PedX
收藏资源简介:
PedX数据集是由密歇根大学开发的一个大规模多模态行人数据集,专注于复杂城市交叉口的行人三维姿态估计。该数据集包含超过5000对高分辨率立体图像和LiDAR数据,以及2D和3D行人标签。数据集的创建过程涉及自动3D标签算法,该算法利用多模态数据和形状及时间先验进行3D标注。PedX数据集的应用领域主要集中在自动驾驶车辆在复杂城市环境中对行人姿态、运动、行为和意图的理解,旨在提高自动驾驶车辆的安全性和效率。
The PedX dataset is a large-scale multimodal pedestrian dataset developed by the University of Michigan, focusing on 3D human pose estimation at complex urban intersections. It contains over 5000 pairs of high-resolution stereo images and LiDAR data, as well as 2D and 3D pedestrian annotations. The development of this dataset involves automated 3D annotation algorithms, which leverage multimodal data, shape priors and temporal priors to perform 3D labeling. The primary application scenarios of the PedX dataset focus on autonomous vehicles' understanding of pedestrian poses, motions, behaviors and intentions in complex urban environments, aiming to improve the safety and efficiency of autonomous driving systems.

- 1PedX: Benchmark Dataset for Metric 3D Pose Estimation of Pedestrians in Complex Urban Intersections密歇根大学 · 2018年



