alibabasglab/VoxCeleb2-mix

A modified version of the VoxCeleb2 Dataset. Original data can be downloaded here. This dataset is used for Audio-visual speaker extraction conditioned on face recordings in the reentry paper, which the code can be found here (ClearVoice repo) or here (Paper repo). Usage cat orig* > orig.tar tar -xvf orig.tar cat audio_clean* > audio_clean.tar tar -xvf audio_clean.tar