forked from Dudnik-Ilia/SoundStream
-
Notifications
You must be signed in to change notification settings - Fork 0
/
dataset.py
69 lines (60 loc) · 2.38 KB
/
dataset.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
import torch
import torchaudio
import os
from torch.utils.data import Dataset
import random
# Accumulate all file names
def get_file_names(start_dir: str):
"""
Look in the start_dir for all files ending with flac or wav
Returns lists with full paths and just names (was meant to be used in conversion and saving)
"""
path_list = []
file_list = []
for root, dirs, files in os.walk(start_dir):
for file in files:
if file.lower().endswith(('.flac', '.wav')):
file_list.append(file)
path_list.append(os.path.join(root, file))
return path_list, file_list
def convert_audio(wav: torch.Tensor, sr: int, target_sr: int, target_channels: int):
"""
Audio resampling to the target sample rate
"""
assert wav.dim() >= 2, "Audio tensor must have at least 2 dimensions"
assert wav.shape[-2] in [1, 2], "Audio must be mono or stereo."
*shape, channels, length = wav.shape
if target_channels == 1:
wav = wav.mean(-2, keepdim=True)
elif target_channels == 2:
wav = wav.expand(*shape, target_channels, length)
elif channels == 1:
wav = wav.expand(target_channels, -1)
else:
raise RuntimeError(f"Impossible to convert from {channels} to {target_channels}")
wav = torchaudio.transforms.Resample(sr, target_sr)(wav)
return wav
class NSynthDataset(Dataset):
"""Dataset to load NSynth data."""
def __init__(self, audio_dir, sample_rate=24000, tensor_cut=36000):
super().__init__()
self.filenames, _ = get_file_names(audio_dir)
self.sr = sample_rate
self.tensor_cut = tensor_cut
def __len__(self):
return len(self.filenames)
def __getitem__(self, index):
"""
During getting an item we transform it to needed sample rate
"""
waveform, sample_rate = torchaudio.load(self.filenames[index])
if sample_rate != self.sr:
waveform = convert_audio(wav=waveform, sr=sample_rate, target_sr=self.sr, target_channels=1)
# Cut the length of audio
if self.tensor_cut > 0:
if waveform.size()[1] > self.tensor_cut:
# random start point
start = random.randint(0, waveform.size()[1] - self.tensor_cut - 1)
# cut tensor
waveform = waveform[:, start:start + self.tensor_cut]
return waveform