HodgeNet: Learning Spectral Geometry on Triangle Meshes
Dmitriy Smirnov, Justin Solomon
SIGGRAPH 2021
To install the necessary dependencies, run:
conda env create -f environment.yml
conda activate HodgeNet
To train the segmentation model, first download the Shape COSEG dataset. Then, run:
python train_segmentation.py --out out_dir --mesh_path path_to_meshes --seg_path path_to_segs
To train the classification model, first download the SHREC 2011 dataset:
wget -O shrec.tar.gz https://www.dropbox.com/s/4z4v1x30jsy0uoh/shrec.tar.gz?dl=0
tar -xvf shrec.tar.gz -C data
Then, run:
python train_classification.py --out out_dir
To train the dihedral angle stress test model, run:
python train_origami.py --out out_dir
To monitor the training, launch a TensorBoard instance with --logdir out_dir
To finetune a model, add the flag --fine_tune to the above training commands.
@article{smirnov2021hodgenet,
title={{HodgeNet}: Learning Spectral Geometry on Triangle Meshes},
author={Smirnov, Dmitriy and Solomon, Justin},
year={2021},
journal={SIGGRAPH}
}
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent