def random_mixed(supervision_manifest: Pathlike, feature_manifest: Pathlike,
output_cut_manifest: Pathlike, snr_range: Tuple[float, float],
offset_range: Tuple[float, float]):
"""
Create a CutSet stored in OUTPUT_CUT_MANIFEST that contains supervision regions from SUPERVISION_MANIFEST
and features supplied by FEATURE_MANIFEST. It first creates a trivial CutSet, splits it into two equal, randomized
parts and mixes their features.
The parameters of the mix are controlled via SNR_RANGE and OFFSET_RANGE.
"""
supervision_set = SupervisionSet.from_json(supervision_manifest)
feature_set = FeatureSet.from_json(feature_manifest)
source_cut_set = CutSet.from_manifests(supervisions=supervision_set,
features=feature_set)
left_cuts, right_cuts = source_cut_set.split(num_splits=2, shuffle=True)
snrs = np.random.uniform(*snr_range, size=len(left_cuts)).tolist()
relative_offsets = np.random.uniform(*offset_range,
size=len(left_cuts)).tolist()
mixed_cut_set = CutSet.from_cuts(
left_cut.mix(right_cut,
offset_other_by=left_cut.duration * relative_offset,
snr=snr)
for left_cut, right_cut, snr, relative_offset in zip(
left_cuts, right_cuts, snrs, relative_offsets))
mixed_cut_set.to_json(output_cut_manifest)
def extract(audio_manifest: Pathlike, output_dir: Pathlike,
segmentation_manifest: Optional[Pathlike], augmentation: str,
feature_manifest: Optional[Pathlike], compressed: bool,
lilcom_tick_power: int, root_dir: Optional[Pathlike],
num_jobs: int):
"""
Extract features for recordings in a given AUDIO_MANIFEST. The features are stored in OUTPUT_DIR,
with one file per recording (or segment).
"""
audio_set = RecordingSet.from_json(audio_manifest)
feature_extractor = (FeatureExtractor.from_yaml(feature_manifest)
if feature_manifest is not None else Fbank())
# TODO: to be used (actually, only the segmentation info will be used, and all supervision info will be ignored)
supervision_set = (SupervisionSet.from_json(segmentation_manifest)
if segmentation_manifest is not None else None)
output_dir = Path(output_dir)
output_dir.mkdir(exist_ok=True, parents=True)
augmenter = None
if augmentation is not None:
sampling_rate = next(iter(audio_set)).sampling_rate
assert all(rec.sampling_rate == sampling_rate for rec in audio_set), \
"Wav augmentation effect chains expect all the recordings to have the same sampling rate at this time."
augmenter = WavAugmenter.create_predefined(name=augmentation,
sampling_rate=sampling_rate)
feature_set_builder = FeatureSetBuilder(
feature_extractor=feature_extractor,
output_dir=output_dir,
root_dir=root_dir,
augmenter=augmenter)
feature_set_builder.process_and_store_recordings(
recordings=audio_set,
segmentation=None, # TODO: implement and use
compressed=compressed,
lilcom_tick_power=lilcom_tick_power,
num_jobs=num_jobs)
def supervision_set():
return SupervisionSet.from_json('test/fixtures/supervision.yml')
def external_supervision_set() -> SupervisionSet:
return SupervisionSet.from_json(
"test/fixtures/supervision.json").with_alignment_from_ctm(
"test/fixtures/supervision.ctm")
def external_supervision_set() -> SupervisionSet:
return SupervisionSet.from_json('test/fixtures/supervision.json')
