AIME

```
from datasets import load_dataset
dataset = load_dataset('disco-eth/AIME')
```

# AIME: AI Music Evaluation Dataset

The AIME dataset contains 6,000 audio tracks generated by 12 music generation models in addition to 500 tracks from [MTG-Jamendo](https://mtg.github.io/mtg-jamendo-dataset/).
The prompts used to generate music are combinations of representative and diverse tags from the MTG-Jamendo dataset.

The AIME dataset consists of two subsets. The AIME audio dataset and the [AIME survey dataset](https://huggingface.co/datasets/disco-eth/AIME-survey).

The dataset contains the following fields:

- **id**: Used to map the survey data to audio files.
- **model**: String with model name used to generate the audio sample (can be one of 'Udio', 'MusicGen Small', 'MTG-Jamendo', 'Suno v3', 'MusicGen Medium', 'MusicGen Large', 'AudioLDM 2 Large', 'AudioLDM 2 Music', 'Riffusion', 'Mustango', 'Stable Audio v1', 'Stable Audio v2', 'Suno v3.5')
- **description**: List of three strings containing the tags used to prompt the music generation model.
- **audio**: Contains the audio samples.

For more information or to cite our work please see [Benchmarking Music Generation Models and Metrics via Human Preference Studies](https://ieeexplore.ieee.org/abstract/document/10887745).

## License

The 6000 music tracks we generated using various models and services are distributed under a [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/) license.
The 500 MTG-Jamendo tracks we distribute here have their own [per-track license](https://mtg.github.io/mtg-jamendo-dataset/#license).
The description tags used to generate music come from MTG-Jamendo and are distributed under [CC BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/).

from datasets import load_dataset
dataset = load_dataset('disco-eth/AIME')

# AIME：AI音乐评估数据集

AIME数据集包含6000段由12种音乐生成模型生成的音频曲目，外加来自[MTG-Jamendo](https://mtg.github.io/mtg-jamendo-dataset/)的500段曲目。用于生成音乐的提示词（prompt）为MTG-Jamendo数据集中兼具代表性与多样性的标签组合。

AIME数据集包含两个子数据集：AIME音频数据集与[AIME调查数据集](https://huggingface.co/datasets/disco-eth/AIME-survey)。

该数据集包含以下字段：
- **id**：用于将调查数据与音频文件进行关联映射。
- **model**：字符串类型，存储用于生成该音频样本的模型名称，可选值包括'Udio'、'MusicGen Small'、'MTG-Jamendo'、'Suno v3'、'MusicGen Medium'、'MusicGen Large'、'AudioLDM 2 Large'、'AudioLDM 2 Music'、'Riffusion'、'Mustango'、'Stable Audio v1'、'Stable Audio v2'、'Suno v3.5'。
- **description**：包含三个字符串的列表，存储用于提示音乐生成模型的标签。
- **audio**：存储音频样本。

如需了解更多信息或引用本研究，请参阅论文《通过人类偏好研究对音乐生成模型及指标进行基准测试（Benchmarking Music Generation Models and Metrics via Human Preference Studies）》，链接为https://ieeexplore.ieee.org/abstract/document/10887745。

## 许可证
我们通过各类模型与服务生成的6000段音乐曲目采用[CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/)许可证进行分发。
我们在此分发的500段MTG-Jamendo曲目拥有各自的[逐曲目许可证](https://mtg.github.io/mtg-jamendo-dataset/#license)。
用于生成音乐的描述标签源自MTG-Jamendo，采用[CC BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/)许可证进行分发。