Files
pyannote-audio/notebook/example.ipynb
T
2020-10-12 16:20:50 +02:00

7.0 KiB

In [1]:
import os
os.environ['PYANNOTE_DATABASE_CONFIG'] = '/Users/bredin/Development/pyannote/pyannote-audio/tests/data/database.yml'
In [2]:
from pyannote.database import get_protocol, FileFinder
protocol = get_protocol('Debug.SpeakerDiarization.Debug', 
                        preprocessors={"audio": FileFinder()})

## Voice activity detection

In [3]:
from pyannote.audio.tasks.voice_activity_detection.task import VoiceActivityDetection
vad = VoiceActivityDetection(protocol, duration=2., batch_size=32, num_workers=4)
In [4]:
from pyannote.audio.models.debug import SimpleSegmentationModel
model = SimpleSegmentationModel(task=vad)
In [5]:
import pytorch_lightning as pl
trainer = pl.Trainer(max_epochs=10)
_ = trainer.fit(model, vad)
GPU available: False, used: False
TPU available: False, using: 0 TPU cores

  | Name       | Type       | Params
------------------------------------------
0 | mfcc       | MFCC       | 0     
1 | lstm       | LSTM       | 18 K  
2 | classifier | Linear     | 130   
3 | activation | LogSoftmax | 0     
/Users/bredin/miniconda3/envs/pyannote-audio-v2/lib/python3.7/site-packages/pytorch_lightning/utilities/distributed.py:45: UserWarning: Your `IterableDataset` has `__len__` defined. In combination with multi-processing data loading (e.g. batch size > 1), this can lead to unintended side effects since the samples will be duplicated.
  warnings.warn(*args, **kwargs)
HBox(children=(HTML(value='Training'), FloatProgress(value=1.0, bar_style='info', layout=Layout(flex='2'), max…

## Speaker tracking

In [6]:
from pyannote.audio.tasks.speaker_tracking.task import SpeakerTracking
spk = SpeakerTracking(protocol, duration=2., batch_size=32, num_workers=2)
In [7]:
model = SimpleSegmentationModel(task=spk)
In [8]:
trainer = pl.Trainer(max_epochs=10)
_ = trainer.fit(model, spk)
GPU available: False, used: False
TPU available: False, using: 0 TPU cores

  | Name       | Type    | Params
---------------------------------------
0 | mfcc       | MFCC    | 0     
1 | lstm       | LSTM    | 18 K  
2 | classifier | Linear  | 1 K   
3 | activation | Sigmoid | 0     
/Users/bredin/miniconda3/envs/pyannote-audio-v2/lib/python3.7/site-packages/pytorch_lightning/utilities/distributed.py:45: UserWarning: The dataloader, train dataloader, does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` (try 4 which is the number of cpus on this machine) in the `DataLoader` init to improve performance.
  warnings.warn(*args, **kwargs)
HBox(children=(HTML(value='Training'), FloatProgress(value=1.0, bar_style='info', layout=Layout(flex='2'), max…

Speaker embedding

In [9]:
from pyannote.audio.tasks.speaker_verification.task import SpeakerEmbeddingArcFace
emb = SpeakerEmbeddingArcFace(protocol, duration=2., batch_size=32, num_workers=4)
In [10]:
from pyannote.audio.models.debug import SimpleEmbeddingModel
model = SimpleEmbeddingModel(task=emb)
In [11]:
trainer = pl.Trainer(max_epochs=10)
_ = trainer.fit(model, emb)
GPU available: False, used: False
TPU available: False, using: 0 TPU cores

  | Name | Type | Params
------------------------------
0 | mfcc | MFCC | 0     
1 | lstm | LSTM | 18 K  
HBox(children=(HTML(value='Training'), FloatProgress(value=1.0, bar_style='info', layout=Layout(flex='2'), max…