pyhks

This is a simple dependency free Python library for the Heat Kernel Signature on triangle meshes. The only dependencies are the numpy/scipy stack. If you want to view the results of the computation, you should also download meshlab.

Running HKS

To see all options, run the script as follows

python hks.py --help

As an example, let's examine the HKS on the "homer" mesh in this repository, at different scales. In each example, we output to a file which can be opened in meshlab, which is the homer mesh colored in grayscale with the values of the HKS

<table> <tr> <td> <code> python hks.py --input homer.off --t 5 --output hks5.off </code> </td> <td> <code> python hks.py --input homer.off --t 20 --output hks20.off </code> </td> <td> <code> python hks.py --input homer.off --t 200 --output hks200.off </code> </td> </tr> <tr> <td> <img src = "hks5.png"> </td> <td> <img src = "hks20.png"> </td> <td> <img src = "hks200.png"> </td> </td> </table>

Notice how at smaller time scales, finer, high frequency curvature detail is present. However, if the time scale is too small, artifacts are present from using a limited number of eigenvectors.

Running Point Sampler

There's a function that comes bundled with this software that samples a triangle mesh uniformly by area. This may be of independent interest to some people. There is a script that can launch this called "PointSampler.py". To see all options, run the script as follows

python PointSampler.py --help

For example, the code

python sampler.py --input homer.off --output out.csv --npoints 1000 --do_plot 1

will evenly sample 1000 points on homer and show a plot, before saving to a csv file called "out.csv"

<img src = "Sampled1000.png">