MeshGraph in PyTorch

Transfroming Mesh data to Mesh Graph Tology through the idea of Finite Element, Paper is publishing soon.

Getting Started

Installation

Clone this repo:

git clone https://github.com/JsBlueCat/MeshGraph.git
cd MeshVertexNet

Install dependencies: nvidia-docker and docker
First intall docker image

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 %]

Train on your Dataset

Coming soon

Credit

MeshGraph: A New Mesh Process Via the Idea of Finite Element

AnPing S,XinYi D

Abstract
Everything will get rapid development once it is related to mathematics. In our current research, a method that combines the presentation of mesh with the topology of Finite Element has been proposed. In previous years, researchers have been thinking about the presentation of data of the 3D shapes.And there have been several popular types of data such as Volumetric Grid, Muiti-View, Point Cloud, Mesh. During to Mesh contains both geometric feature information and spatial feature information of 3D data, it has been more popular in rescent years.Although Mesh data has the properties of complexity and irregularity, fortunetily, researchers have already make much tranformation to Mesh data so as to avoid the compleity and irregularity problem. Basing on The Finite Element Concept in Fluid Physics, we re-discussed the organization of Mesh, and proposed a new Organization of Mesh data —— Mesh Topology, which is based on the mini-batch idea of the graph, and using the convolution method to extract the adjacent location information according to the basic structure of the topological network graph. As well as the network structure which integrates geometric and physical features has been finally proposed. In this paper, the advantages of this network are discussed in theory, and many experiments are designed to prove the validity and accuracy of this network,finally the network is achieved a classification accuracy of 94.3%, which can be achieved in https://github.com/JsBlueCat/MeshGraph.

[Paper] [Authors' Implementation]

@article{meshgraph,
title={MeshGraph: A New Mesh Process Via the Idea of Finite Element },
author={XinYi D},
journal = {arXiv},
year={2019}
}

junhua-zhang/MeshGraph