Iterative Poisson Surface Reconstruction (iPSR) for Unoriented Points
Fei Hou, Chiyu Wang, Wencheng Wang, Hong Qin, Chen Qian, Ying He
ACM Transactions on Graphics, 41, 4, Article 128, 13 pages, 2022. (SIGGRAPH 2022)

iPSR extends the popular Poisson Surface Reconstruction (https://github.com/mkazhdan/PoissonRecon). iPSR has no more need of oriented normals as input, but infers the normals in an iterative manner. It is used to reconstruct surface from only points input.
project page

Windows: The code is tested by Visual Studio. The ipsr.vcxproj is an example to configure the project.
Linux: The code is tested by GCC and Clang with makefile.
Executable: Win64

--in <input ply file name>
The input file should be in .ply format and only 3D point coordinates are needed.

--out <output ply file name>
The output file name. It should be in .ply format.

[--iters <maximum number of iterations>]
The maximum number of iterations. The default value of this parameter is 30.

[--pointWeight <interpolation weight>]
The pointWeight parameter of screened Poisson surface reconstruction. The default value for this parameter is 10.

[--depth <reconstruction depth>]
The depth parameter of screened Poisson surface reconstruction. It is the maximum possible depth of the octree. The default value of this parameter is 10.

[--neighbors <number of neighbors>]
The number of the closest sample points to search from every reconstructed triangle face. The suggested range is between 10 and 20. The default value of this parameter is 10.

The models in the data folder all could run with default parameters. Smaller depth helps with noisy points.