def test_amplitude_to_DB_reversible(self, shape):
"""Round trip between amplitude and db should return the original for various shape
This implicitly also tests `DB_to_amplitude`.
"""
amplitude_mult = 20.
power_mult = 10.
amin = 1e-10
ref = 1.0
db_mult = math.log10(max(amin, ref))
torch.manual_seed(0)
spec = torch.rand(*shape, dtype=self.dtype, device=self.device) * 200
# Spectrogram amplitude -> DB -> amplitude
db = F.amplitude_to_DB(spec,
amplitude_mult,
amin,
db_mult,
top_db=None)
x2 = F.DB_to_amplitude(db, ref, 0.5)
self.assertEqual(x2, spec, atol=5e-5, rtol=1e-5)
# Spectrogram power -> DB -> power
db = F.amplitude_to_DB(spec, power_mult, amin, db_mult, top_db=None)
x2 = F.DB_to_amplitude(db, ref, 1.)
self.assertEqual(x2, spec)
def test_DB_to_amplitude(self):
# Make some noise
x = torch.rand(1000)
spectrogram = torchaudio.transforms.Spectrogram()
spec = spectrogram(x)
amin = 1e-10
ref = 1.0
db_multiplier = math.log10(max(amin, ref))
# Waveform amplitude -> DB -> amplitude
multiplier = 20.
power = 0.5
db = F.amplitude_to_DB(torch.abs(x),
multiplier,
amin,
db_multiplier,
top_db=None)
x2 = F.DB_to_amplitude(db, ref, power)
self.assertEqual(x2, torch.abs(x), atol=5e-5, rtol=1e-5)
# Spectrogram amplitude -> DB -> amplitude
db = F.amplitude_to_DB(spec,
multiplier,
amin,
db_multiplier,
top_db=None)
x2 = F.DB_to_amplitude(db, ref, power)
self.assertEqual(x2, spec, atol=5e-5, rtol=1e-5)
# Waveform power -> DB -> power
multiplier = 10.
power = 1.
db = F.amplitude_to_DB(x, multiplier, amin, db_multiplier, top_db=None)
x2 = F.DB_to_amplitude(db, ref, power)
self.assertEqual(x2, torch.abs(x), atol=5e-5, rtol=1e-5)
# Spectrogram power -> DB -> power
db = F.amplitude_to_DB(spec,
multiplier,
amin,
db_multiplier,
top_db=None)
x2 = F.DB_to_amplitude(db, ref, power)
self.assertEqual(x2, spec, atol=5e-5, rtol=1e-5)
def db_to_amplitude(spectrogram: Tensor, hp: HParams) -> Tensor:
r"""Wrapper around torchaudio.functional.DB_to_amplitude().
Args:
spectrogram (Tensor): spectrogram in the decibel scale.
Shape: [B, FREQ, FRAMES] or [B, N_MELS, FRAMES] if it is a melspectrogram.
hp (HParams): parameters. Parameters needed are power.
Returns:
spectrogram (Tensor): spectrogram in power/amplitude scale.
Shape: Shape: [B, FREQ, FRAMES] or [B, N_MELS, FRAMES] if it is a melspectrogram.
"""
assert len(spectrogram.size()) == 3, \
"Dimensions of spectrogram should be 3: [B, FREQ, FRAMES] or [B, N_MELS, FRAMES], " \
"but found {}".format(len(spectrogram.size()))
# power_exp calculated according to torchaudio.functional.DB_to_amplitude docs
power_exp = 1 if hp.audio.spectrogram_type == 'power' else 0.5
return F.DB_to_amplitude(spectrogram, ref=1, power=power_exp)
def __getitem__(self, idx):
if self.num_frames != -1:
idx = idx % self.num_frames
low_index, high_index = binarySearch(self.cumulative_sum, idx+1)
file_name = self.df.iloc[high_index]['file_name']
emb_path = os.path.join(self.embeddings_dir, file_name)
spec_path = os.path.join(self.spectrograms_dir, file_name)
audio_prep_path = os.path.join(self.audio_pred_dir, file_name)
if low_index == 0 and high_index == 0:
frame_idx = idx
else:
frame_idx = idx - self.cumulative_sum[low_index]
with open(emb_path, 'rb') as f:
emb = np.load(f)
with open(spec_path, 'rb') as f:
spec = np.load(f)
with open(audio_prep_path, 'rb') as f:
audio_prep = np.load(f)
emb_tensor = torch.from_numpy(emb[frame_idx]).float()
spec_tensor = torch.from_numpy(spec[frame_idx]).float().permute(2, 0, 1)
audio_prep_tensor = torch.from_numpy(audio_prep[frame_idx]).float()
if self.emb_means is not None and self.emb_stds is not None:
emb_tensor = ( emb_tensor - torch.tensor(self.emb_means) ) / torch.tensor(self.emb_stds)
if self.return_amp is True:
spec_tensor_amp = F.DB_to_amplitude(x = spec_tensor, ref = 1, power = 0.5)
if self.spec_means is not None and self.spec_stds is not None:
spec_tensor_amp = ( spec_tensor_amp - torch.tensor(self.spec_means) ) / torch.tensor(self.spec_stds)
spec_tensor_amp = spec_tensor_amp.float()
return emb_tensor, spec_tensor_amp, audio_prep_tensor, file_name, torch.tensor(frame_idx)
else:
return emb_tensor, spec_tensor, audio_prep_tensor, file_name, torch.tensor(frame_idx)
def func(tensor):
ref = 1.
power = 1.
return F.DB_to_amplitude(tensor, ref, power)
