Description

Dockerfiles for some SOTA SLAM algorithms (mainly Visual Inertial Odometry with SLAM capabilities). Compiled images could be found here.

The repository is part of the publication, which was published in the Journal of Intelligent & Robotic Systems.

Sharafutdinov, D., Griguletskii, M., Kopanev, P., Kurenkov, M., Ferrer, G., Burkov, A., Gonnochenko, A., & Tsetserukou, D. (2023). Comparison of modern open-source visual SLAM approaches. Journal of Intelligent & Robotic Systems, 107(3), 43. (arXiv preprint arXiv:2108.01654. PDF.)

If you use this repository in academic work, please cite:

@article{sharafutdinov2023comparison,
  title={Comparison of modern open-source visual SLAM approaches},
  author={Sharafutdinov, Dinar and Griguletskii, Mark and Kopanev, Pavel and Kurenkov, Mikhail and Ferrer, Gonzalo and Burkov, Aleksey and Gonnochenko, Aleksei and Tsetserukou, Dzmitry},
  journal={Journal of Intelligent \& Robotic Systems},
  volume={107},
  number={3},
  pages={43},
  year={2023},
  publisher={Springer}
}

Available algorithms

PS some dockerfiles fail to compile automatically due to the big size or building is in progress

Kimera

Docker for Kimera-VIO - Visual Inertial Odometry with SLAM capabilities and 3D Mesh generation.

Possible commands (building, pulling, cleaning):

make help

Run container:

./run.sh

Maplab

Docker for Maplab - An open visual-inertial mapping framework. Dockerfile was created using this.

To run, use make:

make

This will build and run the docker container with the maplab components.

Or you can pull compiled image from Docker Hub:

docker pull kopanev/maplab:version3

This version includes CLion and vim for development.

Run container using this command:

xhost +
docker run -v /tmp/.X11-unix:/tmp/.X11-unix -e DISPLAY=$DISPLAY -h $HOSTNAME -v $HOME/.Xauthority:/home/lyonn/.Xauthority kopanev/maplab:version3

VINS-Mono

Docker for VINS-Mono - A Robust and Versatile Monocular Visual-Inertial State Estimator.

You need to put this folder into your project (as it was done in the original repo), or modify Dockerfile (to clone repo into the container)

Run container:

make build

./run.sh LAUNCH_FILE_NAME   # ./run.sh euroc.launch

VINS-Fusion

Docker for VINS-Fusion - An optimization-based multi-sensor state estimator.

You need to put this folder into your project (as it was done in the original repo), or modify Dockerfile (to clone repo into the container)

Run container:

make build

# Euroc Monocualr camera + IMU
./run.sh ~/catkin_ws/src/VINS-Fusion/config/euroc/euroc_mono_imu_config.yaml

# Euroc Stereo cameras + IMU with loop fusion
./run.sh -l ~/catkin_ws/src/VINS-Fusion/config/euroc/euroc_mono_imu_config.yaml

# KITTI Odometry (Stereo)
./run.sh -k ~/catkin_ws/src/VINS-Fusion/config/kitti_odom/kitti_config00-02.yaml YOUR_DATASET_FOLDER/sequences/00/

# KITTI Odometry (Stereo) with loop fusion
./run.sh -kl ~/catkin_ws/src/VINS-Fusion/config/kitti_odom/kitti_config00-02.yaml YOUR_DATASET_FOLDER/sequences/00/

#  KITTI GPS Fusion (Stereo + GPS)
./run.sh -kg ~/catkin_ws/src/VINS-Fusion/config/kitti_raw/kitti_10_03_config.yaml YOUR_DATASET_FOLDER/2011_10_03_drive_0027_sync/

OpenVINS

Docker for OpenVINS - An open source platform for visual-inertial navigation research. Note that this container launches with flag --net=host (gives the container full access to local system services such as D-bus and is therefore considered insecure). ROS master will run on the host PC and will see all topics published in the container.

Possible commands (building, pulling, cleaning):

make help

To run the simulation from the examples:

./run.sh pgeneva_serial_eth.launch

VIORB

Docker for VIORB - Visual Inertial ORB SLAM based on ORB-SLAM2.

Possible commands (building, pulling, cleaning):

make help

To run the container:

./run.sh

Then inside the container:

export ROS_PACKAGE_PATH=${ROS_PACKAGE_PATH}:$CATKIN_WS/src/Examples/ROS 
roslaunch ORB_VIO testeuroc.launch

OpenVSLAM

Possible commands (building, pulling, cleaning):

make help

To run the container:

xhost +local
docker run -it --rm -e DISPLAY=$DISPLAY -v /tmp/.X11-unix/:/tmp/.X11-unix:ro ros:openvslam

ORB_SLAM2

Pull docker image:

docker pull youyu/orb_slam2:ubuntu18

(this fork is used)

To run the container:

xhost +local:
sudo docker run --name orb -it --rm -e DISPLAY=$DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix:ro \
youyu/orb_slam2:ubuntu18