MAC++: Going Further with Maximal Cliques for 3D Registration

Source code of MAC++: Going Further with Maximal Cliques for 3D Registration.

Introduction

Maximal cliques (MAC) represent a novel state-of-the-art approach for 3D registration from correspondences, however, it still suffers from extremely severe outliers. In this paper, we introduce a robust learning-free estimator called MAC++, exploring maximal cliques for 3D registration from the following two perspectives:

1. A novel hypothesis generation method utilizing putative seeds through voting to guide the construction of maximal clique pools, effectively preserving more potential correct hypotheses.
2. A progressive hypothesis evaluation method that continuously reduces the solution space in a ''global-clusters-cluster-individual'' manner rather than traditional one-shot techniques, greatly alleviating the issue of missing good hypotheses.

Experiments conducted on U3M, 3DMatch/3DLoMatch, and KITTI-LC datasets show the new state-of-the-art performance of MAC++. MAC++ demonstrates the capability to handle extremely low inlier ratio data where MAC fails (e.g., showing 27.1%/30.6% registration recall improvements on 3DMatch/3DLoMatch with <1% inliers).

Build

MAC depends on PCL (version 1.12.1 is recommended) and igraph(=0.10.6). Please install these libraries first. To build MAC, you need CMake and, of course, a compiler that supports >= C++11.

• Option 1 (purely on the command line): Use CMake to generate Makefiles and then make.
  • You can simply run

    $ cd path-to-root-dir-of-MAC
    $ mkdir Release
    $ cd Release
    $ cmake -DCMAKE_BUILD_TYPE=Release ..
    $ make
    

• Option 2: Use any IDE that can directly handle CMakeLists files to open the CMakeLists.txt in the root directory of MAC. Then you should have obtained a usable project and just build it. I recommend using CLion.
• NOTICE: Please compile in RELEASE mode!

Usage: ./Boot --args

• --help list all usages.
• --demo run the demo.

Required args:

• --output_path output path for saving results.
• --input_path input data path.
• --dataset_name dataset name.
• --descriptor descriptor name.
• --start_index(begin from 0) run from given index.

Optional args:

• --no_logs forbid generation of log files.

Datasets

All tested datasets can be found at this link, password：6666.

Citation

If you find this code useful for your work or use it in your project, please consider citing:

Reference

• Zhang et al., 3D Registration with Maximal Cliques (CVPR 2023)