🏔 Geo-CNN-Pytorch-PYG

A Pytorch re-implementation of “Modeling Local Geometric Structure of 3D Point Clouds using Geo-CNN”

• This repository is a reproduction of the GeoCNN, which can support multiple GPUs.
• My enviroment:
  • Ubuntu 18.04
  • Anaconda Python 3.7
  • Pytorch 1.5.0
  • PYG 1.5.0
  • Cuda 10.2
  • Cudnn 7.6.5
  • GPU Memory >= 8G
• If you like graph neural network, too. Welcome to our 🐧 QQ group: 832405795

Accuracy on ModelNet40

this implementation	original paper
93.2	93.4

How to Use This Code

• Prepare Data
  • Download ModelNet40 data set
  • Move modelnet40_normal_resampled.zip into data/ModelNet40_10000
  • Unzip modelnet40_normal_resampled.zip
  • Rename modelnet40_normal_resampled to raw
• Train
  • We can change args in the Configuration part of the code if you want
  • Then let’s start training: python geocnn.py
• Test
  • Uncomment this line and replace the weight path
  • Set only_test as True
  • Then let’s start testing: python geocnn.py

Bibtex

@article{DBLP:journals/corr/abs-1811-07782,
  author    = {Shiyi Lan and
              Ruichi Yu and
              Gang Yu and
              Larry S. Davis},
  title     = {Modeling Local Geometric Structure of 3D Point Clouds using Geo-CNN},
  journal   = {CoRR},
  volume    = {abs/1811.07782},
  year      = {2018},
  url       = {http://arxiv.org/abs/1811.07782},
  archivePrefix = {arXiv},
  eprint    = {1811.07782},
  timestamp = {Mon, 26 Nov 2018 12:52:45 +0100},
  biburl    = {https://dblp.org/rec/bib/journals/corr/abs-1811-07782},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}