Mulin Yu*, Tao Lu*, Linning Xu, Lihan Jiang, Yuanbo Xiangli✉, Bo Dai

[Project Page][arxiv]

Representing 3D scenes from multiview images remains a core challenge in computer vision and graphics, requiring both reliable rendering and reconstruction, which often conflicts due to the mismatched prioritization of image quality over precise underlying scene geometry. Although both neural implicit surfaces and explicit Gaussian primitives have advanced with neural rendering techniques, current methods impose strict constraints on density fields or primitive shapes, which enhances the affinity for geometric reconstruction at the sacrifice of rendering quality. To address this dilemma, we introduce GSDF, a dual-branch architecture combining 3D Gaussian Splatting (3DGS) and neural Signed Distance Fields (SDF). Our approach leverages mutual guidance and joint supervision during the training process to mutually enhance reconstruction and rendering. Specifically, our method guides the Gaussian primitives to locate near potential surfaces and accelerates the SDF convergence. This implicit mutual guidance ensures robustness and accuracy in both synthetic and real-world scenarios. Experimental results demonstrate that our method boosts the SDF optimization process to reconstruct more detailed geometry, while reducing floaters and blurry edge artifacts in rendering by aligning Gaussian primitives with the underlying geometry.

We tested on a server configured with Ubuntu 22.04, cuda 11.8 and gcc 11.4.0. Other similar configurations should also work, but we have not verified each one individually.

1. Clone this repo:

git clone https://github.com/city-super/GSDF.git --recursive
cd GSDF

2. Install dependencies

SET DISTUTILS_USE_SDK=1 # Windows only
conda env create --file environment.yml
conda activate gsdf

We are working to accelerate the training speed and will release the update soon.

First, create a data/ folder inside the project path by

mkdir data

The data structure will be organised as follows:

data/
├── dataset_name
│   ├── scene1/
│   │   ├── images
│   │   │   ├── IMG_0.jpg
│   │   │   ├── IMG_1.jpg
│   │   │   ├── ...
│   │   ├── sparse/
│   │       └──0/
│   ├── scene2/
│   │   ├── images
│   │   │   ├── IMG_0.jpg
│   │   │   ├── IMG_1.jpg
│   │   │   ├── ...
│   │   ├── sparse/
│   │       └──0/
...

The MipNeRF360 scenes are provided by the paper author here. And we test on scenes bicycle, bonsai, counter, garden, kitchen, room, stump. The SfM data sets for Tanks&Temples and Deep Blending are hosted by 3D-Gaussian-Splatting here. Download and uncompress them into the data/ folder. The DTU scenes are downloaded from here.

For custom data, you should process the image sequences with Colmap to obtain the SfM points and camera poses. Then, place the results into data/ folder.

For training a single scene, modify the path and configurations in train.sh accordingly and run it:

bash ./train.sh

• exp_dir: user-defined experiment directory;
• config: path of config file;
• gpu: specify the GPU id to run the code. '-1' denotes using the most idle GPU.
• train/eval: training/eval mode;
• tag: user-defined experiment name.

This script will store the log (with running-time code) into outputs/${tag} (for GS branch) and exp/scenename/${tag} (for SDF branch) automatically.

We've integrated the rendering and metrics calculation process into the training code. So, when completing training, the rendering results, fps and quality metrics will be printed automatically. And the rendering results will be save in the log dir. Mind that the fps is roughly estimated by

torch.cuda.synchronize();t_start=time.time()
rendering...
torch.cuda.synchronize();t_end=time.time()

which may differ somewhat from the original 3D-GS, but it does not affect the analysis.

Meanwhile, we keep the manual rendering function with a similar usage of the counterpart in 3D-GS, one can run it by

python render.py -m <path to trained model> # Generate renderings
python metrics.py -m <path to trained model> # Compute error metrics on renderings

For the geomtry reconstruction evaluation, please refer to 2DGS for calculating the Chamfer Distance.

• Mulin Yu: [email protected]
• Tao Lu: [email protected]

If you find our work helpful, please consider citing:

@article{yu2024gsdf,
  title={Gsdf: 3dgs meets sdf for improved rendering and reconstruction},
  author={Yu, Mulin and Lu, Tao and Xu, Linning and Jiang, Lihan and Xiangli, Yuanbo and Dai, Bo},
  journal={arXiv preprint arXiv:2403.16964},
  year={2024}
}

Please follow the LICENSE of 3D-GS.

The GS-branch is built based on Scaffold-GS and the SDF-branch is built based on Instant-NSR, we thank all the authors for the excellent work.