---
task_categories:
  - audio-classification
tags:
  - audio
  - bioacoustics
  - dolphin
  - whistle
---

# Dolphin whistle classification

Short whistle clips with metadata, fundamental-frequency tracks, F0 spectrograms, and integer class IDs in `main_category`.

**Whistle sub-category:** fine-grained whistle identity is carried in the **`whistle_type`** column (integer IDs). This is the primary label column for sub-category work; `main_category` groups whistles at a coarser level.

Splits (`train`, `validation`, `test`) are disjoint by recording session.

## Columns

- `audio`, `main_category`, `name`, `onset`, `offset`, `duration`, `recording_duration`, **`whistle_type`**, `whistle_name`, `f0_time`, `f0_hz`, `f0_conf`, `f0_ok`, `f0_bad_reason`, `f0_spectrogram`, `snr_db`

## Usage

```python
from datasets import load_dataset

# Default: one `datasets.Dataset` per split (`train`, `validation`, `test`)
ds = load_dataset("dolphinteam/Whistle-Classification")
```

Load **all rows in a single dataset** (no split separation; train, validation, and test concatenated in order):

```python
from datasets import load_dataset

full = load_dataset("dolphinteam/Whistle-Classification", split="train+validation+test")
```

Equivalent if you already loaded the split dict and want one table:

```python
from datasets import concatenate_datasets

ds = load_dataset("dolphinteam/Whistle-Classification")
full = concatenate_datasets([ds["train"], ds["validation"], ds["test"]])
```