Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

fix index errors in separation pipeline #1769

Draft
wants to merge 1 commit into
base: develop
Choose a base branch
from
Draft

fix index errors in separation pipeline #1769

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
17 changes: 13 additions & 4 deletions pyannote/audio/pipelines/speech_separation.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -91,7 +91,7 @@ class SpeechSeparation(SpeakerDiarizationMixin, Pipeline):

Usage
-----
>>> pipeline = SpeakerDiarization()
>>> pipeline = SpeechSeparation()
>>> diarization, separation = pipeline("/path/to/audio.wav")
>>> diarization, separation = pipeline("/path/to/audio.wav", num_speakers=4)
>>> diarization, separation = pipeline("/path/to/audio.wav", min_speakers=2, max_speakers=10)
Expand Down Expand Up @@ -236,7 +236,7 @@ def get_segmentations(
segmentations, separations = file[self.CACHED_SEGMENTATION]
else:
segmentations, separations = self._segmentation(file, hook=hook)
file[self.CACHED_SEGMENTATION] = segmentations
file[self.CACHED_SEGMENTATION] = (segmentations, separations)
else:
segmentations, separations = self._segmentation(file, hook=hook)

Expand Down Expand Up @@ -441,7 +441,6 @@ def reconstruct(
clustered_segmentations, segmentations.sliding_window
)
return clustered_segmentations
return self.to_diarization(clustered_segmentations, count)

def apply(
self,
Expand Down Expand Up @@ -583,7 +582,7 @@ def apply(

# reconstruct discrete diarization from raw hard clusters

# keep track of inactive speakers
# keep track of inactive speakers at chunk level
inactive_speakers = np.sum(binarized_segmentations.data, axis=1) == 0
# shape: (num_chunks, num_speakers)

Expand All @@ -594,17 +593,27 @@ def apply(
count,
)
discrete_diarization = self.to_diarization(discrete_diarization, count)

# remove inactive speakers at the audio level from the diarization
active_speakers = np.sum(discrete_diarization, axis=0) > 0
# shape: (num_speakers, )
discrete_diarization.data = discrete_diarization.data[:, active_speakers]
# shape: (num_frames, num_active_speakers)

hook("discrete_diarization", discrete_diarization)
clustered_separations = self.reconstruct(separations, hard_clusters, count)
frame_duration = separations.sliding_window.duration / separations.data.shape[1]
frames = SlidingWindow(step=frame_duration, duration=2 * frame_duration)

sources = Inference.aggregate(
clustered_separations,
frames=frames,
hamming=True,
missing=0.0,
skip_average=True,
)
sources.data = sources.data[:, active_speakers]

# zero-out sources when speaker is inactive
# WARNING: this should be rewritten to avoid huge memory consumption
if self.separation.leakage_removal:
Expand Down
Loading