from malaya_speech.model.webrtc import WebRTC
from malaya_speech.supervised import classification
from malaya_speech.utils import describe_availability
from herpetologist import check_type
_availability = {
'vggvox-v1': {
'Size (MB)': 70.8,
'Quantized Size (MB)': 17.7,
'Accuracy': 0.80984375,
},
'vggvox-v2': {
'Size (MB)': 31.1,
'Quantized Size (MB)': 7.92,
'Accuracy': 0.8196875,
},
'speakernet': {
'Size (MB)': 20.3,
'Quantized Size (MB)': 5.18,
'Accuracy': 0.7340625,
},
'marblenet-factor1': {
'Size (MB)': 0.526,
'Quantized Size (MB)': 0.232,
'Accuracy': 0.8491875,
},
'marblenet-factor3': {
'Size (MB)': 3.21,
'Quantized Size (MB)': 0.934,
'Accuracy': 0.83855625,
},
'marblenet-factor5': {
'Size (MB)': 8.38,
'Quantized Size (MB)': 2.21,
'Accuracy': 0.843540625,
}
}
[docs]def available_model():
"""
List available VAD deep models.
"""
return describe_availability(_availability)
[docs]@check_type
def webrtc(
aggressiveness: int = 3,
sample_rate: int = 16000,
minimum_amplitude: int = 100,
):
"""
Load WebRTC VAD model.
Parameters
----------
aggressiveness: int, optional (default=3)
an integer between 0 and 3.
0 is the least aggressive about filtering out non-speech, 3 is the most aggressive.
sample_rate: int, optional (default=16000)
sample rate for samples.
minimum_amplitude: int, optional (default=100)
abs(minimum_amplitude) to assume a sample is a voice activity. Else, automatically False.
Returns
-------
result : malaya_speech.model.webrtc.WebRTC class
"""
try:
import webrtcvad
except BaseException:
raise ModuleNotFoundError(
'webrtcvad not installed. Please install it by `pip install webrtcvad` and try again.'
)
vad = webrtcvad.Vad(aggressiveness)
return WebRTC(vad, sample_rate, minimum_amplitude)
[docs]@check_type
def deep_model(model: str = 'marblenet-factor1', quantized: bool = False, **kwargs):
"""
Load VAD model.
Parameters
----------
model : str, optional (default='vggvox-v2')
Check available models at `malaya_speech.vad.available_model()`.
quantized : bool, optional (default=False)
if True, will load 8-bit quantized model.
Quantized model not necessary faster, totally depends on the machine.
Returns
-------
result : malaya_speech.supervised.classification.load function
"""
model = model.lower()
if model not in _availability:
raise ValueError(
'model not supported, please check supported models from `malaya_speech.vad.available_model()`.'
)
settings = {
'vggvox-v1': {'frame_len': 0.005, 'frame_step': 0.0005},
'vggvox-v2': {'hop_length': 24, 'concat': False, 'mode': 'eval'},
'speakernet': {'frame_ms': 20, 'stride_ms': 1.0},
'marblenet-factor1': {'feature_type': 'mfcc'},
'marblenet-factor3': {'feature_type': 'mfcc'},
'marblenet-factor5': {'feature_type': 'mfcc'},
}
return classification.load(
model=model,
module='vad',
extra=settings[model],
label=[False, True],
quantized=quantized,
**kwargs
)
