Dataset

VoxCeleb 1&2

Please refer to this website to download VoxCeleb 1&2 dataset.

FFSVC 2020

Note: Since FFSVC2020 and FFSVC2022 have concluded, please send an email to xiaoyi.qin@dukekunshan.edu.cn to request the FFSVC2020 dataset. The following FFSVC2020 supplementary set link is still active.

The FFSVC 2020 challenge dataset is part of the DMASH dataset. The whole FFSVC2020 challenge data includes the training/development/evaluation data and dev/eval trial files with keys. Please visit FFSVC2020 challenge paper and FFSVC2020 evaluation plan for more information anout FFSVC2020 dataset.

In this challenge, we release an additional supplementary set of FFSVC2020, in which the speaker information is the same as the FFSVC2020 training set. In contrast to the FFSVC2020 training set, the supplementary set of FFSVC2020 consists of the far-field tablet/telephone data, which is the same as the evaluation data for FFSVC2022. You can download the supplementary data through the following link.

The email address of the above data application should be the same as the account email of the challenge submission platform to facilitate confirmation as soon as possible.

Development and test set of FFSVC2022

A small trials file and wav files with accurate speaker information will be provided for participants as the FFSVC22 development set. The FFSVC2022 development data has the same data distribution as the FFSVC2022 evaluation data. However, the FFSVC2022 development set is only allowed to tune hyper-parameters and test the model performance. Any circumstances of using FFSVC2022 development set in the training is not allowed.

Unlike the FFSVC2020 challenge, FFSVC2022 includes more recording devices and all trial pairs are single-channel speech segments.

Note: If your region has restrictions on the current download links, please email to ffsvc.challenge@gmail.com. We can provide additional download links of Myairbridge and Baidu Pan.