오디오 데이터셋

저자: Moto Hira 번역: 백승엽

``torchaudio``는 공용으로 접근할 수 있는 일반적인 데이터셋에 쉽게 접근할 수 있는 기능을 제공합니다. 사용할 수 있는 데이터셋 목록은 공식 문서를 참고하세요.

import torch
import torchaudio

print(torch.__version__)
print(torchaudio.__version__)

2.3.1
2.3.1

import os

import IPython

import matplotlib.pyplot as plt


_SAMPLE_DIR = "_assets"
YESNO_DATASET_PATH = os.path.join(_SAMPLE_DIR, "yes_no")
os.makedirs(YESNO_DATASET_PATH, exist_ok=True)


def plot_specgram(waveform, sample_rate, title="Spectrogram"):
    waveform = waveform.numpy()

    figure, ax = plt.subplots()
    ax.specgram(waveform[0], Fs=sample_rate)
    figure.suptitle(title)
    figure.tight_layout()

여기서 torchaudio.datasets.YESNO 데이터셋의 사용 방법을 볼 수 있습니다.

dataset = torchaudio.datasets.YESNO(YESNO_DATASET_PATH, download=True)

  0%|          | 0.00/4.49M [00:00<?, ?B/s]
  3%|2         | 128k/4.49M [00:00<00:24, 190kB/s]
  6%|5         | 256k/4.49M [00:00<00:14, 298kB/s]
  8%|8         | 384k/4.49M [00:01<00:11, 360kB/s]
 11%|#1        | 512k/4.49M [00:01<00:10, 418kB/s]
 14%|#3        | 640k/4.49M [00:01<00:08, 456kB/s]
 17%|#6        | 768k/4.49M [00:01<00:08, 484kB/s]
 22%|##2       | 1.00M/4.49M [00:02<00:05, 646kB/s]
 25%|##5       | 1.12M/4.49M [00:02<00:05, 618kB/s]
 28%|##7       | 1.25M/4.49M [00:02<00:05, 609kB/s]
 33%|###3      | 1.50M/4.49M [00:02<00:04, 753kB/s]
 36%|###6      | 1.62M/4.49M [00:03<00:04, 707kB/s]
 42%|####1     | 1.88M/4.49M [00:03<00:03, 836kB/s]
 47%|####7     | 2.12M/4.49M [00:03<00:02, 922kB/s]
 53%|#####2    | 2.38M/4.49M [00:03<00:02, 978kB/s]
 59%|#####8    | 2.62M/4.49M [00:04<00:01, 1.00MB/s]
 61%|######1   | 2.75M/4.49M [00:04<00:02, 865kB/s]
 70%|######9   | 3.12M/4.49M [00:04<00:01, 1.10MB/s]
 75%|#######5  | 3.38M/4.49M [00:04<00:01, 1.12MB/s]
 81%|########  | 3.62M/4.49M [00:05<00:00, 1.14MB/s]
 86%|########6 | 3.88M/4.49M [00:05<00:00, 1.13MB/s]
 95%|#########4| 4.25M/4.49M [00:05<00:00, 1.25MB/s]
100%|##########| 4.49M/4.49M [00:05<00:00, 851kB/s]

i = 1
waveform, sample_rate, label = dataset[i]
plot_specgram(waveform, sample_rate, title=f"Sample {i}: {label}")
IPython.display.Audio(waveform, rate=sample_rate)

i = 3
waveform, sample_rate, label = dataset[i]
plot_specgram(waveform, sample_rate, title=f"Sample {i}: {label}")
IPython.display.Audio(waveform, rate=sample_rate)

i = 5
waveform, sample_rate, label = dataset[i]
plot_specgram(waveform, sample_rate, title=f"Sample {i}: {label}")
IPython.display.Audio(waveform, rate=sample_rate)

i = 5
waveform, sample_rate, label = dataset[i]
plot_specgram(waveform, sample_rate, title=f"Sample {i}: {label}")
IPython.display.Audio(waveform, rate=sample_rate)

• 
•