Skip to content

Latest commit

 

History

History

4_8

Generalized ICP (G-ICP), Normal Distributions Transform (NDT), and TEASER++

  • Load Velodyne lidar points from KITTI dataset.
  • Perform G-ICP on the LiDAR points.
  • Perform NDT on the LiDAR points.
  • Perform TEASER++ on the LiDAR points.

How to build & run

Requirement: PCL, TEASER++

Local build

mkdir build && cd build
cmake ..
make -j
./gicp /data/kitti/sequences/00/velodyne
./ndt /data/kitti/sequences/00/velodyne
./teaser_plus_plus

Docker build

Requires base build

docker build . -t slam:4_8
docker run -it --env DISPLAY=$DISPLAY -v /kitti:/data/ -v /tmp/.X11-unix/:/tmp/.X11-unix:ro slam:4_8

# Inside docker container
cd fastcampus_slam_codes/4_8
./gicp /data/sequences/00/velodyne
./ndt /data/sequences/00/velodyne
./teaser_plus_plus

Output

Red: Target point cloud Green: Transformed point cloud (Not really visible, as the points lie behind the target point cloud)

If you can see green point clouds, it would actually mean that the registration has not worked perfectly.

G-ICP

gicp

NDT

ndt

TEASER++

teaser_plus_plus

Fast-GICP

Original repo: https://github.com/SMRT-AIST/fast_gicp

How to build

docker build . -f Dockerfile_fast_gicp -t slam:fast_gicp

How to run

xhost +local:docker
docker run -it --env DISPLAY=$DISPLAY -v /kitti:/data -v /tmp/.X11-unix/:/tmp/.X11-unix:ro slam:fast_gicp

# Inside docker container
cd fast_gicp/build
./gicp_kitti /data/sequences/00/velodyne

Output


CUDA accelerated FAST-GICP

How to build

docker build . -f Dockerfile_fast_gicp_cuda -t slam:fast_gicp_cuda

How to run

xhost +local:docker
docker run -it --env DISPLAY=$DISPLAY --privileged --runtime nvidia --gpus all -v /kitti:/data -v /tmp/.X11-unix/:/tmp/.X11-unix:ro slam:fast_gicp_cuda

# Inside docker container
cd fast_gicp/
vim src/kitti.cpp (And edit code line 86-91 to select your algorihtm)
cd build
make -j
./gicp_kitti /data/sequences/00/velodyne

Output