RT-Pose
收藏资源简介:
RT-Pose数据集由国立成功大学和华盛顿大学联合创建,是一个集成4D雷达张量、LiDAR点云和RGB图像的多模态数据集,专门用于人体姿态估计和定位。该数据集包含72000帧数据,分布在240个序列中,涵盖六种不同复杂度的动作。数据集的创建过程结合了RGB图像和LiDAR点云进行精确的3D人体骨骼标注。RT-Pose数据集主要应用于增强/虚拟现实、人机交互和医疗保健等领域,旨在解决复杂场景下的人体姿态估计问题。
RT-Pose, a multimodal dataset jointly developed by National Cheng Kung University and the University of Washington, integrates 4D radar tensors, LiDAR point clouds and RGB images, and is specifically designed for human pose estimation and localization. It comprises 72,000 frames of data across 240 sequences, covering six action categories with varying levels of complexity. The construction of the RT-Pose dataset leverages RGB images and LiDAR point clouds to generate accurate 3D human skeleton annotations. Primarily applied in domains including augmented/virtual reality, human-computer interaction and healthcare, the RT-Pose dataset aims to address the challenges of human pose estimation in complex scenarios.

- 1RT-Pose: A 4D Radar Tensor-based 3D Human Pose Estimation and Localization Benchmark国立成功大学,华盛顿大学 · 2024年



