设为首页  |  加入收藏
当前位置: 网站首页>>科学研究
天津理工大学手语识别和翻译数据集

天津理工大学手语识别和翻译数据集

TJUT Sign Language Recognition and Translation Dataset

 

该数据集为基于国家通用手语标准的自然手语视频集。以手语语言学专家录制的示范视频为指导,先后邀请到共计120名社会聋人、聋人大学生及听人手语爱好者参与多样本手语视频数据的采集工作。被采集者采用手机摄像头纵向拍摄,拍摄时姿势为站立,双手的开始和结束位置为身体两侧。拍摄场景尽量光照条件良好,不限室内与室外,但不应出现逆光等情况,且要求服装颜色与背景颜色不能相同或相近,背景不能过于杂乱,背景中的远景可以出现行人、旁观者等其他没有在打手语的人。

 

This data set is a natural sign language video set based on Chinese national common sign language standard. A total of 120 signers, including the deaf from society, deaf college students and sign language enthusiasts, were invited to participate in the collection of multi sample sign language video data. They were guided by demonstration videos recorded by sign language linguists. The signers used a mobile phone camera to shoot longitudinally. The position of shooting was standing, and the start and end positions of their hands were on both sides of their bodies. The lighting conditions of the shooting scene should be good as far as possible, not limited to indoor or outdoor, but there should be no backlight. Their clothing color and background color should not be the same or similar, the background should not be too messy. People are allowed to appear in the vision of the background, but only one person is allowed to sign in the scene.

 

如希望索取该数据集,请联系tslrtd@126.com

If you want to obtain this data set, please contact tslrtd@126.com

 科研专业 
 科研项目 
Baidu
map