Audio Embedding with data2vec

author: David Wang

Description

This operator extracts features for audio with data2vec. The core idea is to predict latent representations of the full input data based on a masked view of the input in a self-distillation setup using a standard Transformer architecture.

Code Example

Generate embeddings for the audio "test.wav".

Write the pipeline in simplified style:

import towhee

(
    towhee.glob('test.wav')
          .audio_decode.ffmpeg()
          .runas_op(func=lambda x:[y[0] for y in x])
          .audio_embedding.data2vec()
          .show()
)

Write a same pipeline with explicit inputs/outputs name specifications:

import towhee

(
    towhee.glob['path']('test.wav')
          .audio_decode.ffmpeg['path', 'frames']()
          .runas_op['frames', 'frames'](func=lambda x:[y[0] for y in x])
          .audio_embedding.data2vec['frames', 'vecs'](model_name="facebook/data2vec-audio-base-960h") 
          .show()
)

Factory Constructor

Create the operator via the following factory method

data2vec(model_name='facebook/data2vec-audio-base')

Parameters:

​ model_name: str

The model name in string. The default value is "facebook/data2vec-audio-base-960h".

Supported model name:

• facebook/data2vec-audio-base-960h
• facebook/data2vec-audio-large-960h
• facebook/data2vec-audio-base
• facebook/data2vec-audio-base-100h
• facebook/data2vec-audio-base-10m
• facebook/data2vec-audio-large
• facebook/data2vec-audio-large-100h
• facebook/data2vec-audio-large-10m

Interface

An audio embedding operator generates vectors in numpy.ndarray given an audio file path or towhee audio frames.

Parameters:

​ data: List[towhee.types.audio_frame.AudioFrame]

​ Input audio data is a list of towhee audio frames. The input data should represent for an audio longer than 0.9s.

Returns: numpy.ndarray

​ The audio embedding extracted by model.