sdialog/sound_events

Sound Events Dataset

This dataset contains 64 audio recordings of 23 different sound events classes.

Dataset Statistics

• Total Audio Files: 64
• Total Classes: 23
• Format: WAV/MP3/OGG (converted to Parquet for Hugging Face Datasets)

Classes and Descriptions

The dataset covers the following sound events:

Label	Description
searching_wooden_drawer	Rummaging through items inside a wooden drawer
pulling_wooden_chair_out	Dragging or moving a wooden chair away from a desk or table
switch_pen_off	Retracting the writing tip of a pen mechanism
switch_light_on	Flipping a switch to activate a light source
open_door	Unlatching and opening a door to allow entry or exit
turning_or_sorting_papers	Shuffling, flipping, or arranging sheets of paper
open_metallic_drawer	Pulling open a standard metal drawer (e.g., filing cabinet)
close_wooden_drawer	Pushing a wooden drawer back into its closed position
close_door	Shutting a door to the closed position
pulling_wooden_chair_in	Dragging or moving a wooden chair toward a desk or table
walking	Moving on foot across a surface
sitting_wooden_chair	The action of sitting down onto a wooden chair
keyboard_typing	Pressing keys on a computer keyboard
moving_pen_tool_metal_holder	Placing a pen/tool into or removing it from a metal container
open_small_metallic_drawer	Pulling open a small metal drawer or compartment
pull_up_door_knob	Pulling a door handle upwards
close_metallic_drawer	Pushing a metal drawer back into its closed position
pull_down_door_knob	Pressing a door handle downwards to unlatch
switch_pen_on	Extending the writing tip of a pen mechanism
knock_door	Striking a door surface with knuckles
open_wooden_drawer	Pulling open a wooden drawer
switch_light_off	Flipping a switch to deactivate a light source
tearing_ripping_medical_roll_paper	Tearing paper from a medical exam table roll or dispenser

Structure

The dataset is organized in a sound_audios/ folder with subdirectories for each class. A metadata.csv file provides the file paths, labels, and descriptions.

Columns

• audio: Path to the audio file.
• label: The category label of the sound event.
• description: A text description of the sound event.

Usage

from datasets import load_dataset

ds = load_dataset("sdialog/sound_events")