Final project of the course Graph Convolutional Networks.
This repository is partly borrowed from original paper's code, and I fixed part of code for running without problem.
This repository is only for the final porject.
There are many package version requirements, so please install a new conda enviroment to run the code.
Install the conda enviroment by running:
conda env create -f deepgcn.yml
conda activate deepgcn
.
├── images # images
├── utils # common useful modules
├── gcn_impl # gcn library
├── sem_seg # code for point clouds semantic segmentation on S3DIS
└── ...
Please look the details in README.md
in sem_seg
folder.
All the information of the code, data, and pretrained models can be found there.
First enter the example directory,
cd sem_seg/
and then run:
CUDA_VISIBEL_DIVICES=0,1,2,3 python train.py --phase train --multi_gpus --batch_size 8
Lower the batch size if out of memory. The batch size will not influence the test results.
Tairen Piao
E-mail: piaotairen@snu.ac.kr