Textured-GS: Gaussian Splatting with Spatially Defined Color and Opacity

Zhentao Huang, Minglun Gong

Full Paper Trained Model

Setting up

Please refer to 3DGS for setting up. For the Textured-GS submodule:

pip install submodules/diff-gaussian-rasterization_texture/

Running the code

python full_eval.py --output_path <mini-spaltting-results (Trained Model)> --mipnerf360 <mipnerf360 folder> --tanksandtemples <tanks and temples folder> --deepblending <deep blending folder>

We select Mini-Splatting as the input of our algorithm, which is stored in iteration_66 in point_cloud folder. The default output of Textured-GS is stored in iteration_14000 folder.

Acknowledgement.

This project is built upon 3DGS.