ertk.preprocessing.audioset.VGGishExtractor

class ertk.preprocessing.audioset.VGGishExtractor(config: VGGishExtractorConfig)

Bases: FeatureExtractor, AudioClipProcessor

A VGGish extractor.

__init__(config: VGGishExtractorConfig) → None

Methods

__init__(config)
finish()	Perform any cleanup necesasry (e.g.
process_batch(batch, **kwargs)	Process a batch of instances.
process_instance(x, **kwargs)	Process a single audio clip.

Inherited Methods

friendly_name()	Get the friendly name for this processor.
get_config_type()	Get the configuration type for this processor.
get_default_config()	Get the default configuration for this processor.
get_processor_class(name)	Get the class for the named processor.
make_processor(name, config)	Create an instance of the named processor.
process_all(xs, batch_size, **kwargs)	Process all instances in batches.
process_file(path[, sr])	Process individual audio file.
process_files(paths, batch_size[, sr])	Process a set of files.
valid_processors()	Get a list of all registered processor names.

Attributes

dim	The dimensionality of the extracted features.
feature_names	The names of the features produced by this processor.
is_sequence	Whether this FeatureExtractor yields sequence features.
config	The configuration for this processor.

config: VGGishExtractorConfig

The configuration for this processor.

property dim: int

The dimensionality of the extracted features.

property feature_names: List[str]

The names of the features produced by this processor.

finish() → None

Perform any cleanup necesasry (e.g. closing files, unloading models, etc.)

property is_sequence: bool

Whether this FeatureExtractor yields sequence features.

process_batch(batch: Iterable[ndarray] | ndarray, **kwargs) → List[ndarray]

Process a batch of instances. By default this simply calls process_instance() on each instance in the batch.

Parameters:

batch: np.ndarray or iterable of arrays

The batch of instances to process.

Returns:

processed: list of np.ndarray

A list of processed instances.

process_instance(x: ndarray, **kwargs) → ndarray

Process a single audio clip.

Parameters:

x: np.ndarray

The audio data to process.

Returns:

result: np.ndarray

The processed instance.