def process(metadata: AudioClipMetadata):
new_path = audio_folder / metadata.path.with_suffix('.wav').name
if not new_path.exists():
audio_data = silent_load(str(metadata.path), self.dataset.sr,
self.dataset.mono)
soundfile.write(str(new_path), audio_data, self.dataset.sr)
metadata.path = new_path
def process(metadata: AudioClipMetadata):
new_path = (audio_folder / metadata.audio_id).with_suffix(".wav")
# TODO:: process function should also take in sample (AudioClipExample)
# and use sample.audio_data when metadata.path does not exist
if not new_path.exists():
audio_data = silent_load(str(metadata.path), self.dataset.sr, self.dataset.mono)
soundfile.write(str(new_path), audio_data, self.dataset.sr)
metadata.path = new_path
def __getitem__(self, idx) -> ClassificationClipExample:
metadata = self.metadata_list[idx]
audio_data = silent_load(str(metadata.path), self.sr, self.mono)
return ClassificationClipExample(
metadata=metadata,
audio_data=torch.from_numpy(audio_data),
sample_rate=self.sr,
label=self.label_map[metadata.transcription])
def __getitem__(self, idx) -> WakeWordClipExample:
metadata = self.metadata_list[idx]
audio_data = silent_load(str(metadata.path), self.sr, self.mono)
return WakeWordClipExample(
metadata=metadata,
audio_data=torch.from_numpy(audio_data),
sample_rate=self.sr,
label_data=self.frame_labeler.compute_frame_labels(metadata))
def __getitem__(self, idx) -> AudioClipExample:
metadata = self.metadata_list[idx]
audio_data = silent_load(str(metadata.path), self.sr, self.mono)
return AudioClipExample(metadata=metadata,
audio_data=torch.from_numpy(audio_data),
sample_rate=self.sr)