The C++ (with cuda) version of the classic Marching Cubes algorithm.

This repo contains the code for the classic Marching Cubes algorithm in C++&Cuda. The input is a SDF tensor with shape (dim-x, dim-y, dim-z). The output is the result reconstructed mesh by the MC algorithm.

1. eigen3
2. cuda-8.0
3. Python3

Clone the repo:

    git clone https://github.com/WanquanF/MarchingCubes_CppCuda.git

Build the code with the given script:

    python run.py

Run the testing sample:

    python test.py

The code has two kinds of implementations of the MC algorithm. One implementation uses the cpu all the time while the other one uses the gpu for parallel. After you build the code by running the "run.py" script, the exe file is in the "./build" folder, named "marching_cubes_cuda". The command line for using it is like this:

    ./build/marching_cubes_cuda  dim-x  dim-y  dim-z  input-sdf-path  output-mesh-path  if-using-cuda

The "dim-x", "dim-y" and "dim-z" are int numbers showing the shape of the input sdf, and the "if-using-cuda" is a 0/1 (0 means running without cuda, 1 means running with cuda) value deciding which version of implementation you use. The "input-sdf-path" is a bin file of the sdf value tensor. The "output-mesh-path" should be in the obj format.