hustep-lab/VoxVietnam-Dataset

VoxVietnam是一个大规模的多语种数据集，用于越南语说话人识别。它包含三个子集：train（官方训练集，包含1256位说话人的161457个样本）、train_small（从train子集采样得到，以匹配Vietnam-Celeb的大小，包含879位说话人的83000个样本）和test（测试集）。此外，还有VoxVietnam-E和VoxVietnam-H由志愿者在没有视觉信息的情况下标注，而VoxVietnam-O是一个由我们团队通过视听验证的独立测试集。数据集的最新更新包括VoxVietnam-O的发布，鼓励研究者使用该测试集进行评估。

VoxVietnam is a large-scale multi-genre dataset for Vietnamese speaker recognition. It consists of three subsets: train (official training set with 1,256 speakers and 161,457 samples), train_small (sampled from the train subset to match the size of Vietnam-Celeb, with 879 speakers and 83,000 samples), and test (test set). Additionally, there are VoxVietnam-E and VoxVietnam-H labeled by volunteers without visual information, and VoxVietnam-O, an independent test set verified by our team through audio-visual validation. The latest update to the dataset includes the release of VoxVietnam-O, which is encouraged for researchers to use for evaluation.