/SED-Net

Surface and Edge Detection for Primitive Fitting of Point Clouds,SIGGRAPH 2023

Primary LanguagePython

Surface and Edge Detection for Primitive Fitting of Point Clouds SIGGRAPH 2023 conference proceedings

install

following parsenet install cmds

data prepare

  • please download our dataset from Onedrive or from Baiduyun (password is meta), and load all files of folder sed_net_data to data

  • please download parsenet datasets from Onedrive or from Baiduyun (password is meta), and load all files of folder parsenet to data_parsenet

  • you can download our pretained models from Onedrive or from Baiduyun (password is meta), and load all weights of folder ckpts to ckpts

  • you can download SplineNet dataset from Gooogle Drive

train & test

  • test model with normal
python generate_predictions_aug.py configs/config_SEDNet_normal.yml NoSave no_multi_vote no_fold5drop
  • train model with normal
python train_sed_net.py configs/config_SEDNet_normal.yml

mesh creation

arg2mesh/arg2mesh.py

reference

  1. parsenet
  2. hpnet

Cite

@inproceedings{li2023surface, title={Surface and edge detection for primitive fitting of point clouds}, author={Li, Yuanqi and Liu, Shun and Yang, Xinran and Guo, Jianwei and Guo, Jie and Guo, Yanwen}, booktitle={ACM SIGGRAPH 2023 Conference Proceedings}, pages={1--10}, year={2023} }