Skip to content

Latest commit

 

History

History
executable file
·
27 lines (19 loc) · 1.43 KB

README.md

File metadata and controls

executable file
·
27 lines (19 loc) · 1.43 KB

LC-CRF-SLAM

Accurate Dynamic SLAM using CRF-based Long-term Consistency

This is the source code of the paper: Accurate Dynamic SLAM using CRF-based Long-term Consistency
Authors: Zheng-Jun Du, Shi-Sheng Huang, Tai-Jiang Mu, Qunhe Zhao, Ralph R. Martin and Kun Xu

Our implementation is based on the framework of ORBSLAM2, any question about the code please contact me via my email: duzj19@mails.tsinghua.edu.cn.

Build

We compile the project on Ubuntu 16.04 LTS, and the compiling is similar to ORBSLAM2, please refer to https://github.com/raulmur/ORB_SLAM2#3-building-orb-slam2-library-and-examples for more detail. All dependency libraries are included in the Thirdparty directory.

Dataset

We test our algorithm on two dynamic dataset: TUM RGB-D dynamic dataset and BONN RGB-D dynamic dataset.

Usage

./rgbd_tum path_to_sequence path_to_association path_to_settings

more detail see the file: rgbd_tum.cc

Evaluate

The tools of ate/rpe evaluation locate in /Examples/RGB-D
usage: (python 2.7)

python evaluate_ate/evaluate_rpe.py groundtruth.txt trajectory.txt --verbose

Notes

If the frame viewers are frozen, close it, it will be reactived.