Skip to content
Point Convolutional Neural Networks by Extension Operators
Python Shell
Branch: master
Clone or download
matanatz Merge pull request #4 from prashantraina/master
Fixed mistake in unit test for convolution
Latest commit f2cb23d Jan 1, 2019
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
confs part segmentation Aug 30, 2018
layers part segmentation Aug 30, 2018
part_segmentation part segmentation data script Sep 12, 2018
unit_tests Fixed mistake in unit test for convolution Dec 29, 2018
.gitignore add segmentation testing file Aug 30, 2018
README.md readme update Aug 21, 2018
evaluate.py fix evaluate Aug 10, 2018
pointcloud_conv_net.py part segmentation Aug 30, 2018
provider.py fix evaluate Aug 10, 2018
teaser.png add teaser file Aug 21, 2018
tf_util.py support rotations Aug 9, 2018
train.py part segmentation Aug 30, 2018

README.md

PCNN - Point Convolutional Neural Networks by Extension Operators

This repository contains an implementation to the SIGGRAPH 2018 paper: PCNN - Point Convolutional Neural Networks by Extension Operators.

PCNN is a novel framework for applying convolutional neural networks to point clouds. The framework consists of two operators: extension and restriction, mapping point cloud functions to volumetric functions and vise-versa. A point cloud convolution is defined by pull-back of the Euclidean volumetric convolution via an extension-restriction mechanism. 

For more details visit: https://arxiv.org/abs/1803.10091.

Installation Requirmenets

The code is compatible with python 3.5 + tensorflow 1.8. In addition, the following packages are required:
pyhocon, h5py.

Usage

  • To run the training procedure on the ModelNet40 classification task:
    python train.py

  • Training outputs are saved in:
    ./exp_results/pcnn/[host_name]/[gpu]/[timestamp]

  • To run evaluation:
    python ./exp_results/pcnn/[host_name]/[gpu]/[timestamp]/evaluate.py

The file pointconv.conf containts additional confguration parameters. To train with a different config file:
python train.py --config file_name

You can’t perform that action at this time.