Skip to content
Local geometric features for voxel-based representation of surfaces in 3D
Branch: master
Clone or download
Latest commit 2989524 Feb 7, 2017
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
convertSTL2OFF first commit Nov 22, 2016
cpp remove some old files Nov 22, 2016
datasets
demo minor description fixes Jan 13, 2017
pics first commit Nov 22, 2016
README.md added link to paper Feb 6, 2017
gpl.txt first commit Nov 22, 2016
setup.py first commit Nov 22, 2016
voxelfeatures.py first commit Nov 22, 2016

README.md

voxelfeatures

Local geometric features for voxel-based representation of surfaces in 3D

Dmitry Yarotsky, Skoltech/IITP, 2016, GPLv3

  • A collection of routines to generate voxel features describing local area, curvature and orientation of the surface
  • Efficient octree-based C++ code supporting spatial sparsity
  • Python bindings with surface classification examples

The paper: D. Yarotsky, Geometric features for voxel-based surface recognition, arXiv:1701.04249

Example: feature visualization (/demo/demoPlotFeatures.py)

The original surface:

original

Voxelizations with different features:

feature examples

Example: training a classification model (/demo/demoXGBoostESB.py)

Predictions of an XGBoost model on the ESB data set, with about 15% test set error:

predictions

Installation

System requirements

  • OS Linux (tested on Ubuntu 14.04/16.04)
  • Python 2.7

Dependencies

Installation of dependencies for Ubuntu 14.04/16.04:

sudo apt-get install libarmadillo-dev sparsehash

Installation as a Python module

python setup.py install

You can’t perform that action at this time.