Awesome

ArXiv Preprint (arXiv 2303.10882)

Accepted in CVPR 2023

Abstract

Refer to: fishmarch repo

The only difference of this repo w.r.t fishmarch repo is using OR-Tools instead of Gurobi to achieve the same result of map compression.

Installation

ORB-SLAM2

The method is tested with ORB-SLAM2. Please follow their instruction to install prerequisites.

Gurobi

The mix-integer programming is solved using GUROBI.

OR-Tools

Google Optimization Tools (a.k.a., OR-Tools) is an open-source, fast and portable software suite for solving combinatorial optimization problems.

If you want to run this repo, please first install OR-Tools on your machine.

Running

The method now is testing with RGBD and stereo mode of ORB-SLAM2, and tested in ICL-NUIM, EuRoC, and KITTI datasets.

Take EuRoC data for example.

1. Build and save an original map

Examples/Stereo/stereo_euroc_savemap Vocabulary/ORBvoc.txt Examples/Stereo/EuRoC.yaml PATH_TO_SEQUENCE/mav0/cam0/data PATH_TO_SEQUENCE/mav0/cam1/data Examples/Stereo/EuRoC_TimeStamps/SEQUENCE.txt PATH_TO_MAP 

2. Compress the original map

Examples/Tools/slam_compress_map Vocabulary/ORBvoc.txt Examples/Stereo/EuRoC.yaml PATH_TO_MAP PATH_TO_COMPRESSED_MAP COMPRESS_METHPD

COMPRESS_METHOD :

• 1: Compress based only on 2D grids
• 2: Compress based on 2D and 3D grids

3. Localize in the compressed map

Examples/Stereo/stereo_euroc_localize Vocabulary/ORBvoc.txt Examples/Stereo/EuRoC.yaml PATH_TO_SEQUENCE/mav0/cam0/data PATH_TO_SEQUENCE/mav0/cam1/data Examples/Stereo/EuRoC_TimeStamps/SEQUENCE.txt PATH_TO_COMPRESSED_MAP

Acknowledgements

• ORB-SLAM2

• Active-ORB-SLAM2

Citation

@InProceedings{Zhang_2023_CVPR,
    author    = {Zhang, Xiaoyu and Liu, Yun-Hui},
    title     = {Efficient Map Sparsification Based on 2D and 3D Discretized Grids},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2023},
    pages     = {12470-12478}
}