Browse Source
Replace resampy with torchaudio
Signed-off-by: Jael Gu <mengjia.gu@zilliz.com>
main
1 changed files with
4 additions and
2 deletions
-
nn_fingerprint.py
|
|
@ -20,8 +20,8 @@ from pathlib import Path |
|
|
from typing import List |
|
|
from typing import List |
|
|
|
|
|
|
|
|
import torch |
|
|
import torch |
|
|
|
|
|
import torchaudio |
|
|
import numpy |
|
|
import numpy |
|
|
import resampy |
|
|
|
|
|
import onnxruntime |
|
|
import onnxruntime |
|
|
|
|
|
|
|
|
from towhee.operator.base import NNOperator |
|
|
from towhee.operator.base import NNOperator |
|
|
@ -119,9 +119,11 @@ class NNFingerprint(NNOperator): |
|
|
audio = audio[None, :] |
|
|
audio = audio[None, :] |
|
|
assert len(audio.shape) == 2 |
|
|
assert len(audio.shape) == 2 |
|
|
audio = self.int2float(audio) |
|
|
audio = self.int2float(audio) |
|
|
|
|
|
audio = torch.from_numpy(audio) |
|
|
|
|
|
|
|
|
if sr != self.params['sample_rate']: |
|
|
if sr != self.params['sample_rate']: |
|
|
audio = resampy.resample(audio, sr, self.params['sample_rate']) |
|
|
|
|
|
|
|
|
resampler = torchaudio.transforms.Resample(sr, self.params['sample_rate'], dtype=audio.dtype) |
|
|
|
|
|
audio = resampler(audio) |
|
|
|
|
|
|
|
|
wav = preprocess_wav(audio, |
|
|
wav = preprocess_wav(audio, |
|
|
segment_size=int(self.params['sample_rate'] * self.params['segment_size']), |
|
|
segment_size=int(self.params['sample_rate'] * self.params['segment_size']), |
|
|
|
