def test_trim_to_unsupervised_segments():
cut_set = CutSet.from_cuts([
# Yields 3 unsupervised cuts - before first supervision,
# between sup2 and sup3, and after sup3.
MonoCut('cut1', start=0, duration=30, channel=0, supervisions=[
SupervisionSegment('sup1', 'rec1', start=1.5, duration=8.5),
SupervisionSegment('sup2', 'rec1', start=10, duration=5),
SupervisionSegment('sup3', 'rec1', start=20, duration=8),
]),
# Does not yield any "unsupervised" cut.
MonoCut('cut2', start=0, duration=30, channel=0, supervisions=[
SupervisionSegment('sup4', 'rec1', start=0, duration=30),
]),
])
unsupervised_cuts = cut_set.trim_to_unsupervised_segments()
assert len(unsupervised_cuts) == 3
assert unsupervised_cuts[0].start == 0
assert unsupervised_cuts[0].duration == 1.5
assert unsupervised_cuts[0].supervisions == []
assert unsupervised_cuts[1].start == 15
assert unsupervised_cuts[1].duration == 5
assert unsupervised_cuts[1].supervisions == []
assert unsupervised_cuts[2].start == 28
assert unsupervised_cuts[2].duration == 2
assert unsupervised_cuts[2].supervisions == []
def mixed_overlapping_cut_set():
"""
Input mixed cut::
|---------------mixedcut--------------------|
|--------rec1 0-30s--------|
|-------rec2 15-45s--------|
|---sup1--| |-----sup3-----|
|sup2|
"""
cut_set = CutSet.from_cuts([
MonoCut(
'cut1', start=0, duration=30, channel=0,
recording=Recording(
id='rec1', sources=[], sampling_rate=16000, num_samples=160000, duration=60.0
),
supervisions=[
SupervisionSegment('sup1', 'rec1', start=1.5, duration=10.5),
SupervisionSegment('sup2', 'rec1', start=10, duration=6),
]
).mix(
MonoCut(
'cut2', start=15, duration=30, channel=0,
recording=Recording(
id='rec2', sources=[], sampling_rate=16000, num_samples=160000, duration=60.0
),
supervisions=[
SupervisionSegment('sup3', 'rec2', start=8, duration=18),
]
),
offset_other_by=15.0
)
])
assert isinstance(cut_set[0], MixedCut)
return cut_set
def overlapping_supervisions_cut():
return MonoCut(
id="cut-1",
start=0.0,
duration=0.5,
channel=0,
features=Features(
recording_id="recording-1",
channels=0,
start=0,
duration=0.5,
type="fbank",
num_frames=50,
num_features=80,
frame_shift=0.01,
sampling_rate=16000,
storage_type="lilcom",
storage_path="test/fixtures/dummy_feats/storage/",
storage_key="e66b6386-aee5-4a5a-8369-fdde1d2b97c7.llc",
),
supervisions=[
SupervisionSegment(
id="s1", recording_id="recording-1", start=0.0, duration=0.2
),
SupervisionSegment(
id="s2", recording_id="recording-1", start=0.1, duration=0.2
),
SupervisionSegment(
id="s3", recording_id="recording-1", start=0.2, duration=0.2
),
SupervisionSegment(
id="s4", recording_id="recording-1", start=0.3, duration=0.2
),
],
)
def libri_cut():
return MonoCut(
channel=0,
duration=16.04,
features=Features(
channels=0,
duration=16.04,
num_features=40,
num_frames=1604,
frame_shift=0.01,
recording_id="recording-1",
sampling_rate=16000,
start=0.0,
storage_path="test/fixtures/libri/storage",
storage_key="30c2440c-93cb-4e83-b382-f2a59b3859b4.llc",
storage_type="lilcom_files",
type="fbank",
),
recording=Recording(
id="recording-1",
sources=[
AudioSource(
type="file",
channels=[0],
source="test/fixtures/libri/libri-1088-134315-0000.wav",
)
],
sampling_rate=16000,
num_samples=256640,
duration=1604,
),
id="849e13d8-61a2-4d09-a542-dac1aee1b544",
start=0.0,
supervisions=[],
)
def cut_with_relative_paths():
return MonoCut('cut', 0, 10, 0,
features=Features(type='fbank', num_frames=1000, num_features=40, sampling_rate=8000,
storage_type='lilcom_files', storage_path='storage_dir', storage_key='feats.llc',
start=0,
duration=10, frame_shift=0.01),
recording=Recording('rec', [AudioSource('file', [0], 'audio.wav')], 8000, 80000, 10.0)
)
def test_mix_same_recording_channels():
recording = Recording('rec', sampling_rate=8000, num_samples=30 * 8000, duration=30, sources=[
AudioSource('file', channels=[0], source='irrelevant1.wav'),
AudioSource('file', channels=[1], source='irrelevant2.wav')
])
cut_set = CutSet.from_cuts([
MonoCut('cut1', start=0, duration=30, channel=0, recording=recording),
MonoCut('cut2', start=0, duration=30, channel=1, recording=recording)
])
mixed = cut_set.mix_same_recording_channels()
assert len(mixed) == 1
cut = mixed[0]
assert isinstance(cut, MixedCut)
assert len(cut.tracks) == 2
