Skip to content
No description, website, or topics provided.
Branch: master
Clone or download
Latest commit 480a4c9 Dec 20, 2016
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
Tools first commit Dec 20, 2016
imgs first commit Dec 20, 2016
README modify readme Dec 20, 2016
README.md Update README.md Dec 21, 2016
README~ first commit Dec 20, 2016
demo.m first commit Dec 20, 2016
vis_2d.m first commit Dec 20, 2016
vis_3d.m first commit Dec 20, 2016

README.md

3D Human Pose Estimation = 2D Pose Estimation + Matching

Contact

Ching-Hang Chen

email: flyawaychase@gmail.com

This matlab code demo the 3D human pose estimation from single RGB image with given 2D pose landmarks. The 2D pose can be derived by state-of-the-art CNN such as CPM (https://github.com/shihenw/convolutional-pose-machines-release) or Hourglass

Demo Videos

Before start, check out the demo videos!

https://www.youtube.com/watch?v=T_gEx7eQqEg

https://www.youtube.com/watch?v=uGBKZ7QQlDI

Getting Started

Download the 3D pose library from the following link: https://drive.google.com/file/d/0BxZCS0CAHYZpRjA2S1FsWDNhcEU/view?usp=sharing

Usage:

  1. Open Matlab, set and include current directory to "Release", put "3D_library.mat" in "Release" folder
  2. Run demo.m (few examples are provided with associated 2D pose by CPM)
  3. Implement 2D pose estimation (ex. CPM or Hourglass) to run your own examples
  4. Change the input image in demo.m (line 26) and load your own 2D pose estimation (line 32)

Reference

3D Human Pose Estimation = 2D Pose Estimation + Matching http://arxiv.org/abs/1612.06524

You can’t perform that action at this time.