PhysDreamer: Physics-Based Interaction with 3D Objects via Video Generation [website]
Install diff-gaussian-rasterization at: https://github.com/graphdeco-inria/diff-gaussian-rasterization
conda create -n physdreamer python
conda activate physdreamer
pip install -r requirements.txt
python setup.py install
Download the scenes and optimized velocity and material fields from: https://huggingface.co/datasets/YunjinZhang/PhysDreamer/tree/main
Put folders of these scenes to data/physics_dreamer/xxx
, e.g. data/physics_dreamer/carnations
Put pretrained models to ./models
.
See dataset_dir
and model_list
in inference/configs/carnation.py
to match the path of dataset and pretrained models.
cd projects/inference
bash run.sh
This codebase used lots of source code from:
- https://github.com/graphdeco-inria/gaussian-splatting
- https://github.com/zeshunzong/warp-mpm
- https://github.com/PingchuanMa/NCLaw
We thank the authors of these projects.
@article{zhang2024physdreamer,
title={{PhysDreamer}: Physics-Based Interaction with 3D Objects via Video Generation},
author={Tianyuan Zhang and Hong-Xing Yu and Rundi Wu and
Brandon Y. Feng and Changxi Zheng and Noah Snavely and Jiajun Wu and William T. Freeman},
journal={arxiv},
year={2024}
}