PrimFit is the implementation of reconstruction method described in paper “Structure–Aware Surface Reconstruction via Primitive Assembly”.

We have released the Windows version of our method.
Feel free to test it and share your feedback.

⚠️ Note: This release is for evaluation only.

• Windows 11
• CLion2024.1.2 + Visual Stdio 2022
• Intel(R) Core i9-13900K

The dependent libraries of our code includes:

• Eigen3 (3.4.0 or later)
• CLI11 (2.4.0 or later), command line.
• nlohmann_json (3.11.3 or later), read json file.
• Easy3D (2.5.2), Please visit the Prof.Nan's repository to obtain it and follow the instructions for installation.
• libigl, A special version, and automatically obtained through the FetchContent function in CMake. Ensure the stability of your VPN if you are in mainland China.

For Eigen3, CLI11 and nlohmann_json, we recommend using vcpkg to install them.

# Eigen3
vcpkg install Eigen3:x64-windwos
# cli11
vcpkg install cli11:x64-windows
# nlohmann_json
vcpkg install nlohmann_json:x64-windows

For Easy3D, you need to set the ${Easy3D_DIR} environment variable to the directory that contains the Easy3DConfig.cmake file.

For libigl, if you're located in mainland China and your VPN connection is unstable, you can download the external.zip file via my Baidu Netdisk link. Please extract external.zip into the project root directory, resulting in the following structure:

PrimFit_project_root/
├── external/                
├── CMakeLists.txt
└── ...

Building our code in CLion:

# File -> Setting -> Build, Execution, Deployment -> CMake -> CMake Option :
-DCMAKE_TOOLCHAIN_FILE=${YOUR_VCPKG_INSTALL_PATH}/scripts/buildsystems/vcpkg.cmake

Making sure that your following settings are correct:

• Toolchains : Visual Stdio
• Architecture : amd64
• Build Type : release

primfit_cli.exe -c config.json

primfit_cli.exe takes a .seg file as input and outputs a .obj file. The .seg format is a custom data format that contains segmented primitives and point cloud data. You can modify the config.json file to change the input and output file paths as well as algorithm parameters (configured based on the details in the paper, typically no adjustments are needed).

Ransac2Seg.exe -c seg_config.json

Ransac2Seg.exe takes a .ply point cloud file (with normals) as input and generates a .seg file, which serves as the input for primfit.exe.

You can modify the seg_config.json file to change the input and output file paths as well as Ransac parameters. The specific parameter tuning details can be referenced from the paper Efficient RANSAC.

vg2seg.exe -i input.vg -o out.seg

vg2Seg.exe converts a .vg file to our custom .seg format, which serves as the input for primfit.exe. The .vg file can be obtained from PolyFit or KSR.

The folder data/sub_abc_seg provides 40 .seg files for testing; data/sub_abc_ours contains our test results; and data/sub_abc_ply holds the original test point cloud files.

We also provide executable files for Windows. Feel free to download and test them.

If you make use of our work, please cite our paper:

@inproceedings{Jiang2023primfit,
  title={Structure–Aware Surface Reconstruction via Primitive Assembly},
  author={Jingen Jiang, Mingyang Zhao, Shiqing Xin, Yanchao Yang, Hanxiao Wang, Xiaohong Jia, Dong-Ming Yan},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  year={2023}
}

This work was partially funded by the National Key Research and Development Program (2021YFB1715900), the CAS Project for Young Scientists in Basic Research (YSBR-034), the National Natural Science Foundation of China (62172415, 62272277, 12022117), and the HKU-100 Research Award.

Our code is inspired the works of BSH, PolyFit and KSR. We would like to thank Dr. Xingyi Du and Prof. Liangliang Nan for their excellent code.

Furthermore, we are grateful to Jiahui Lv from Shenzhen University for his valuable advice in this work.

If any problem, please contact me via [email protected].