| 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 (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 | |
| ``` | |