Hint

Please change branch to Bunker-DVI-Dataset-reg-1 for quick experiment.

Example Dataset:

Download the dataset from Bunker DVI Dataset

benchmark-FORM-to-HDMapping

Runs the FORM LiDAR odometry algorithm on a ROS bag file and converts the output to an HDMapping session.

FORM (Fixed-Lag Odometry with Reparative Mapping) is a real-time LiDAR odometry system from Carnegie Mellon University's Robotics Perception Lab (2025).

Prerequisites

• Docker
• A ROS bag containing a /velodyne_points topic (sensor_msgs/PointCloud2) — ideally with a ring field (Velodyne driver standard)

Step 1 — Clone with submodules

git clone https://github.com/MapsHD/benchmark-FORM-to-HDMapping.git --recursive
cd benchmark-FORM-to-HDMapping

Note: Internet access is required at docker build time because FORM fetches the tsl::robin_map dependency via CMake FetchContent.

Step 2 — Build the Docker image

docker build -t form_noetic .

This installs:

• Ubuntu 20.04 + ROS Noetic
• GTSAM 4.2 (from the borglab PPA)
• Intel TBB, Eigen3, PCL
• FORM C++ library (compiled from submodule)
• catkin workspace with form_ros_node and form-to-hdmapping

The build takes several minutes on first run.

Step 3 — Run the pipeline

chmod +x docker_session_run-ros1-form.sh
./docker_session_run-ros1-form.sh /path/to/input.bag /path/to/output/dir

Or with no arguments to use a GUI file selector (requires zenity):

./docker_session_run-ros1-form.sh

What happens:

The script opens a Docker container with a tmux session containing four panes:

Pane	Role
0	roscore
1	form_ros_node — reads /velodyne_points, publishes /form/odometry + /form/registered_cloud
2	rosbag record — captures the two published topics
3	rosbag play — plays your input bag with simulated clock

After playback completes, recording is stopped and a second Docker run converts recorded-form.bag into the HDMapping session format.

Step 4 — Open in HDMapping

Output files appear in <output_dir>/output_hdmapping-form/:

lio_initial_poses.reg
poses.reg
scan_lio_0.laz
scan_lio_1.laz
...
session.json
trajectory_lio_0.csv
trajectory_lio_1.csv
...

Open session.json with the multi_view_tls_registration_step_2 application.

Notes on point cloud format

The form_ros_node handles three input cloud types automatically:

Cloud type	Handling
Organized + ring field	Ring field used as row index (best accuracy, Velodyne standard)
Organized, no ring field	Row index = point_index / width
Unorganized (height == 1)	Elevation-angle bucketing assigns synthetic rings

Adjust num_rows and num_columns in src/form-ros-node/launch/form.launch to match your sensor (default: 64 rows × 1024 columns for Velodyne HDL-64E).

Sensor geometry presets

Sensor	num_rows	num_columns
Velodyne HDL-64E	64	1024
Velodyne VLP-16	16	1800
Ouster OS1-64	64	1024
Ouster OS0-128	128	1024

Contact

januszbedkowski@gmail.com