MIRACL-VC1
收藏arXiv2025-09-30 收录
下载链接：
https://abenhamadou.github.io/miraclvc1/index.html
下载链接
链接失效反馈官方服务：
资源简介：
该数据集（MIRACL-VC1）是业界公认的用于评估视觉语音识别模型的标准数据集，尤其在口语密码的识别场景中具有重要作用。该数据集被用来展示所提出模型的效能，在对其进行测试时，该模型实现了96.1%的准确率。此外，该数据集的任务是基于视觉语音识别，对人员-单词组合进行二分类。
This dataset (MIRACL-VC1) is a widely recognized standard dataset for evaluating visual speech recognition models across academic and industrial circles, and holds particular importance in the scenario of spoken password recognition. It is utilized to showcase the effectiveness of proposed models, with the tested model achieving an accuracy of 96.1% when evaluated using this dataset. Furthermore, the task of this dataset is binary classification of person-word pairs based on visual speech recognition.
提供机构：
Research community
