Learning to Estimate 3D Human Pose from Point Cloud

TL;DR

This paper introduces a novel deep learning approach that directly estimates 3D human pose from point cloud data, outperforming existing CNN-based methods on public datasets.

Contribution

The paper presents a new deep network that models 3D human surface structures directly from point clouds for pose estimation, diverging from traditional image-based methods.

Findings

Achieves higher accuracy than previous methods on ITOP and EVAL datasets.

Effectively models complex human surface structures from point clouds.

Demonstrates the effectiveness of point cloud-based 3D pose estimation.

Abstract

3D pose estimation is a challenging problem in computer vision. Most of the existing neural-network-based approaches address color or depth images through convolution networks (CNNs). In this paper, we study the task of 3D human pose estimation from depth images. Different from the existing CNN-based human pose estimation method, we propose a deep human pose network for 3D pose estimation by taking the point cloud data as input data to model the surface of complex human structures. We first cast the 3D human pose estimation from 2D depth images to 3D point clouds and directly predict the 3D joint position. Our experiments on two public datasets show that our approach achieves higher accuracy than previous state-of-art methods. The reported results on both ITOP and EVAL datasets demonstrate the effectiveness of our method on the targeted tasks.