The official Pytorch implementation of Fast Context-based Pitch Estimation (FCPE)
Project description
TorchFCPE
Useage
from torchfcpe import spawn_bundled_infer_model
import torch
import librosa
# configure device and target hop_size
device = 'cpu'
sr = 16000
hop_size = 160
# load audio
audio, sr = librosa.load('test.wav', sr=sr)
audio = librosa.to_mono(audio)
audio_length = len(audio)
f0_target_length=(audio_length // hop_size) + 1
audio = torch.from_numpy(audio).float().unsqueeze(0).unsqueeze(-1).to(device)
# load model
model = spawn_bundled_infer_model(device=device)
# infer
'''
audio: wav, torch.Tensor
sr: sample rate
decoder_mode: [Optional] 'local_argmax' is recommended
threshold: [Optional] threshold for V/UV decision, 0.006 is recommended
f0_min: [Optional] minimum f0
f0_max: [Optional] maximum f0
interp_uv: [Optional] whether to interpolate unvoiced frames
output_interp_target_length: [Optional] If not None, the output f0 will be
interpolated to the target length
'''
f0 = model.infer(
audio,
sr=sr,
decoder_mode='local_argmax',
threshold=0.006,
f0_min=80,
f0_max=880,
interp_uv=False,
output_interp_target_length=f0_target_length,
)
print(f0)
# the model is son class of torch.nn.Module, so you can use it as a normal pytorch model
# example: change device
model = model.to(device)
# example: compile model
model = torch.compile(model)
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distributions
No source distribution files available for this release.See tutorial on generating distribution archives.
Built Distribution
torchfcpe-0.0.4-py3-none-any.whl
(40.2 MB
view hashes)
Close
Hashes for torchfcpe-0.0.4-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | f042c463d850d76c6f4899a0b84f0b694bb560adf05f4de951097a756d17472d |
|
MD5 | 4db27f64a096b8ee959742bde09c9523 |
|
BLAKE2b-256 | 159866da447498a2835b01cb3ae851e1333035a9fbf228e5c6f3e3c4351c06b6 |