The project requires: OpenGL 4.2 or later; 

We provide GLFW in "includes" and "lib" dirs. 
Create "build" dir and use cmake to build the project. Note that the paths of shaders should be correct.

Specify the paths of position and normal of the point cloud. 

We provide an example point cloud in the "example" dir. We pre-compute the normals of the points. 

The program generates supervoxel segmentation results in txt file, including labels and pseudo-color result. 
You can specify the "save_type" of the txt result. Please use software "CloudCompare" to see the pseudo-color result. 


You can download the point cloud datasets from following websites:
Okaland: http://www.cs.cmu.edu/~vmr/datasets/oakland_3d/cvpr09/doc/
Semantic3D: http://www.semantic3d.net/
NYUV2: https://cs.nyu.edu/~silberman/datasets/nyu_depth_v2.html

The codes in this repository are under the GNU AFFERO GENERAL PUBLIC LICENSE as specified by the LICENSE file.

@article{dong2022gpu,
  title={GPU-based Supervoxel Segmentation for 3D Point Clouds},
  author={Dong, Xiao and Xiao, Yanyang and Chen, Zhonggui and Yao, Junfeng and Guo, Xiaohu},
  journal={Computer Aided Geometric Design},
  pages={102080},
  year={2022},
  publisher={Elsevier}
}