yanx27 / GACNet

Pytorch implementation of 'Graph Attention Convolution for Point Cloud Segmentation'

Graph Attention Convolution for Point Cloud Segmentation

This is personal pytorch implmentation of GACNet on S3DIS dataset, but not official version.

Download Data

Run download_data.sh and save dataset in ./indoor3d_sem_seg_hdf5_data/

Train Model

Run python train_semseg.py

Announcement

It is only a personal implmentation, and the experimental results do not represent the model in paper. There are still many hyper parameters that need to be adjusted when the author publishes the source code.

Environments

Ubuntu 16.04
Python 3.6.5
Pytorch 0.4.1