|
--- |
|
license: cc-by-sa-4.0 |
|
tags: |
|
- speaker extraction |
|
- speech separation |
|
- lip |
|
- face |
|
- multimodal |
|
- cocktail party problem |
|
--- |
|
|
|
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](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 |
|
``` |
|
|
