GYdevy/volleyball-whistles

Volleylitics Whistle Detection Dataset

Overview

This dataset contains full match audio recordings and whistle annotations for training and evaluating whistle detection models in indoor volleyball environments.

The dataset is designed for:

• DSP-based whistle detection
• Audio classification (CNN)
• Acoustic event detection research

---

Audio Format

• Format: WAV
• Sample rate: 22.05kHz

---

Structure

match1.wav match2.wav
...

whistles_all.json whistles_all_anchored.json whistles_all_reanchored.json

---

Annotation Format

Each annotation JSON contains:

{
        "match_id": "match1",
        "whistle_id": 0,
        "time": 395.705,
        "type": "other",
        "t_raw": 395.705,
        "global_id": 0,
        "t_anchor": 395.663
    },
    {
        "match_id": "match1",
        "whistle_id": 1,
        "time": 396.473,
        "type": "other",
        "t_raw": 396.473,
        "global_id": 1,
        "t_anchor": 396.303
    },

t_raw is the time of a manual labeling to a whistle sound. t_anchor is the time of anchoring by some features (band_peak)