SynHand
收藏资源简介:
SynHand数据集由香港中文大学的研究团队创建,旨在为全身姿态估计中的手部姿态评估提供一个全面的基准。该数据集包含462,800条数据,涵盖了多样化的手部姿态,特别是在近距离人体拍摄中的手部姿态。数据集中的手部姿态被精确标注为全身SMPL-X标签的一部分。通过使用SynHand数据集,研究人员能够更好地评估模型在手部姿态估计中的表现,尤其是在复杂场景下的泛化能力。该数据集的应用领域主要集中在动画、游戏和时尚产业中的人体姿态捕捉与重建。
SynHand Dataset was developed by the research team at The Chinese University of Hong Kong, aiming to provide a comprehensive benchmark for hand pose evaluation in full-body pose estimation. This dataset contains 462,800 samples, covering a wide range of hand poses, particularly those captured in close-up human imagery. The hand poses in the dataset are precisely annotated as part of the full-body SMPL-X annotation suite. By leveraging the SynHand Dataset, researchers can more effectively evaluate the performance of models for hand pose estimation, especially their generalization capabilities in complex scenarios. The primary application scenarios of this dataset focus on human pose capture and reconstruction in the animation, gaming, and fashion industries.

- 1SMPLest-X: Ultimate Scaling for Expressive Human Pose and Shape Estimation香港中文大学 · 2025年



