps(threshold=0.1)
expected = LabelList(idx='test', labels=[
Label('c', 9.0, 14.0),
])
assert ll == expected
def test_split_label_within_cutting_points_is_included(self):
ll = LabelList(idx='test', labels=[
Label('a', 0.0, 4.0),
Label('b', 4.0, 8.0),
Label('c', 9.0, 12.0)
])
res = ll.split([1.9, 10.5])
assert len(res[1]) == 3
assert sorted(res[1])[1].value == 'b'
assert sorted(res[1])[1].start == 4.0
assert sorted(res[1])[1].end == 8.0
def test_all_tokens(self):
ll = LabelList(labels=[
Label('some text'),
Label('more text'),
Label('text again'),
])
assert sorted(ll.all_tokens()) == [
'again',
'more',
'some',
'text',
]
def test_label_count(self):
ll = LabelList(labels=[
Label('a', 3.2, 4.5),
Label('b', 5.1, 8.9),
Label('c', 7.2, 10.5),
Label('a', 10.5, 14),
Label('c', 13, 14)
])
res = ll.label_count()
assert 2 == res['a']
assert 1 == res['b']
assert 2 == res['c']
def test_label_total_duration(self):
ll = LabelList(labels=[
Label('a', 3.2, 4.5),
Label('b', 5.1, 8.9),
Label('c', 7.2, 10.5),
Label('a', 10.5, 14),
Label('c', 13, 14)
])
res = ll.label_total_duration()
assert res['a'] == pytest.approx(4.8)
assert res['b'] == pytest.approx(3.8)
assert res['c'] == pytest.approx(4.3)
def test_update(self):
ll = LabelList()
assert len(ll) == 0
label_a = Label('some text')
label_b = Label('more text')
label_c = Label('text again')
ll.update([label_a, label_b, label_c])
assert len(ll) == 3
assert label_a.label_list == ll
assert label_b.label_list == ll
assert label_c.label_list == ll
def test_relabel_removes_unwanted_labels(self):
projections = {
('a', ): '',
('b', ): 'b',
}
label_list = LabelList(
labels=[Label('a', 3.2, 4.4),
Label('b', 4.4, 5.1)])
actual = relabeling.relabel(label_list, projections)
expected = LabelList(labels=[Label('b', 4.4, 5.1)])
assert actual == expected
def test_split_first_label_not_splitted(self):
ll = LabelList(idx='test',
labels=[Label('a', 0.0, 9.0),
Label('c', 9.0, 12.0)])
res = ll.split([11.2], shift_times=True)
assert len(res) == 2
assert len(res[0]) == 2
assert sorted(res[0])[0] == Label('a', 0.0, 9.0)
assert sorted(res[0])[1] == Label('c', 9.0, 11.2)
assert len(res[1]) == 1
assert sorted(res[1])[0] == Label('c', 0.0, pytest.approx(0.8))
def test_labels_in_range_returns_only_fully_included(self):
ll = LabelList(labels=[
Label('a', 3.2, 4.5),
Label('b', 5.1, 8.9),
Label('c', 7.2, 10.5),
Label('a', 10.5, 14),
Label('c', 13, 15)
])
in_range = ll.labels_in_range(7.2, 14.99, fully_included=True)
assert sorted(in_range) == [
Label('c', 7.2, 10.5),
Label('a', 10.5, 14),
]
def test_apply(self):
ll = LabelList(labels=[
Label('some text'),
Label('more text'),
Label('text again'),
])
def apply_func(label):
label.value = 'app {}'.format(label.value)
ll.apply(apply_func)
labels = sorted(ll)
assert labels[0].value == 'app more text'
assert labels[1].value == 'app some text'
assert labels[2].value == 'app text again'
def test_labels_in_range(self):
ll = LabelList(labels=[
Label('a', 3.2, 4.5),
Label('b', 5.1, 8.9),
Label('c', 7.2, 10.5),
Label('a', 10.5, 14),
Label('c', 13, 14)
])
in_range = ll.labels_in_range(8.2, 12.5)
assert sorted(in_range) == [
Label('b', 5.1, 8.9),
Label('c', 7.2, 10.5),
Label('a', 10.5, 14)
]
