philgzl/tut2020

TUT Urban Acoustic Scenes 2020 Mobile Development and Evaluation datasets

This is a mirror of the TUT Urban Acoustic Scenes 2020 Mobile Development and Evaluation datasets. The original files were converted from WAV to Opus to reduce the size and accelerate streaming.

• Sampling rate: 48 kHz (resampled from 44.1 kHz to support Opus format)
• Channels: 1
• Format: Opus
• Splits:
  • Dev: 64 hours, 23035 clips.
  • Eval: 33 hours, 11880 clips.
• License: Custom. See LICENSE file.
• Source: https://zenodo.org/records/3819968 and https://zenodo.org/records/3685828
• Paper: TUT Database for Acoustic Scene Classification and Sound Event Detection

See here for more details about the dataset.

Usage

import io

import soundfile as sf
from datasets import Features, Value, load_dataset

for item in load_dataset(
    "philgzl/tut2020",
    split="dev",
    streaming=True,
    features=Features({"audio": Value("binary"), "name": Value("string")}),
):
    print(item["name"])
    buffer = io.BytesIO(item["audio"])
    x, fs = sf.read(buffer)
    # do stuff...

Citation

@inproceedings{mesaros2016tut,
  title = {{TUT} database for acoustic scene classification and sound event detection},
  author = {Mesaros, Annamaria and Heittola, Toni and Virtanen, Tuomas},
  booktitle = {Proc. EUSIPCO},
  pages = {1128--1132},
  year = {2016},
}