How2Sign
收藏资源简介:
How2Sign数据集是由卡内基梅隆大学等机构联合创建的大型多模态美国手语(ASL)数据集,包含超过80小时的连续手语视频及相应的语音、英文文本和深度数据。该数据集通过与专业ASL手语者合作,精心录制并标注,涵盖了2500多个教学视频,涉及广泛的主题类别。数据集的创建旨在推动手语识别、翻译和生成等领域的研究,特别是在缺乏大规模标注数据集的情况下。此外,数据集还包括详细的3D姿态估计,为计算机视觉系统理解和分析手语提供了重要资源。
The How2Sign dataset is a large-scale multimodal American Sign Language (ASL) dataset jointly created by Carnegie Mellon University and other institutions. It contains over 80 hours of continuous sign language videos along with corresponding speech, English text and depth data. The dataset was meticulously recorded and annotated in collaboration with professional ASL signers, covering more than 2,500 instructional videos spanning a wide range of topic categories. The dataset was developed to advance research in fields such as sign language recognition, translation and generation, particularly in scenarios where large-scale annotated datasets are scarce. Furthermore, the dataset includes detailed 3D pose estimation, providing an important resource for computer vision systems to understand and analyze sign language.

- 1How2Sign: A Large-scale Multimodal Dataset for Continuous American Sign Language卡内基梅隆大学 · 2021年



