| from typing import List | |
| from .base import SequentialDataPipe | |
| from .common_pipes import LoadAudio, RandomCrop, SetOutputKeys | |
| class SpeakerVerificationPipe(SequentialDataPipe): | |
| """ | |
| each item in the input dataset should have: | |
| wav_path: str | |
| label: str | |
| """ | |
| def __init__( | |
| self, | |
| audio_sample_rate: int = 16000, | |
| audio_channel_reduction: str = "first", | |
| random_crop_secs: float = -1, | |
| sox_effects: List[List] = None, | |
| ): | |
| pipes = [ | |
| LoadAudio( | |
| audio_sample_rate=audio_sample_rate, | |
| audio_channel_reduction=audio_channel_reduction, | |
| sox_effects=sox_effects, | |
| ), | |
| ] | |
| output_keys = dict( | |
| x="wav", | |
| x_len="wav_len", | |
| label="label", | |
| unique_name="id", | |
| ) | |
| if random_crop_secs != -1: | |
| pipes.append( | |
| RandomCrop(sample_rate=audio_sample_rate, max_secs=random_crop_secs) | |
| ) | |
| output_keys["x"] = "wav_crop" | |
| output_keys["x_len"] = "wav_crop_len" | |
| pipes.append(SetOutputKeys(output_keys)) | |
| super().__init__(*pipes) | |