The paper has been released and can be found at Dual-Level Precision Edges Guided Multi-View Stereo with Accurate Planarization.

If you find this project useful for your research, please cite:

@inproceedings{chen2025dual,
  title={Dual-Level Precision Edges Guided Multi-View Stereo with Accurate Planarization},
  author={Chen, Kehua and Yuan, Zhenlong and Mao, Tianlu and Wang, Zhaoqi},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={39},
  number={2},
  pages={2105--2113},
  year={2025}
}

The code has been tested on Ubuntu 20.04 with Nvidia RTX 3090.

• Cuda >= 10.2
• OpenCV >= 3.3.0
• Boost >= 1.62.0
• cmake >= 2.8

Besides make sure that your GPU Compute Capability matches the CMakeList.txt!!! Otherwise you won't get the depth results! For example, according to GPU Compute Capability, RTX3080's Compute Capability is 8.6. So you should set the cuda compilation parameter 'arch=compute_86,code=sm_86' or add a '-gencode arch=compute_86,code=sm_86'.

• Compile

mkdir build & cd build
cmake ..
make

• Test

Use script colmap2mvsnet_acm.py to convert COLMAP SfM result to MVS input   
Run ./DPE $data_folder to get reconstruction results.
The result will be saved in the folder $data_folder/DPE, and the point cloud is saved as "DPE.ply"

If you need to filter out the sky during point cloud fusion, you can use a segmentation approach. Please refer to MP-MVS and save the segmentation results in the $data_folder/blocks directory.

This code largely benefits from the following repositories: APD-MVS, HPM-MVS. Thanks to their authors for opening the source of their excellent works!