glomap

⚠️ [DEPRECATED] This project is no longer maintained and has been fully migrated to COLMAP, where GLOMAP functionality is exposed as the "global" mapper.

GLOMAP: Global Structure-from-Motion Revisited

Project page | Paper

About

GLOMAP is a general purpose global structure-from-motion pipeline for image-based reconstruction. GLOMAP requires a COLMAP database as input and outputs a COLMAP sparse reconstruction. As compared to COLMAP, this project provides a much more efficient and scalable reconstruction process, typically 1-2 orders of magnitude faster, with on-par or superior reconstruction quality.

If you use this project for your research, please cite

@inproceedings{pan2024glomap,
    author={Pan, Linfei and Barath, Daniel and Pollefeys, Marc and Sch\"{o}nberger, Johannes Lutz},
    title={{Global Structure-from-Motion Revisited}},
    booktitle={European Conference on Computer Vision (ECCV)},
    year={2024},
}

To use the seperate rotation averaging module, refer to this README.

Getting Started

To build GLOMAP, first install COLMAP dependencies and then build GLOMAP using the following commands:

mkdir build
cd build
cmake .. -GNinja
ninja && ninja install

Pre-compiled Windows binaries can be downloaded from the official release page.

After installation, one can run GLOMAP by (starting from a database)

glomap mapper --database_path DATABASE_PATH --output_path OUTPUT_PATH --image_path IMAGE_PATH

For more details on the command line interface, one can type glomap -h or glomap mapper -h for help.

We also provide a guide on improving the obtained reconstruction, which can be found here.

Note:

• GLOMAP depends on two external libraries - COLMAP and PoseLib. With the default setting, the library is built automatically by GLOMAP via FetchContent. However, if a self-installed version is preferred, one can also disable the FETCH_COLMAP and FETCH_POSELIB CMake options.
• To use FetchContent, the minimum required version of cmake is 3.28. If a self-installed version is used, cmake can be downgraded to 3.10.
• If your system does not provide a recent enough CMake version, you can install it as:
  bash复制

  wget https://github.com/Kitware/CMake/releases/download/v3.30.1/cmake-3.30.1.tar.gz
  tar xfvz cmake-3.30.1.tar.gz && cd cmake-3.30.1
  ./bootstrap && make -j$(nproc) && sudo make install

End-to-End Example

In this section, we will use datasets from this link as examples. Download the datasets and put them under data folder.

From database

If a COLMAP database already exists, GLOMAP can directly use it to perform mapping:

glomap mapper \
    --database_path ./data/gerrard-hall/database.db \
    --image_path    ./data/gerrard-hall/images \
    --output_path   ./output/gerrard-hall/sparse

From images

To obtain a reconstruction from images, the database needs to be established first. Here, we utilize the functions from COLMAP:

colmap feature_extractor \
    --image_path    ./data/south-building/images \
    --database_path ./data/south-building/database.db
colmap exhaustive_matcher \
    --database_path ./data/south-building/database.db
glomap mapper \
    --database_path ./data/south-building/database.db \
    --image_path    ./data/south-building/images \
    --output_path   ./output/south-building/sparse

Visualize and use the results

The results are written out in the COLMAP sparse reconstruction format. Please refer to COLMAP for more details.

The reconstruction can be visualized using the COLMAP GUI, for example:

colmap gui --import_path ./output/south-building/sparse/0

Alternatives like rerun.io also enable visualization of COLMAP and GLOMAP outputs.

If you want to inspect the reconstruction programmatically, you can use pycolmap in Python or link against COLMAP's C++ library interface.

Notes

• For larger scale datasets, it is recommended to use sequential_matcher or vocab_tree_matcher from COLMAP.

colmap sequential_matcher --database_path DATABASE_PATH
colmap vocab_tree_matcher --database_path DATABASE_PATH --VocabTreeMatching.vocab_tree_path VOCAB_TREE_PATH

• Alternatively, one can use hloc for image retrieval and matching with learning-based descriptors.

Acknowledgement

We are highly inspired by COLMAP, PoseLib, Theia. Please consider also citing them, if using GLOMAP in your work.

Support

Please, use GitHub Discussions at https://github.com/colmap/glomap/discussions for questions and the GitHub issue tracker at https://github.com/colmap/glomap for bug reports, feature requests/additions, etc.

Contribution

Contributions (bug reports, bug fixes, improvements, etc.) are very welcome and should be submitted in the form of new issues and/or pull requests on GitHub.

License

…