ssssjiang / iRotAvg

iRotAvg (incremental rotation averaging) incrementally solves rotation averaging. iRotAvg is the optimisation core of L-infinity SLAM.

iRotAvg

Author: Álvaro Parra

iRotAvg (incremental rotation averaging) incrementally solves rotation averaging. iRotAvg is the optimisation core of L-infinity SLAM presented in [Á. Parra, T.-J. Chin, A. Eriksson, I. Reid: Visual SLAM: Why bundle adjust?, ICRA 2019].

This work was supported by Maptek and the ARC Linkage Project LP140100946.

     

Related Publication:

Á. Parra, T.-J. Chin, A. Eriksson, I. Reid: Visual SLAM: Why bundle adjust?, ICRA 2019

License

iRotAvg is released under a GPLv3 license.

For a closed-source version of iRotAvg (e.g., for commercial purposes), please contact the author.

For an academic use of iRotAvg, please cite Á. Parra, T.-J. Chin, A. Eriksson, I. Reid: Why bundle adjust?, ICRA 2019.

Dependencies

• Eigen
• SuiteSparse
• opencv
• Boost (Filesystem)

In Mac:

• brew install eigen
• brew install suite-sparse
• brew install opencv
• brew install boost

In Ubuntu:

• sudo apt install libeigen3-dev
• sudo apt-get install libsuitesparse-dev
• sudo apt-get install libboost-all-dev
• For OpenCV wise, https://www.pyimagesearch.com/2018/05/28/ubuntu-18-04-how-to-install-opencv/ provides a good guide.

Compilation

• mkdir build
• cd build
• cmake ..
• make

(binary is compiled inside src)

Execution

General usage follows ORB-SLAM input:

./irotavg orb_vocab.txt config.yaml sequence_path

where

• orb_vocab.txt Orb vocabulary as in ORB-SLAM.
• config.yaml Configuration file as in ORB-SLAM.
• sequence_path Directory containing the frame sequence.

To see more usage options simply execute:

./irotavg

Testing with the KITTI dataset

To test iRotAvg with the KITTI dataset:

1. Download the grayscale sequences from here.
2. Download a configuration file from the ORB-SLAM monocular exampes. E.g.: KITTI00-02.yaml for sequences 0 to 2.
3. Download and uncompress the orb vocabulary file from here.
4. Execute irotavg passing the corresponding paths: ./irotavg path/to/orb_vocab.txt path/to/config.yaml path/to/sequence/

The output is saved to:

• rotavg_poses_ids.txt Contains the ids (starting in 1) of frames in the view-graph (selected frames).
• rotavg_poses.txt Contains the computed absolute poses.

Watch a demonstration in the KITTI dataset