Skip to content

Repository files navigation

3D Shape Completion using Multi-resolution Spectral Encoding

This is the working code for WACV 2025(Oral) paper 3D Shape Completion using Multi-resolution Spectral Encoding.

Overview of our work

Create environment

conda env create -f environment.yaml
conda activate MSSC

Download datasets

Download the processed datasets and the priors from PatchComplete.

Training

  1. Train for patches R=4, R=8, and R=32 on ShapeNet dataset.
sh train_script_patch.sh
  1. Train multi-resolution on ShapeNet.
sh train_script_multi.sh
  1. Fine-tune on ScanNet dataset
sh train_script_fine.sh

Generation and Evaluation

  1. Generation -
sh generation_script.sh
  1. Evaluation -
cd evaluation
sh evaluation_script.sh

Please cite us using the following if you found our work useful -

@InProceedings{Deka_2025_WACV,
    author    = {Deka, Pallabjyoti and Bhattacharya, Saumik and Sen, Debashis and Biswas, Prabir Kumar},
    title     = {3D Shape Completion using Multi-Resolution Spectral Encoding},
    booktitle = {Proceedings of the Winter Conference on Applications of Computer Vision (WACV)},
    month     = {February},
    year      = {2025},
    pages     = {54-63}
}

About

No description, website, or topics provided.

Resources

Stars

2 stars

Watchers

1 watching

Forks

Releases

Packages

Contributors

Languages