참고자료:
깨끗한
TFDS에 이 데이터세트를 로드하려면 다음 명령어를 사용하세요.
ds = tfds.load('huggingface:librispeech_asr/clean')
- 설명 :
LibriSpeech is a corpus of approximately 1000 hours of read English speech with sampling rate of 16 kHz,
prepared by Vassil Panayotov with the assistance of Daniel Povey. The data is derived from read
audiobooks from the LibriVox project, and has been carefully segmented and aligned.87
Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .flac format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:
python
import soundfile as sf
def map_to_array(batch):
speech_array, _ = sf.read(batch["file"])
batch["speech"] = speech_array
return batch
dataset = dataset.map(map_to_array, remove_columns=["file"])
- 라이센스 : 알려진 라이센스 없음
- 버전 : 2.1.0
- 분할 :
| 나뉘다 | 예 |
|---|---|
'test' | 2620 |
'train.100' | 28539 |
'train.360' | 104014 |
'validation' | 2703 |
- 특징 :
{
"file": {
"dtype": "string",
"id": null,
"_type": "Value"
},
"audio": {
"sampling_rate": 16000,
"mono": true,
"decode": true,
"id": null,
"_type": "Audio"
},
"text": {
"dtype": "string",
"id": null,
"_type": "Value"
},
"speaker_id": {
"dtype": "int64",
"id": null,
"_type": "Value"
},
"chapter_id": {
"dtype": "int64",
"id": null,
"_type": "Value"
},
"id": {
"dtype": "string",
"id": null,
"_type": "Value"
}
}
다른
TFDS에 이 데이터세트를 로드하려면 다음 명령어를 사용하세요.
ds = tfds.load('huggingface:librispeech_asr/other')
- 설명 :
LibriSpeech is a corpus of approximately 1000 hours of read English speech with sampling rate of 16 kHz,
prepared by Vassil Panayotov with the assistance of Daniel Povey. The data is derived from read
audiobooks from the LibriVox project, and has been carefully segmented and aligned.87
Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .flac format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:
python
import soundfile as sf
def map_to_array(batch):
speech_array, _ = sf.read(batch["file"])
batch["speech"] = speech_array
return batch
dataset = dataset.map(map_to_array, remove_columns=["file"])
- 라이센스 : 알려진 라이센스 없음
- 버전 : 2.1.0
- 분할 :
| 나뉘다 | 예 |
|---|---|
'test' | 2939 |
'train.500' | 148688 |
'validation' | 2864 |
- 특징 :
{
"file": {
"dtype": "string",
"id": null,
"_type": "Value"
},
"audio": {
"sampling_rate": 16000,
"mono": true,
"decode": true,
"id": null,
"_type": "Audio"
},
"text": {
"dtype": "string",
"id": null,
"_type": "Value"
},
"speaker_id": {
"dtype": "int64",
"id": null,
"_type": "Value"
},
"chapter_id": {
"dtype": "int64",
"id": null,
"_type": "Value"
},
"id": {
"dtype": "string",
"id": null,
"_type": "Value"
}
}
모두
TFDS에 이 데이터세트를 로드하려면 다음 명령어를 사용하세요.
ds = tfds.load('huggingface:librispeech_asr/all')
- 설명 :
LibriSpeech is a corpus of approximately 1000 hours of read English speech with sampling rate of 16 kHz,
prepared by Vassil Panayotov with the assistance of Daniel Povey. The data is derived from read
audiobooks from the LibriVox project, and has been carefully segmented and aligned.87
- 라이센스 : 알려진 라이센스 없음
- 버전 : 2.1.0
- 분할 :
| 나뉘다 | 예 |
|---|---|
'test.clean' | 2620 |
'test.other' | 2939 |
'train.clean.100' | 28539 |
'train.clean.360' | 104014 |
'train.other.500' | 148688 |
'validation.clean' | 2703 |
'validation.other' | 2864 |
- 특징 :
{
"file": {
"dtype": "string",
"id": null,
"_type": "Value"
},
"audio": {
"sampling_rate": 16000,
"mono": true,
"decode": true,
"id": null,
"_type": "Audio"
},
"text": {
"dtype": "string",
"id": null,
"_type": "Value"
},
"speaker_id": {
"dtype": "int64",
"id": null,
"_type": "Value"
},
"chapter_id": {
"dtype": "int64",
"id": null,
"_type": "Value"
},
"id": {
"dtype": "string",
"id": null,
"_type": "Value"
}
}