ResGEM

Official code for the TVCG paper "ResGEM: Multi-scale Graph Embedding Network for Residual Mesh Denoising"

Environment

Pytorch and Pytorch Geometric are required.

Usage

We temporally provide PREPROCESSED Synthetic data and checkpoint of the first normal regression iteration, which can be run by the following command:

python infer.py --testset ./data/test --data_type Synthetic --graph_type DENSE_FACE --resume ./ckpt/Synthetic_nn.pth --save_dir ./data/results

Citation

@ARTICLE{ResGEM2024,
  author={Zhou, Ziqi and Yuan, Mengke and Zhao, Mingyang and Guo, Jianwei and Yan, Dong-Ming},
  journal={IEEE Transactions on Visualization and Computer Graphics}, 
  title={ResGEM: Multi-scale Graph Embedding Network for Residual Mesh Denoising}, 
  year={2024},
  pages={1-17},
  doi={10.1109/TVCG.2024.3378309}}