assert cut.tracks[0].cut == cut_set[0]
assert cut.tracks[1].cut == cut_set[1]
def dummy_cut(unique_id: int,
start: float = 0.0,
duration: float = 1.0,
supervisions=None):
return MonoCut(
id=f'dummy-cut-{unique_id:04d}',
start=start,
duration=duration,
channel=0,
recording=dummy_recording(unique_id),
features=dummy_features(unique_id),
supervisions=supervisions if supervisions is not None else [],
)
def cut2(dummy_features, dummy_recording):
return MonoCut(
id="cut-2",
start=180.0,
duration=10.0,
channel=0,
features=dummy_features,
recording=dummy_recording,
supervisions=[
SupervisionSegment(id="sup-3",
recording_id="irrelevant",
start=3.0,
duration=2.5)
],
)
def cut1(dummy_features, dummy_recording):
return MonoCut(
id="cut-1",
start=0.0,
duration=10.0,
channel=0,
features=dummy_features,
recording=dummy_recording,
supervisions=[
SupervisionSegment(id="sup-1",
recording_id="irrelevant",
start=0.5,
duration=6.0),
SupervisionSegment(id="sup-2",
recording_id="irrelevant",
start=7.0,
duration=2.0),
],
)
def cut_with_relative_paths():
return MonoCut(
"cut",
0,
10,
0,
features=Features(
type="fbank",
num_frames=1000,
num_features=40,
sampling_rate=8000,
storage_type="lilcom_files",
storage_path="storage_dir",
storage_key="feats.llc",
start=0,
duration=10,
frame_shift=0.01,
),
recording=Recording("rec", [AudioSource("file", [0], "audio.wav")],
8000, 80000, 10.0),
)
def overlapping_supervisions_cut():
return MonoCut(id='cut-1',
start=0.0,
duration=0.5,
channel=0,
features=Features(
recording_id='recording-1',
channels=0,
start=0,
duration=0.5,
type='fbank',
num_frames=50,
num_features=80,
frame_shift=0.01,
sampling_rate=16000,
storage_type='lilcom',
storage_path='test/fixtures/dummy_feats/storage/',
storage_key='e66b6386-aee5-4a5a-8369-fdde1d2b97c7.llc'),
supervisions=[
SupervisionSegment(id='s1',
recording_id='recording-1',
start=0.0,
duration=0.2),
SupervisionSegment(id='s2',
recording_id='recording-1',
start=0.1,
duration=0.2),
SupervisionSegment(id='s3',
recording_id='recording-1',
start=0.2,
duration=0.2),
SupervisionSegment(id='s4',
recording_id='recording-1',
start=0.3,
duration=0.2)
])
def test_trim_to_supervisions_simple_cuts(keep_overlapping, num_jobs):
cut_set = CutSet.from_cuts([
MonoCut(
"cut1",
start=0,
duration=30,
channel=0,
supervisions=[
SupervisionSegment("sup1", "rec1", start=1.5, duration=10.5),
SupervisionSegment("sup2", "rec1", start=10, duration=5),
SupervisionSegment("sup3", "rec1", start=20, duration=8),
],
recording=dummy_recording(1, duration=30),
),
MonoCut(
"cut2",
start=0,
duration=30,
channel=0,
supervisions=[
SupervisionSegment("sup4", "rec1", start=0, duration=30),
],
recording=dummy_recording(2, duration=30),
),
])
cuts = cut_set.trim_to_supervisions(keep_overlapping=keep_overlapping,
num_jobs=num_jobs)
assert len(cuts) == 4
# Note: expected results diverge here depending on the value of keep_overlapping flag
cut = cuts[0]
assert cut.start == 1.5
assert cut.duration == 10.5
if keep_overlapping:
assert len(cut.supervisions) == 2
sup = cut.supervisions[0]
assert sup.id == "sup1"
assert sup.start == 0
assert sup.duration == 10.5
sup = cut.supervisions[1]
assert sup.id == "sup2"
assert sup.start == 8.5
assert sup.duration == 5
else:
assert len(cut.supervisions) == 1
sup = cut.supervisions[0]
assert sup.id == "sup1"
assert sup.start == 0
assert sup.duration == 10.5
# Note: expected results diverge here depending on the value of keep_overlapping flag
cut = cuts[1]
assert cut.start == 10
assert cut.duration == 5
if keep_overlapping:
assert len(cut.supervisions) == 2
sup = cut.supervisions[0]
assert sup.id == "sup1"
assert sup.start == -8.5
assert sup.duration == 10.5
sup = cut.supervisions[1]
assert sup.id == "sup2"
assert sup.start == 0
assert sup.duration == 5
else:
assert len(cut.supervisions) == 1
sup = cut.supervisions[0]
assert sup.id == "sup2"
assert sup.start == 0
assert sup.duration == 5
# Note: both test cases have same results
cut = cuts[2]
assert len(cut.supervisions) == 1
assert cut.start == 20
assert cut.duration == 8
assert cut.supervisions[0].id == "sup3"
# Note: both test cases have same results
cut = cuts[3]
assert len(cut.supervisions) == 1
assert cut.start == 0
assert cut.duration == 30
assert cut.supervisions[0].id == "sup4"
