FSD50K: An open dataset of human-labeled sound events

This is a mirror of the FSD50K sound event dataset. The original files were converted from WAV to Opus to reduce the size and accelerate streaming.

• Sampling rate: 48 kHz
• Channels: 1
• Format: Opus
• Splits:
  • Dev: 80 hours, 40966 clips.
  • Eval: 28 hours, 10231 clips.
• License: FSD50K is released under CC-BY. However, each clip has its own licence. Clip licenses include CC0, CC-BY, CC-BY-NC and CC Sampling+. Clip licenses are specified in dev_clips_info_FSD50K.json and eval_clips_info_FSD50K.json.
• Source: https://zenodo.org/records/4060432
• Paper: FSD50K: An Open Dataset of Human-Labeled Sound Events

Usage

import io

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

for item in load_dataset(
    "philgzl/fsd50k",
    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

@article{fonseca2022fsd50k,
  title = {{FSD50K}: {An} open dataset of human-labeled sound events},
  author = {Fonseca, Eduardo and Favory, Xavier and Pons, Jordi and Font, Frederic and Serra, Xavier},
  journal = {IEEE/ACM} Trans. Audio, Speech, Lang. Process.},
  volume = {30},
  pages = {829--852},
  year = {2022},
}