Skip to content

Latest commit

 

History

History
107 lines (85 loc) · 3.61 KB

README.md

File metadata and controls

107 lines (85 loc) · 3.61 KB
Raw

Reconstruction and Simulation of Elastic Objects with Spring-Mass 3D Gaussians

Licheng Zhong · Hong-Xing "Koven" Yu · Jiajun Wu · Yunzhu Li

ECCV 2024

Project Page | arXiv | Data

Logo
Spring-Gaus_demos.mp4

🚀 Dependencies

git clone https://github.com/Colmar-zlicheng/Spring-Gaus.git --recursive
cd Spring-Gaus

# create conda environment
conda create -n Spring_Gaus python=3.7
conda activate Spring_Gaus
conda install pytorch torchvision torchaudio pytorch-cuda=11.6 cudatoolkit=11.6 -c pytorch -c nvidia
pip install -r requirements.txt

# install submodules for 3D Gaussian Splatting
pip install submodules/diff-gaussian-rasterization
pip install submodules/simple-knn

# install pytorch3d (https://pytorch3d.org/)
pip install "git+https://github.com/facebookresearch/pytorch3d.git"

💿 Datasets

Our synthetic data and real world data are provided at Google Drive. Download and extract the data to the data folder. If you don't have one, make it mkdir data.

├── data
│   ├── mpm_synthetic
│   │   ├── ... 
│   ├── real_capture
│   │   ├── ...

🚂 Train

General command

python train.py -g 0 --cfg ${CONFIG_FILE} --eval_cam ${EVAL_CAM_ID} --exp_id ${EXP_ID}

Command line arguments

  • -g, --gpu_id, visible GPUs for training, e.g. -g 0. Only supports single GPU.
  • -c, --cfg, e.g. --cfg ${CONFIG_FILE}. Config file path.

Optional:

  • --eval_cam, e.g. --eval_cam 0. Camera id used in evaluation. If not provided, will use a random id.
  • --exp_id specify the name of experiment, e.g. --exp_id ${EXP_ID}. It defaults to 'default'. All results will be saved in exp/${EXP_ID}_{timestamp}.

For example

# mpm_synthetic: torus
python train.py -g 0 --cfg config/mpm_synthetic/torus.yaml --eval_cam 5 --exp_id torus
# real_capture:  burger
python train.py -g 0 --cfg config/real_capture/burger.yaml --eval_cam 0 --exp_id burger

Checkpoint

All the training checkpoints are saved at exp/${EXP_ID}_{timestamp}/checkpoints/

🧪 Test

  • set ${PATH_TO_CHECKPOINT} as the path to the checkpoint (Spring_Mass.pth.tar) to be loaded
python test.py -g 0 --cfg ${CONFIG_FILE} --exp_id ${EXP_ID} --dy_reload ${PATH_TO_CHECKPOINT}

🔑 License

The code and model provided herein are available for usage as specified in the LICENSE file. By downloading and using the code you agree to the terms in the LICENSE.

🌏 Citation

@article{zhong2024springgaus,
    title     = {Reconstruction and Simulation of Elastic Objects with Spring-Mass 3D Gaussians},
    author    = {Zhong, Licheng and Yu, Hong-Xing and Wu, Jiajun and Li, Yunzhu},
    journal   = {European Conference on Computer Vision (ECCV)},
    year      = {2024}
}

For more questions, please contact Licheng Zhong: zlicheng@sjtu.edu.cn