---
dataset_info:
  features:
  - name: 'Unnamed: 0'
    dtype: int64
  - name: path
    dtype: audio
  - name: label
    dtype: int64
  - name: fold
    dtype: int64
  splits:
  - name: train
    num_bytes: 529286612.8
    num_examples: 1200
  - name: valid
    num_bytes: 176428850.0
    num_examples: 400
  - name: test
    num_bytes: 176428850.0
    num_examples: 400
  - name: train_low
    num_bytes: 176428850.0
    num_examples: 400
  download_size: 926457942
  dataset_size: 1058573162.8
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: valid
    path: data/valid-*
  - split: test
    path: data/test-*
  - split: train_low
    path: data/train_low-*
---
# Dataset Card for "beans_esc50"

## Dataset Description

**Paper:** https://doi.org/10.1145/2733373.2806390
          
## Dataset Summary

This dataset contains environmental audio recordings including animal, nature, human (non-speech), interior/domestic, and exterior/urban sounds with splits and preprocessing like described in BEANS. It is used for **classification** tasks and is an **auxiliary** dataset in BEANS.

## Data Splits

 | train | train_low | valid | test | 
 |------|-----| ------ | ------ | 
 | 1200 |  400 | 400 | 400 |