Struct SpeechCommandsDataset 

pub struct SpeechCommandsDataset { /* private fields */ }

Speech Commands dataset from Huggingface v0.02. See Speech Commands dataset.

The data is downloaded from Huggingface and stored in a SQLite database (3.0 GB). The dataset contains 99,720 audio samples of 2,607 people saying 35 different words.

NOTE: The most samples are under 1 second long but there are some with pure background noise that need splitting into shorter segmants.

The labels are 20 target words, silence and other words.

The dataset is split into 3 parts:

• train: 84,848 audio files
• test: 4,890 audio files
• validation: 9,982 audio files

Implementations

impl SpeechCommandsDataset

pub fn new(split: &str) -> SpeechCommandsDataset

Create a new dataset with the given split.

pub fn train() -> SpeechCommandsDataset

Create a new dataset with the train split.

pub fn test() -> SpeechCommandsDataset

Create a new dataset with the test split.

pub fn validation() -> SpeechCommandsDataset

Create a new dataset with the validation split.

pub fn num_classes() -> usize

Returns the number of classes in the dataset

Trait Implementations

impl Dataset<SpeechItem, SqliteDatasetError> for SpeechCommandsDataset

fn get(&self, index: usize) -> Result<SpeechItem, SqliteDatasetError>

Gets the item at the given index.

fn len(&self) -> usize

Gets the number of items in the dataset.

fn get_many(&self, indexes: Vec<usize>) -> Result<Vec<I>, E>

Gets the items at given indexes

fn is_empty(&self) -> bool

Checks if the dataset is empty.

fn iter(&self) -> DatasetIterator<'_, I, E>

where Self: Sized,

Returns an iterator over the dataset.