Abstract
We propose an approach of human joints extraction using part segmentation. Six human segmentations obtained by PointNet++ are further utilized to extract the 3D joints. The average error of 14 joints is < 4.2 cm.
© 2021 The Author(s)
PDF ArticleMore Like This
Tianxu Xu, Dong An, Zhonghan Wang, Sicheng Jiang, Chengnuo Meng, Qiang Wang, Zhongqi Pan, and Yang Yue
FW1F.6 Frontiers in Optics (FiO) 2020
J. F. Flórez, C.R. Contreras, and J. E. Meneses
LTh4A.21 Latin America Optics and Photonics Conference (LAOP) 2014
Prasant Potuluri and David J. Brady
OFA5 Optics in Computing (IP) 2003