Implements Parallel Implementation of 3D mesh simplification using Quadric Error Metrics and Edge Length
Install CGAL.
sudo apt install libcgal-dev
Build with make.
make
Run the program with some input.
./Simplify <input_file> <fraction of points to remove> <decimation method (elen/qem)> <grid_resolution> <no. of threads>
Please use the Object File Format (.off) as input file. You can convert most mesh formats to OFF using MeshLab.
Example usage:
./Simplify data/bunny_zipper.off 0.25 qem 64 16
This simplification on the bunny_zipper model to remove 25% of the points using the Quadric Error Metric (QEM) with uniform grid resolution of 64
running on 16 threads.
The model in data/bunny_zipper.off is provided here for simplicity and to facilitate using this program, converted using MeshLab.
The original model is the Stanford Bunny. I encourage you to read about it.
After 5 years of being done with this project, I've revisited this repository and thought of some improvements I could do to make the code more interesting.
When I worked on this project (for my MSc thesis), our best computer was a 64GB Intel 16-core i5 (maybe?), and we had some relatively powerful GPUs. The GPUs could barely hold triangular meshes with a few million points. I believe some of the models used in the experiments were too large to be loaded in the GPU. But technological improvement is not all that has happened since then. I sometimes wonder how much I could have made this project more interesting and usable had I had the experience I acquired in these 5 years.
That being said, here is a list of simple improvements:
Python wrapper: Wrap these methods in Python 3. Should only take a few. Might even work better. Object viewer: Since we'll be using Python, create an object viewer that can show loaded objects in real-time, and show the simplification intermediate steps as if in a video or animation. GPU: Using GPU with Python has become quite simpler these days. PyCuda and scikit-cuda are very helpful.