---
license: apache-2.0
---

# DECRO Dataset

Please use the download scripts from https://github.com/XIAOYixuan/AUDDT/tree/yixuan-dev
to download and process the dataset.

```
chmod +x download/get_decro.sh
./download/get_decro.sh
```

## Label Distribution

### Chinese (ch) Splits

**ch_train** (26,850 samples):
- spoof: 17,850 samples
- bonafide: 9,000 samples

**ch_dev** (18,124 samples):
- spoof: 12,015 samples
- bonafide: 6,109 samples

**ch_eval** (18,124 samples):
- spoof: 12,015 samples
- bonafide: 6,109 samples

### English (en) Splits

**en_train** (22,541 samples):
- spoof: 17,413 samples
- bonafide: 5,128 samples

**en_dev** (13,552 samples):
- spoof: 10,503 samples
- bonafide: 3,049 samples

**en_eval** (19,190 samples):
- spoof: 14,884 samples
- bonafide: 4,306 samples

## Data Attributes

```
         ID                                                     path  label dataset_name
0  20032114  petrichorwq-DECRO-dataset-6fc9884/ch_train/20032114.wav  spoof        DECRO
1  20032116  petrichorwq-DECRO-dataset-6fc9884/ch_train/20032116.wav  spoof        DECRO
2  20032118  petrichorwq-DECRO-dataset-6fc9884/ch_train/20032118.wav  spoof        DECRO
```

## How to Import

```python
import pandas as pd

# Load a split
df = pd.read_parquet("ch_train.parquet")
print(df.head())
```