This repository provides code, data and pretrained models for SSCF.
[Project Webpage] | [Paper]
- Create a new conda env
conda create -n sscf python=3.9 cmake make compilers pybind11 eigen -c conda-forge
- Install pyigl
pip install git+https://github.com/zzilch/pyigl
# pip install git+https://gitee.com/zzilch/pyigl
- Install
Pytorch
andPytorch3D
based on your cuda version. (OurPytorch
version is 2.4.0 andcuda
version is 12.2)
conda install pytorch==2.4.0 torchvision==0.19.0 torchaudio==2.4.0 pytorch-cuda=12.1 -c pytorch -c nvidia
conda install pytorch3d -c pytorch3d
- Install the human models with
smplx
andmano
.
pip install smplx
pip install git+https://github.com/lixiny/manotorch.git
- Install other dependencies.
pip install trimesh[all] pyvista[all] point-cloud-utils h5py scikit-learn cycpd chumpy scikit-image matplotlib imageio plotly opencv-python open3d warp-lang
pip install numpy==1.23.0
We directly provide the processed data and checkpoints, please download them and organize them as following:
sscf/
└── data/
├── ckpts/
├── hoi/
├── human_model/
├── shapenet/
For details on data preprocessing, please refer to preprocess_data.
If you want to evaluate our model on human-chair interaction, please run:
python generate_chairs_interaction.py
If you want to evaluate our model on hand-mug interaction, please run:
python generate_mugs_interaction.py
We also provide our reproduction code of other baselines, please refer to baselines for more details.
If you find SSCF useful in your research, please cite our paper:
@article{InterTransfer24,
title={Spatial and Surface Correspondence Field for Interaction Transfer},
author={Zeyu Huang and Honghao Xu and Haibin Huang and Chongyang Ma and Hui Huang and Ruizhen Hu},
journal={ACM Transactions on Graphics (Proceedings of SIGGRAPH)},
volume={43},
number={4},
pages={83:1--83:12},
year={2024},
}
We thank for the following excellent open source projects: