Yumeng He, Yunbo Wang†, Xiaokang Yang
This repository contains the official code for our paper: GS-Phong: Meta-Learned 3D Gaussians for Relightable Novel View Synthesis.
-
Create an environment
conda create -n gs-phong python=3.10 conda activate gs-phong
-
Install dependencies
git clone https://github.com/ymhe12/GS-Phong.git cd GS-Phong git submodule update --init --recursive pip install -r requirements.txt pip install -e submodules/depth-diff-gaussian-rasterization pip install -e submodules/simple-knn pip install ./bvh
-
Stage 1&2: Gaussian initialization & Normal finetuning
python train.py -s <path_to_your_dataset> -m <path_to_ouput_folder> --eval
-
Stage 3: Meta-learning
python train_meta.py -s <path_to_your_dataset> -m <path_to_ouput_folder> --eval
-
Generate NVS renderings
python render.py -m <path_to_ouput_folder>
-
Calculate error metrics
python metrics.py -m <path_to_ouput_folder>
You can run the full experiment using: (remember ro edit the $DATADIR and $OUTPUT location)
sh run.sh
We appreciate the following github repos where we borrow code from:
Thanks for their amazing works!
If you find our work helps, please cite our paper:
@article{he2024gs,
title={GS-Phong: Meta-Learned 3D Gaussians for Relightable Novel View Synthesis},
author={He, Yumeng and Wang, Yunbo and Yang, Xiaokang},
journal={arXiv preprint arXiv:2405.20791},
year={2024}
}