PrimFit is the implementation of reconstruction method described in paper “Structure–Aware Surface Reconstruction via Primitive Assembly”.
- 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)
- Easy3D (2.5.2 or later), 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.
After successfully configuring all the required dependent libraries, you can proceed to run the example. To do so, simply modify the string variables input_path
(in .seg format) and output_path
(in .obj format) in the cli/primfit_cli.cpp
file.
Once the paths are updated, execute the target primfit_cli.exe
to test our algorithm.
The .seg format is our custom format.
The folder data/sub_abc_seg
provides 40 .seg files for testing; data/sub_abc_ours
contains our test results; and data/sub_abc_xyz
holds the original test point cloud files.
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].