RadProPoser 数据集
收藏资源简介:
RadProPoser 数据集是一个基于雷达数据的人体姿态估计数据集,由德国埃尔朗根-纽伦堡弗莱德里希-亚历山大大学的研究团队创建。该数据集包含光学动作捕捉的真实数据,用于训练和评估人体姿态估计模型。数据集包含26个三维关节位置信息,并能够量化每个关节的不确定性。数据集的创建过程涉及对原始雷达数据进行处理,并通过变分推理方法进行关键点回归。RadProPoser 数据集旨在解决基于雷达的人体姿态估计问题,为可解释和可靠的人体运动分析提供基础。
The RadProPoser dataset is a radar-based human pose estimation dataset created by the research team from Friedrich-Alexander University Erlangen-Nuremberg, Germany. It provides ground-truth optical motion capture data for training and evaluating human pose estimation models. The dataset encompasses 26 sets of 3D joint position information, and enables the quantification of uncertainty for each joint. The development of the RadProPoser dataset involves processing raw radar data and performing keypoint regression via variational inference methods. This dataset aims to address radar-based human pose estimation challenges, and serves as a foundation for explainable and reliable human motion analysis.

- 1RadProPoser: A Framework for Human Pose Estimation with Uncertainty Quantification from Raw Radar Data德国埃尔朗根-纽伦堡弗莱德里希-亚历山大大学(Friedrich-Alexander-Universität Erlangen-Nürnberg) · 2025年



