Sovereign 8s.is mirror of standardgalactic/SVGS
  • Python 76%
  • Cuda 19.2%
  • C++ 4.5%
  • CMake 0.2%
Find a file
Repository files (latest commit first)
Filename Latest commit message Latest commit date
2026-03-31 17:05:45 +08:00
arguments Add files via upload 2024-11-14 20:25:34 +08:00
assets Add files via upload 2024-11-14 20:25:34 +08:00
gaussian_renderer Add files via upload 2024-11-14 20:25:34 +08:00
lpipsPyTorch Add files via upload 2024-11-14 20:25:34 +08:00
scene Add files via upload 2024-11-14 20:25:34 +08:00
scripts Add files via upload 2024-11-14 20:25:34 +08:00
submodules 111 2024-11-14 20:28:50 +08:00
utils Add files via upload 2024-11-14 20:26:27 +08:00
convert.py Add files via upload 2024-11-14 20:25:34 +08:00
environment.yml Update environment.yml 2026-03-31 17:05:45 +08:00
LICENSE.md Add files via upload 2024-11-14 20:25:34 +08:00
metrics.py Add files via upload 2024-11-14 20:25:34 +08:00
README.md Update README.md 2026-03-31 17:05:35 +08:00
render.py Add files via upload 2024-11-14 20:25:34 +08:00
train.py Add files via upload 2024-11-14 20:25:34 +08:00
view.py Add files via upload 2024-11-14 20:25:34 +08:00

SVGS: Enhancing Gaussian Splatting Using Primitives with Spatially Varying Colors

Teaser image [Project] [Arxiv]

Gaussian Splattings demonstrate impressive results in multi-view reconstruction based on Gaussian explicit representations. However, the current Gaussian primitives only have a single view-dependent color and an opacity to represent the appearance and geometry of the scene, resulting in a non-compact representation. In this paper, we introduce a new method called SVGS that utilizes spatially varying colors and opacity in a single Gaussian primitive to improve its representation ability. We have implemented bilinear interpolation, movable kernels, and even tiny neural networks as spatially varying functions. Quantitative and qualitative experimental results demonstrate that all three functions outperform the baseline, with the best movable kernels achieving superior novel view synthesis performance on multiple datasets, highlighting the strong potential of spatially varying functions.

News: We have release the MLP version of SVGS! Find our code [here]!

Installation

# download
git clone

# if you have an environment used for 3dgs, use it
# if not, create a new environment
conda env create --file environment.yml
conda activate SVGS

Training

To train a scene, simply use

python train.py -s <path to COLMAP or NeRF Synthetic dataset>

Testing

Bounded Mesh Extraction

python render.py -m <path to pre-trained model> -s <path to COLMAP dataset> 

Quick Examples

Assuming you have downloaded [MipNeRF360], simply use

python train.py -s <path to m360>/<garden> -m output/m360/garden
# use our unbounded mesh extraction!!
python render.py -s <path to m360>/<garden> -m output/m360/garden --unbounded --skip_mesh --skip_train 


**Custom Dataset**: We use the same COLMAP loader as 3DGS, you can prepare your data following 3DGS.


#### Novel View Synthesis
For novel view synthesis on [MipNeRF360] (which also works for other colmap datasets), use
```bash
python scripts/mipnerf_eval.py -m360 <path to the MipNeRF360 dataset>