Semi-Automatic Indoor Window Reconstruction

This repository provides the implementation for reconstruction algorithm proposed in the paper Semi-Automatic Indoor Window Reconstruction.

Folder Structure

• evaluation/: contains script for evaluating the reconstruction from three aspects: 1. geometric accuracy with Chamfer distance and Hausdorff distance(geometric_comparison.py), 2. grid-based daylight availability with advanced 2-phase method(advanced_2phase.py), and 3. visual comfort performance with glare analysis(glare_render.py). The I/O data for the evaluation is available at .

• The remaining folders contain the source code and CMake configuration files for the reconstruction program.

Usage

Build

Option 1

Use CMake to generate Makefiles, then compile with make (on Linux/macOS) or nmake (on Windows with Microsoft Visual Studio).

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

On Windows (use the x64 Native Tools Command Prompt for VS XXXX, not x86):

cd path-to-root-dir
mkdir Release
cd Release
cmake -G "NMake Makefiles" -DCMAKE_BUILD_TYPE=Release ..
nmake

Option 2 (IDE):

Use an IDE that supports CMakeLists.txt directly (e.g., CLion). Open the root CMakeLists.txt file, configure, and build the project.

Running and usage

The program relies on user interaction in the following order:

1. Run the code, you should see a window with point cloud and PolyFit output.
2. Press CTRL + left click on the walls that contain windows.
3. Press CTRL + D to run the Delaunay triangulation on the selected walls.
4. Press SHIFT + S to start drawing the window regions on each selected wall, when done with a wall, press SHIFT + right click to switch to the next wall until all the walls are complete.
5. After completing the window selection, find the win_boundary.obj, result.obj, and extruded{n}.obj files in the bin directory. These files contain the reconstructed window boundaries, the walls with windows, and the extruded windows, respectively.

Test data preparation

A test dataset is provided at /resources/data/example.bvg. To make your own test data, you should (1) Remove outlier points (optional), (2) Estimate normals, and (3) Extract planes with RANSAC and (4) Downsample the point cloud for quicker processing (optional). We used CloudCompare and Mapple to do this. When adding your own data, update the input path in ./code/main.cpp accordingly.

Solver

This program relies on the commercial solver Gurobi for its main optimization process. To use Gurobi, you must first install it and ensure that CMake can locate its headers and libraries. This can be achieved by setting the appropriate paths in the FindGUROBI.cmake file. Please note that a license is required to use Gurobi, although it is available free of charge for academic purposes.

Citation

If you find this code useful for your research, please consider citing the following paper:

@article{FOROUZANDEH2025114045,
title = {Semi-automated indoor geometry reconstruction for daylight simulation},
journal = {Building and Environment},
pages = {114045},
year = {2025},
issn = {0360-1323},
doi = {https://doi.org/10.1016/j.buildenv.2025.114045},
url = {https://www.sciencedirect.com/science/article/pii/S0360132325015112},
author = {Nima Forouzandeh and Jin Huang and Liangliang Nan and Eleonora Brembilla and Jantien Stoter},
}

The I/O data for the evaluation is available at .

License

This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 3 of the License or (at your option) any later version. The full text of the license can be found in the accompanying LICENSE file.

Should you have any questions, comments, or suggestions, please feel free to contact us at: N.Forouzandeh@tudelft.nl, jinhuang.nuaa@gmail.com.

Nima Forouzandeh, Jin Huang.

Copyright © 2025 3D geoinformation group, TU Delft, Nima Forouzandeh and Jin Huang. All rights reserved.