RepeatReal / README.md
Hamozwa's picture
Update README.md
bb3f79c verified
|
Raw History Blame Contribute Delete
2.25 kB
metadata
configs:
  - config_name: clocks
    data_dir: clocks
  - config_name: heartbeats
    data_dir: heartbeats
  - config_name: dolphins
    data_dir: dolphins/test_padded
license: cc-by-nc-4.0

Dataset Card for RepeatReal

Real-world portion of the audio datasets referenced in Class-Agnostic Audio Repetition Counting.

Code: https://github.com/Hamozwa/audio-counting

Dataset Summary

Unlike the synthetic RepeatSynth data, these are real, variable-length recordings across three domains, used as held-out test sets for models trained only on synthetic data.

Config Domain Source
clocks Mechanical (ticking clocks) BBC Sound Effects
heartbeats Medical (heart sounds) PhysioNet/CinC Challenge 2016
dolphins Ecological (dolphin click trains) Song et al., Springer Nature Figshare

Each config has a single test split.

from datasets import load_dataset
ds = load_dataset("Hamozwa/RepeatReal", "clocks", split="test")

Dataset Structure

Field Type Description
audio Audio The (variable-length) wav clip
location string Original filename
repetitions int Ground-truth repetition count
start_time float (clocks only) seconds to trim from the start of the clip before scoring

Dataset Sources

Licensing

Each domain comes from a different source with different terms. Treat the dataset as a whole under the most restrictive of the three (Clocks) unless you separate out only the domains you need.

Clocks — BBC Sound Effects. Full terms: https://sound-effects.bbcrewind.co.uk/licensing

Heartbeats — PhysioNet/CinC Challenge 2016. Licensed under the Open Data Commons Attribution License v1.0

Dolphins — Song et al., Springer Nature Figshare. Licensed under CC BY 4.0

Citation

[Your paper's BibTeX entry here]