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_no_missing_projections_if_projection_complete(self):
projections = {
('b', ): 'foo',
(
'a',
'b',
): 'foo',
(
'a',
'b',
'c',
): 'foo',
(
'a',
'c',
): 'foo',
('c', ): 'bar'
}
label_list = LabelList(labels=[
Label('b', 3.2, 4.5),
Label('a', 4.0, 4.9),
Label('c', 4.2, 5.1)
])
unmapped_combinations = relabeling.find_missing_projections(
label_list, projections)
assert len(unmapped_combinations) == 0
def test_all_projections_missing_if_no_projections_defined(self):
label_list = LabelList(labels=[
Label('b', 3.2, 4.5),
Label('a', 4.0, 4.9),
Label('c', 4.2, 5.1)
])
unmapped_combinations = relabeling.find_missing_projections(
label_list, {})
assert len(unmapped_combinations) == 5
assert ('b', ) in unmapped_combinations
assert (
'a',
'b',
) in unmapped_combinations
assert (
'a',
'b',
'c',
) in unmapped_combinations
assert (
'a',
'c',
) in unmapped_combinations
assert ('c', ) in unmapped_combinations
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_len(self):
ll = LabelList(labels=[
Label('some text'),
Label('more text'),
Label('text again'),
])
assert len(ll) == 3
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 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_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 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_end(self):
ll = LabelList(idx='test', labels=[
Label('c', 9.0, 12.0),
Label('a', 0.2, 4.0),
Label('b', 4.0, 8.0),
Label('c', 12.05, 14.0),
Label('b', 7.9, 9.3),
])
assert ll.end == 14.0
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_total_length(self):
ll = LabelList(labels=[
Label('abc', 3.2, 4.5),
Label('bg', 5.1, 8.9),
Label('caaf', 7.2, 10.5),
Label('yxva', 10.5, 14),
Label('cy', 13, 14)
])
assert ll.total_length == 15
def test_end_returns_inf(self):
ll = LabelList(idx='test', labels=[
Label('c', 9.0, 12.0),
Label('a', 0.2, 4.0),
Label('b', 4.0, float('inf')),
Label('c', 12.05, 14.0),
Label('b', 7.9, 9.3),
])
assert ll.end == float('inf')
def test_split_unsorted_label_list(self):
ll = LabelList(idx='test',
labels=[
Label('a', 0.0, 4.0),
Label('c', 9.0, 12.0),
Label('b', 4.0, 8.0)
])
res = ll.split([1.9, 6.2, 10.5])
assert res == [
LabelList(idx='test', labels=[
Label('a', 0.0, 1.9),
]),
LabelList(idx='test',
labels=[
Label('a', 1.9, 4.0),
Label('b', 4.0, 6.2),
]),
LabelList(idx='test',
labels=[
Label('b', 6.2, 8.0),
Label('c', 9.0, 10.5),
]),
LabelList(idx='test', labels=[
Label('c', 10.5, 12.0),
]),
]
def test_split_with_overlap(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, 6.2, 10.5], overlap=2.0)
assert res == [
LabelList(idx='test', labels=[
Label('a', 0.0, 3.9)
]),
LabelList(idx='test', labels=[
Label('a', 0, 4.0),
Label('b', 4.0, 8.0)
]),
LabelList(idx='test', labels=[
Label('b', 4.2, 8.0),
Label('c', 9.0, 12.0)]
),
LabelList(idx='test', labels=[
Label('c', 9.0, 12.0)
]),
]
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_with_label_values(self):
ll = LabelList.with_label_values([
'a',
'b',
'c',
])
assert ll == LabelList(labels=[
Label('a'),
Label('b'),
Label('c'),
])
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_is(self):
ll_a = LabelList(idx='test', labels=[
Label('a', 0.0, 4.0),
Label('b', 4.0, 8.0),
Label('b', 7.9, 9.3),
Label('c', 9.0, 12.0),
Label('c', 12.05, 14.0),
])
ll_b = ll_a
assert ll_a is ll_b
def test_iter(self):
ll = LabelList(labels=[
Label('some text'),
Label('more text'),
Label('text again'),
])
labels = sorted(ll)
assert labels[0].value == 'more text'
assert labels[1].value == 'some text'
assert labels[2].value == 'text again'
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_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_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_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_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