Environments

sudo apt-get update && sudo apt-get install -y \
    libsuitesparse-dev libeigen3-dev libboost-all-dev \
    libglm-dev libjsoncpp-dev libvtk9-dev libpcl-dev \
    libopencv-dev libglfw3-dev libglew-dev libzip-dev libflann-dev

Install CUDA and PyTorch respectively. We used CUDA Version: 11.8 and PyTorch Version: 2.2.2

Datasets

-Download

cd dataset
bash download_replica.sh
bash download_tum.sh

-Preprocess

bash preprocess.sh

Run

Build

mkdir build && cd build
cmake ..
make -j$(nproc)

Replica

cd experiments_bash
bash replica.sh

TUM-RGBD

cd experiments_bash
bash tum.sh

If use_gaussian_viewer is set as 1, gaussian viewer will appear.

Evaluation

cd experiments_bash
bash replica_eval_rendering.sh
bash replica_eval_depth.sh