VoxCeleb2-mix
收藏魔搭社区2026-01-02 更新2025-01-11 收录
下载链接:
https://modelscope.cn/datasets/alibabasglab/VoxCeleb2-mix
下载链接
链接失效反馈官方服务:
资源简介:
A modified version of the VoxCeleb2 Dataset. Original data can be downloaded [here](https://www.robots.ox.ac.uk/~vgg/data/voxceleb/vox2.html).
This dataset is used for Audio-visual speaker extraction conditioned on face recordings in [the reentry paper](https://ieeexplore.ieee.org/document/9721129), which the code can be found [here (ClearVoice repo)](https://github.com/modelscope/ClearVoice) or [here (Paper repo)](https://github.com/zexupan/reentry).
# Usage
``` sh
cat orig* > orig.tar
tar -xvf orig.tar
cat audio_clean* > audio_clean.tar
tar -xvf audio_clean.tar
```
本数据集为VoxCeleb2数据集(VoxCeleb2 Dataset)的修改版本。原始数据可通过[此处](https://www.robots.ox.ac.uk/~vgg/data/voxceleb/vox2.html)下载。
本数据集被用于[《Reentry》论文](https://ieeexplore.ieee.org/document/9721129)中提出的、以人脸录音为条件的音视觉说话人提取(Audio-visual speaker extraction)任务,相关代码可通过[此处(ClearVoice仓库)](https://github.com/modelscope/ClearVoice)或[此处(论文配套仓库)](https://github.com/zexupan/reentry)获取。
# 使用方法
sh
cat orig* > orig.tar
tar -xvf orig.tar
cat audio_clean* > audio_clean.tar
tar -xvf audio_clean.tar
提供机构:
maas
创建时间:
2025-01-06



