feat(pipeline): add Waveform and SampleRate preprocessors (pyannote#1593

)

CHANGELOG.md

@@ -1,9 +1,10 @@
 # Changelog
 
-## develop branch
+## develop
 
 ### New features
 
+- feat(pipeline): add `Waveform` and `SampleRate` preprocessors
 - feat(model): add `num_frames` and `receptive_field` to segmentation models
 
 ## Version 3.1.1 (2023-12-01)

pyannote/audio/utils/preprocessors.py

@@ -27,12 +27,13 @@
 # Hervé BREDIN - http://herve.niderb.fr
 
 from functools import reduce
-from itertools import chain
 from typing import Dict, List, Optional, Set
 
 from pyannote.core import Annotation, Segment
 from pyannote.database import ProtocolFile
 
+from pyannote.audio.core.io import Audio, get_torchaudio_info
+
 
 class LowerTemporalResolution:
     """Artificially degrade temporal resolution of reference annotation
@@ -50,7 +51,6 @@ def __init__(self, resolution: float = 0.1):
         self.resolution = resolution
 
     def __call__(self, current_file: ProtocolFile) -> Annotation:
-
         annotation = current_file["annotation"]
         new_annotation = annotation.empty()
 
@@ -128,3 +128,17 @@ def __call__(self, current_file: ProtocolFile) -> Annotation:
                 derived[seg] = intersect_label
 
         return derived
+
+
+class Waveform:
+    def __init__(self):
+        self._audio = Audio()
+
+    def __call__(self, file: ProtocolFile):
+        waveform, _ = self._audio(file)
+        return waveform
+
+
+class SampleRate:
+    def __call__(self, file: ProtocolFile):
+        return get_torchaudio_info(file).sample_rate