README.md

As-Plausible-As-Possible (APAP)

Project Page | Paper | arXiv

Seungwoo Yoo*¹, Kunho Kim*¹, Vladimir G. Kim², Minhyuk Sung¹ (* co-first authors)

¹KAIST, ²Adobe Research

This is the reference implementation of As-Plausible-As-Possible: Plausibility-Aware Mesh Deformation Using 2D Diffusion Priors (CVPR 2024).

We have modified the code in this repository to remove the 2D diffusion prior for our use case.

Get Started

Clone the repository and create a Python environment:

conda create --name apap python=3.9
conda activate apap
conda install pytorch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2 pytorch-cuda=11.7 -c pytorch -c nvidia
conda install pytorch-sparse -c pyg
pip install wandb
pip install diffusers==0.19.0
pip install accelerate transformers ninja
pip install cholespy libigl
pip install 'imageio[ffmpeg]' jaxtyping tyro
pip install fpsample trimesh pymeshlab pyrender
pip install -r requirements.txt

The authors of APAP provide data necessary to run the codes via Google Drive. Specifically, you can download

• APAP-Bench 3D

Download the files from the link and place them under the directory data. After that, the directory structure should look like:

APAP
├── data
│   ├── apap_3d  # APAP-Bench 3D
│   └── ...
├── ext
├── scripts
├── src
├── environment.yaml
└── README.md

Making Deformations using APAP-Bench

To run 3D mesh deformation experiments using APAP-Bench (3D), run:

python scripts/exp/batch/batch_deform_meshes.py \
--data-list-path configs/deform_meshes/data/apap_3d.txt
--out-root outputs/apap-3d
--gpu-ids 0

Note that the experiments can be parallelized by specifying multiple GPU IDs via the argument --gpu-ids.

Citation

Please consider citing the following work if you find this codebase useful:

@inproceedings{yoo2024apap,
  title = {{As-Plausible-As-Possible: Plausibility-Aware Mesh Deformation Using 2D Diffusion Priors}},
  author = {Yoo, Seungwoo and Kim, Kunho and Kim, Vladimir G. and Sung, Minhyuk},
  booktitle = {CVPR},
  year = {2024},
}