/Gaussian_Splatting_Realtime_Render

3D Gaussian Splatting Real-time Rendered Image

Primary LanguagePython

3D Gaussian Splatting Real-time Renderer from VINS Pose

Prerequisites

1.1 Ubuntu and ROS

Ubuntu 64-bit 16.04 or 18.04. ROS Kinetic or Melodic. ROS Installation

1.2 Catkin Tools

Install catkin_tools. Follow catkin_tools Installation

1.3 Anaconda

Install Latest Version. Follow Anaconda Installation

1.4. Ceres Solver

Follow Ceres Installation.

Build

Clone the repository and catkin_make:

    $ cd ~/catkin_ws/src
    $ git clone https://github.com/SungJaeShin/Gaussian_Splatting_Realtime_Render.git
    $ cd ../
    $ catkin build camera_models
    $ source ~/catkin_ws/devel/setup.bash
    $ catkin build vins
    $ source ~/catkin_ws/devel/setup.bash

Run

You must have a pre-trained model and have bag data and camera parameters !!!

    # Terminal 1
    $ rosrun vins vins_node ./VINS-Fusion/config/euroc/euroc_stereo_imu_config.yaml

    # Terminal 2
    $ python3 realtime_render.py -m <TRAINED_MODEL_PATH> -s <DATASET_PATH>
 

Results (KAIST N25 Building Indoor Dataset)

  • Random Pose from VINS-Fusion Keyframe Pose
    random_pose

  • Fixed z position from VINS-Fusion Keyframe Pose
    fixed_z

Reference

[1] VINS-Fusion
[2] Gaussian Splatting