emg2pose
收藏资源简介:
emg2pose数据集是由Meta的Reality Labs创建的,旨在促进从表面肌电图(sEMG)进行手势估计模型的开发。该数据集包含193名用户的高质量手势标签和手腕sEMG记录,总计370小时,涵盖29种多样化的手势类别。数据集通过26摄像头的动作捕捉系统和sEMG-RD腕带同时记录,提供了高分辨率的sEMG数据和精确的手势标签。创建过程中,参与者按照标准化协议进行数据收集,确保了数据的多样性和广泛性。该数据集主要应用于开发通用sEMG-to-pose模型,旨在解决手势估计在虚拟和增强现实中的应用问题。
The emg2pose dataset was developed by Meta's Reality Labs to facilitate the development of gesture estimation models based on surface electromyography (sEMG). This dataset includes high-quality gesture labels and wrist sEMG recordings from 193 unique users, totaling 370 hours of data, and covers 29 distinct gesture categories. Collected simultaneously using a 26-camera motion capture system and sEMG-RD wristbands, the dataset provides high-resolution sEMG data and precise gesture annotations. During the data collection process, all participants followed a standardized protocol, ensuring the dataset's diversity and broad representativeness. This dataset is primarily intended for developing general-purpose sEMG-to-pose models, with the aim of addressing practical challenges of gesture estimation in virtual reality (VR) and augmented reality (AR).

- 1emg2pose: A Large and Diverse Benchmark for Surface Electromyographic Hand Pose EstimationMeta的Reality Labs · 2024年



