Skip to content

Latest commit

 

History

History
63 lines (46 loc) · 1.52 KB

dynamic-3dgs-tutorial.md

File metadata and controls

63 lines (46 loc) · 1.52 KB

Step-by-step tutorial for getting dynamic 3dgs running

Make sure you have installed Nvidia driver (version > 530), CUDA (version > 12) and Docker.

Pull the docker images:

docker pull zmu913/mm811project:latest

Then clone dynamic 3dgs code into your current working directory:

git clone --recursive https://github.com/mzgsxs/3dgs-uoa-mm.git
cd 3dgs-uoa-mm
git checkout d3dgs-soft-bg
cd ..

Then clone render code into your current working directory:

git clone --recursive https://github.com/JonathonLuiten/diff-gaussian-rasterization-w-depth.git

Download pretrained models, and training data

wget https://omnomnom.vision.rwth-aachen.de/data/Dynamic3DGaussians/output.zip  
unzip output.zip
wget https://omnomnom.vision.rwth-aachen.de/data/Dynamic3DGaussians/data.zip
unzip data.zip

Start your docker in nvidia-runtime mode with source code mounted in container:

docker run --rm --runtime=nvidia --gpus all \
        -e DISPLAY=$DISPLAY  -v /tmp/.X11-unix:/tmp/.X11-unix \
        -v $PWD/3dgs-uoa-mm:/root/Dynamic3DGaussians \
        -v $PWD/diff-gaussian-rasterization-w-depth:/root/Dynamic3DGaussians/diff-gaussian-rasterization-w-depth \
        -v $PWD/data:/root/Dynamic3DGaussians/data \
        -v $PWD/output:/root/Dynamic3DGaussians/output \
        -it zmu913/mm811project:0.1 bash

In side docker, go to directory

cd /root/Dynamic3DGaussians/

Visualize pre-trianed dynamic 3dgs

python visualize.py

Trian soft bg dynamic 3dgs by you self

python train.py