Skip to content

Latest commit

 

History

History
93 lines (77 loc) · 4.2 KB

README.md

File metadata and controls

93 lines (77 loc) · 4.2 KB
Raw

MeshGraph in PyTorch

Transfroming Mesh data to Mesh Graph Tology through the idea of Finite Element, Paper is published on IEEE url.

Transform to Topology Graph

transfrom

Network Structure

network

Getting Started

Installation

  • Clone this repo:
git clone https://github.com/JsBlueCat/MeshGraph.git
cd MeshVertexNet
cd docker
docker build -t your/docker:meshgraph.
  • then run docker image
docker run --rm -it --runtime=nvidia --shm-size 16G -e DISPLAY=unix$DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix -v /your/path/to/MeshGraph/:/meshgraph your/docker:meshgraph bash

3D Shape Classification on ModelNet40

  • get the dataset if you fail download it from python [Model40]
  • put the zip file in datasets/modelnet40_graph
cd /meshvertex
sh script/modelnet40_graph/train.sh

Classification Acc. On ModelNet40

Method Modality Acc
3DShapeNets volume 77.3%
Voxnet volume 83%
O-cnn volume 89.9%
Mvcnn view 90.1%
Mlvcnn view 94.16%
Pointnet Point cloud 89.2%
Meshnet Mesh 91%
Ours with SAGE Mesh 94.3% +0.5%
  • run test
  • get the dataset if you fail download it from python [Model40]
  • put the zip file in datasets/modelnet40_graph
  • download the weight file from [weights] and put it in your ckpt_root/40_graph/ and run
sh script/modelnet40_graph/test.sh
  • the result will be like
root@2730e382330f:/meshvertex# sh script/modelnet40_graph/test.sh 
Running Test
loading the model from ./ckpt_root/40_graph/final_net.pth
epoch: -1, TEST ACC: [94.49 %]

result1 result2

Train on your Dataset

Coming soon

Some 3D Reconstruct conducted on 3d face

face

Credit

MeshGraphNet: An Effective 3D Polygon Mesh Recognition With Topology Reconstruction

An Ping Song ; Xin Yi Di; Xiao Kang Xu; Zi Heng Song

Abstract
Three-dimensional polygon mesh recognition has a significant impact on current computer graphics. However, its application to some real-life fields, such as unmanned driving and medical image processing, has been restricted due to the lack of inner-interactivity, shift-invariance, and numerical uncertainty of mesh surfaces. In this paper, an interconnected topological dual graph that extracts adjacent information from each triangular face of a polygon mesh is constructed, in order to address the above issues. On the basis of the algebraic topological graph, we propose a mesh graph neural network, called MeshGraphNet, to effectively extract features from mesh data. In this concept, the graph node-unit and correlation between every two dual graph vertexes are defined, the concept of aggregating features extracted from geodesically adjacent nodes is introduced, and a graph neural network with available and effective blocks is proposed. With these methods, MeshGraphNet performs well in 3D shape representation by avoiding the lack of inner-interactivity, shift-invariance, and the numerical uncertainty problems of mesh data. We conduct extensive 3D shape classification experiments and provide visualizations of the features extracted from the fully connected layers. The results demonstrate that our method performs better than state-of-the-art methods and improves the recognition accuracy by 4–4.5%.

If you find this code useful, please consider citing our paper

@article{Song2020MeshGraphNetAE,
  title={MeshGraphNet: An Effective 3D Polygon Mesh Recognition With Topology Reconstruction},
  author={An Ping Song and Xin Yi Di and X. Xu and Zi Heng Song},
  journal={IEEE Access},
  year={2020},
  volume={8},
  pages={205181-205189}
}