Files
pyannote-audio/notebook/example.ipynb
T
Hervé BREDIN 06d3b6116c feat: add Hydra-based "pyannote-audio-train" CLI
* closes #476 (CLI)
* closes #514 (support for AutoLR)
* closes #485 (hyper-parameter optimization with Hydra’s Ax sweeper)
* closes #412 (log graph to Tensorboard)
2020-11-18 10:59:39 +01:00

4.3 KiB

In [ ]:
from pyannote.database import get_protocol, FileFinder
protocol = get_protocol('Debug.SpeakerDiarization.Debug', 
                        preprocessors={"audio": FileFinder()})
In [ ]:
from pyannote.audio.models.debug import SimpleSegmentationModel
from pyannote.audio.models.debug import MultiTaskSegmentationModel
from pyannote.audio.models.debug import SimpleEmbeddingModel
import pytorch_lightning as pl
In [ ]:
from pyannote.audio.utils.preview import listen
listen(next(protocol.train()))

## Voice activity detection

In [ ]:
from pyannote.audio.tasks import VoiceActivityDetection
vad = VoiceActivityDetection(protocol, duration=2., batch_size=32, num_workers=4)
model = SimpleSegmentationModel(task=vad)
trainer = pl.Trainer(max_epochs=1)
_ = trainer.fit(model, vad)

## Speaker change detection

In [ ]:
from pyannote.audio.tasks import SpeakerChangeDetection
scd = SpeakerChangeDetection(protocol, duration=2., batch_size=32, num_workers=4)
model = SimpleSegmentationModel(task=scd)
trainer = pl.Trainer(max_epochs=1)
_ = trainer.fit(model, scd)

Overlapped speech detection

In [ ]:
from pyannote.audio.tasks import OverlappedSpeechDetection
ovl = OverlappedSpeechDetection(protocol, duration=2., batch_size=32, num_workers=4)
model = SimpleSegmentationModel(task=ovl)
trainer = pl.Trainer(max_epochs=1)
_ = trainer.fit(model, ovl)

Multi-task segmentation

In [ ]:
from pyannote.audio.tasks import MultiTaskSegmentation
seg = MultiTaskSegmentation(protocol, vad=True, scd=True, osd=True, duration=2., batch_size=32, num_workers=4)
model = MultiTaskSegmentationModel(task=seg)
trainer = pl.Trainer(max_epochs=1)
_ = trainer.fit(model, seg)

## Speaker tracking

In [ ]:
from pyannote.audio.tasks import SpeakerTracking
spk = SpeakerTracking(protocol, duration=2., batch_size=32, num_workers=4)
model = SimpleSegmentationModel(task=spk)
trainer = pl.Trainer(max_epochs=1)
_ = trainer.fit(model, spk)

Speaker embedding

In [ ]:
from pyannote.audio.tasks.speaker_verification.task import SpeakerEmbeddingArcFace
emb = SpeakerEmbeddingArcFace(protocol, duration=2., num_workers=4)
model = SimpleEmbeddingModel(task=emb)
trainer = pl.Trainer(max_epochs=1)
_ = trainer.fit(model, emb)