SignAvatars
收藏资源简介:
SignAvatars数据集由帝国理工学院和腾讯AI实验室合作创建,是首个大规模的3D手语运动数据集。该数据集包含70,000个视频,总计8.34百万帧,覆盖了孤立和连续的手语动作,并提供了多种提示类型,如HamNoSys、口语和词汇。数据集通过创新的自动标注流程实现了3D整体标注,包括身体、手和面部的网格和生物力学有效的姿势,以及2D和3D关键点。SignAvatars数据集旨在弥合听障人士的沟通差距,支持多种任务,如3D手语识别和从文本脚本、单个词汇和HamNoSys符号生成3D手语。
The SignAvatars dataset, co-created by Imperial College London and Tencent AI Lab, is the first large-scale 3D sign language motion dataset. It contains 70,000 videos totaling 8.34 million frames, covering both isolated and continuous sign language gestures, and provides multiple annotation types such as HamNoSys, spoken language, and vocabulary. The dataset realizes holistic 3D annotations via an innovative automatic annotation pipeline, including mesh models and biomechanically valid poses for the body, hands, and face, as well as 2D and 3D keypoints. The SignAvatars dataset aims to bridge the communication gap for deaf and hard-of-hearing individuals, supporting a range of tasks including 3D sign language recognition and 3D sign language generation from text scripts, single vocabulary items, and HamNoSys symbols.




