Skip to content
GitLab
Commit b20c6a25 authored by Gaël Le Lan's avatar Gaël Le Lan
Browse files

add_noise

parent 71e8b37c
Hide whitespace changes
Inline Side-by-side
nnet/augmentation.py
View file @ b20c6a25
......@@ -105,6 +105,8 @@ class AddNoise(object):
noises = []
left = original_duration
noise = numpy.zeros_like(data)
while left > 0:
# select noise file at random
file = random.choice(self.noises)
......@@ -117,23 +119,23 @@ class AddNoise(object):
duration = noise_signal.shape[0]
# if noise file is longer than what is needed, crop it
if duration > left:
noise = crop(noise_signal, left)
if duration >= left:
noise[-left:] = normalize(crop(noise_signal, left))
left = 0
# otherwise, take the whole file
else:
noise = noise_signal
noise[-left:-left+duration] = normalize(noise_signal)
left -= duration
# Todo Downsample if needed
# if sample_rate > fs:
#
noise = normalize(noise)
noises.append(noise.squeeze())
#noise = normalize(noise)
#noises.append(noise.squeeze())
# concatenate
noise = numpy.hstack(noises)
#noise = numpy.hstack(noises)
# select SNR at random
snr = (self.snr_max - self.snr_min) * numpy.random.random_sample() + self.snr_min
......
nnet/res_net.py
View file @ b20c6a25
......@@ -607,21 +607,21 @@ class PreFastResNet34(torch.nn.Module):
self.in_planes = 16
self.speaker_number = speaker_number
#torchaudio.transforms.MelSpectrogram(sample_rate=16000,
# n_fft=2048,
# f_min=133.333,
# f_max=6855.4976,
# win_length=400,
# hop_length=160,
# window_fn=torch.hann_window,
# power=2,
# n_mels=80)
self.MelSpec = torchaudio.transforms.MelSpectrogram(sample_rate=melkwargs['sample_rate'],
n_fft=melkwargs['n_fft'],
f_min=melkwargs['f_min'],
f_max=melkwargs['f_max'],
win_length=melkwargs['win_length'],
hop_length=melkwargs['hop_length'],
window_fn=melkwargs['window_fn'],
n_mels=melkwargs['n_mels'])
self.PreEmphasis = PreEmphasis()
self.MFCC = torchaudio.transforms.MFCC(sample_rate=sample_rate,
n_mfcc=1 + n_mfcc,
log_mels=True,
melkwargs = melspec_dict)
#self.MFCC = torchaudio.transforms.MFCC(sample_rate=sample_rate,
# n_mfcc=1 + n_mfcc,
# log_mels=True,
# melkwargs = melkwargs)
self.CMVN = torch.nn.InstanceNorm1d(n_mfcc)
......@@ -648,9 +648,10 @@ class PreFastResNet34(torch.nn.Module):
with torch.no_grad():
with torch.cuda.amp.autocast(enabled=False):
out = self.PreEmphasis(x)
out = self.MFCC(x)[:, 1:, :]
out = self.CMVN(x)
out = torch.nn.functional.relu(self.bn1(self.conv1(x)))
out = self.MelSpec(out)+1e-6
out = torch.log(out)
out = self.CMVN(out).unsqueeze(1)
out = torch.nn.functional.relu(self.bn1(self.conv1(out)))
out = self.layer1(out)
out = self.layer2(out)
out = self.layer3(out)
......
nnet/xsets.py
View file @ b20c6a25
......@@ -274,7 +274,6 @@ class MFCC(object):
"""
sig = sample[0][:, numpy.newaxis] # ajout
framed = framing(sample[0], self.window_length, win_shift=self.window_length - self.overlap).copy()
framed = framing(sample[0], self.window_length, win_shift=self.window_length - self.overlap).copy()
# Pre-emphasis filtering is applied after framing to be consistent with stream processing
framed = pre_emphasis(framed, self.prefac)
# Windowing has been changed to hanning which is supposed to have less noisy sidelobes
......@@ -453,9 +452,13 @@ class SpkSet(Dataset):
current_speaker = self._spk_index[int(math.fmod(index, len(self._spk_index)))]
segment_index = numpy.random.choice(self._spk_dict[current_speaker]['num_segs'], p=self._spk_dict[current_speaker]['p'])
self._spk_dict[current_speaker]['p'][segment_index] /= 2
self._spk_dict[current_speaker]['p'][segment_index] = 0 #/= 2
current_segment = self._spk_dict[current_speaker]['segments'][segment_index]
self._spk_dict[current_speaker]['p'] = self._spk_dict[current_speaker]['p']/numpy.sum(self._spk_dict[current_speaker]['p'])
if numpy.sum(self._spk_dict[current_speaker]['p']) > 0:
self._spk_dict[current_speaker]['p'] = self._spk_dict[current_speaker]['p']/numpy.sum(self._spk_dict[current_speaker]['p'])
else:
self._spk_dict[current_speaker]['p'] += 1/self._spk_dict[current_speaker]['num_segs']
nfo = soundfile.info(f"{self.data_path}/{current_segment['file_id']}{self.data_file_extension}")
if self._windowed:
......@@ -465,16 +468,16 @@ class SpkSet(Dataset):
start_frame = int(current_segment['start'] * self.sample_rate)
stop_frame = int(current_segment['duration'] * self.sample_rate)
sig, sample_rate2 = torchaudio.load(f"{self.data_path}/{current_segment['file_id']}{self.data_file_extension}",
frame_offset=start_frame,
num_frames=stop_frame)
#sig, _ = soundfile.read(f"{self.data_path}/{current_segment['file_id']}{self.data_file_extension}",
# start=start_frame,
# stop=stop_frame,
# dtype=wav_type
# )
#sig = sig.astype(numpy.float32)
#sig += 0.0001 * numpy.random.randn(sig.shape[0])
#sig, sample_rate2 = torchaudio.load(f"{self.data_path}/{current_segment['file_id']}{self.data_file_extension}",
# frame_offset=start_frame,
# num_frames=stop_frame)
sig, _ = soundfile.read(f"{self.data_path}/{current_segment['file_id']}{self.data_file_extension}",
start=start_frame,
stop=stop_frame,
dtype=wav_type
)
sig = sig.astype(numpy.float32)
sig += 0.0001 * numpy.random.randn(sig.shape[0])
speaker_idx = self._spk_dict[current_speaker]["speaker_idx"]
......
