Semantic Parametric Body Shape Estimation from Noisy RGB-D Sequences

Robotics and Autonomous Systems

 



Publication

Alexandru-Eugen Ichim and Federico Tombari

Semantic Parametric Body Shape Estimation from Noisy RGB-D Sequences

In Review for Robotics and Autonomous Systems, Special Issue on 3D Robot Perception with the Point Cloud Library.



Click on the image to download the full version of the paper:




Video

Click on the image to download the video:




Source Code

We provide source code set up to experiment with the pipeline described in the paper, as well as offer the code for data acquisition. The project requires PCL 1.7, and are build using CMake. Please check the Point Cloud Library Tutorials on how to set up your system and become familiar with the basic building concepts: http://pointclouds.org/documentation/tutorials/

GitHub Repo



Datasets

Sample Asus Xtion PRO Datasets

Each dataset is a sequence of a person moving in front of the camera. Each frame is composed of: a depth image, RGB image, binary user map, as well as a text file containing the detected NiTE features.


COMING SOON

COMING SOON

COMING SOON

COMING SOON



Other Datasets



The IAS-Lab at the University of Padova, Italy provides multiple RGB-D body tracking datasets:
Kinect Tracking Precision (KTP) Dataset
IAS-Lab People Tracking Dataset
BIWI RGBD-ID Dataset
IAS-Lab RGBD-ID Dataset

For more datasets, please visit the pcl dataset webpage, where some of the most popular 3D datasets are indexed.


Authors:

For further questions or suggestions, please do not hesitate to contact the authors via e-mail.