nnet/xvector.py
View file @ b20c6a25
......@@ -424,25 +424,27 @@ class Xtractor(torch.nn.Module):
self.input_nbdim = 2
self.preprocessor = None
melkargs = dict()
melkargs['n_fft'] = 2048
melkargs['f_min'] = 133.333
melkargs['f_max'] = 6855.4976
melkargs['win_length'] = 400
melkargs['hop_length'] = 160
melkargs['window_fn'] = torch.hann_window
melkargs['power'] = 2
melkargs['n_mels'] = 81
self.sequence_network = PreFastResNet34(n_mfcc=80, melkargs=melkargs)
melkwargs = dict()
melkwargs['sample_rate'] = 16000
melkwargs['n_fft'] = 1024 #2048
melkwargs['f_min'] = 90 #133.333
melkwargs['f_max'] = 7600 #6855.4976
melkwargs['win_length'] = 1024 #400
melkwargs['hop_length'] = 256 #160
melkwargs['window_fn'] = torch.hann_window
#melkwargs['power'] = 2
melkwargs['n_mels'] = 80
self.sequence_network = PreFastResNet34(melkwargs=melkwargs)
self.embedding_size = 256
self.before_speaker_embedding = torch.nn.Linear(in_features = 2560,
out_features = 512)
out_features = self.embedding_size)
self.stat_pooling = MeanStdPooling()
self.stat_pooling_weight_decay = 0
self.embedding_size = 512
self.loss = "aam"
self.after_speaker_embedding = ArcMarginProduct(self.embedding_size,
......@@ -818,7 +820,7 @@ def xtrain(speaker_number,
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
# Start from scratch
if model_name is None and model_yaml in ["xvector", "rawnet2", "resnet34"]:
if model_name is None and model_yaml in ["xvector", "rawnet2", "resnet34", "fastresnet34"]:
# Initialize a first model
if model_yaml == "xvector":
model = Xtractor(speaker_number, "xvector", loss=loss)
......@@ -826,6 +828,8 @@ def xtrain(speaker_number,
model = Xtractor(speaker_number, "rawnet2")
elif model_yaml == "resnet34":
model = Xtractor(speaker_number, "resnet34")
elif model_yaml == "fastresnet34":
model = Xtractor(speaker_number, "fastresnet34")
model_archi = model_yaml
else:
with open(model_yaml, 'r') as fh:
......@@ -1009,18 +1013,18 @@ def xtrain(speaker_number,
best_eer = 100
curr_patience = patience
logging.critical("Compute EER before starting")
val_acc, val_loss, val_eer = cross_validation(model,
validation_loader,
device,
[validation_set.__len__(),
embedding_size],
mixed_precision)
test_eer = test_metrics(model, device, speaker_number, num_thread, mixed_precision)
#logging.critical("Compute EER before starting")
#val_acc, val_loss, val_eer = cross_validation(model,
# validation_loader,
# device,
# [validation_set.__len__(),
# embedding_size],
# mixed_precision)
logging.critical(f"***{time.strftime('%H:%M:%S', time.localtime())} Initial metrics - Cross validation accuracy = {val_acc} %, EER = {val_eer * 100} %")
logging.critical(f"***{time.strftime('%H:%M:%S', time.localtime())} Initial metrics - Test EER = {test_eer * 100} %")
#test_eer = test_metrics(model, device, speaker_number, num_thread, mixed_precision)
test_eer = 100
#logging.critical(f"***{time.strftime('%H:%M:%S', time.localtime())} Initial metrics - Cross validation accuracy = {val_acc} %, EER = {val_eer * 100} %")
#logging.critical(f"***{time.strftime('%H:%M:%S', time.localtime())} Initial metrics - Test EER = {test_eer * 100} %")
for epoch in range(1, epochs + 1):
# Process one epoch and return the current model
......@@ -1782,7 +1786,7 @@ def eer(negatives, positives):
n_index = n_index - next_n_jump
if next_p_jump == 0 and next_n_jump == 0:
break
p_score = positives[p_index]
n_score = negatives[n_index]
next_p_jump = next_p_jump//2
......
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment