Skip to content

slamlab-offline-importer

Converts the sensor data of a ROS 2 bag into the input files of slamlab-offline: point clouds, IMU and GNSS measurements, reference trajectories and images. It reads the bag directly and needs no ROS installation.

Terminal window
slamlab-offline-importer <config_filepath>

The single required argument is the path to a YAML configuration file. The executable reads the slamlab_offline_importer section.

Terminal window
slamlab-offline-importer --check <config_filepath>

checks the configuration against the bag (are the topics in the bag, do they have a supported type) without converting anything. --help prints the usage, --version the version.

  • Storage: MCAP (.mcap) and SQLite (.db3) bags, as single files or as directories of files, including bags split into several files (read in the order of their index, bag_2 before bag_10). metadata.yaml is not needed: the topics, their types and the time range come from the bag files. MCAP files without summary section (e.g. of interrupted recordings) are read too; listing their topics takes an extra pass over the file.
  • Messages:
SectionMessage types
cloudssensor_msgs/PointCloud2, point_cloud_interfaces/CompressedPointCloud2 (Cloudini)
imusensor_msgs/Imu
gnsssensor_msgs/NavSatFix, gps_msgs/GPSFix
trajectoriesnav_msgs/Odometry
imagessensor_msgs/Image (mono8, rgb8, bgr8, rgba8, bgra8), sensor_msgs/CompressedImage (JPEG, PNG, H.264, H.265)

The bag is read once for all topics; point clouds and images are converted in parallel.

All files are written below output_dirpath. A topic is named by its name without the leading / and with - for /, e.g. /ouster/points becomes ouster-points.

SectionFiles
cloudsclouds/<topic>/<sec>.<nsec>.laz: one LAZ file per message (see point cloud file formats), named after the header stamp. Points with the integer attributes row, col and intensity and a stamp per point, sorted by stamp.
imuimu/<topic>.csv: sec,nsec,angular_velocity_x,angular_velocity_y,angular_velocity_z,linear_acceleration_x,linear_acceleration_y,linear_acceleration_z
gnssgnss/<topic>.csv: sec,nsec,x,y,z,cov_00,...,cov_22 with UTM coordinates plus the shift; gnss/<topic>.shift: the applied shift and the UTM zone
trajectoriestrajectory/<topic>.txt: TUM (timestamp tx ty tz qx qy qz qw) or KITTI
imagesimages/<topic>/<sec>.<nsec>.jpg

The clouds directory and the imu and gnss files are the inputs clouds_dirpath, imu_measurements_filepath and absolute_position_measurements_filepath of slamlab_offline.

The importer logs the fields and the point stamps of the first cloud of each topic, e.g. point stamps from +0.000 ms to +99.954 ms relative to the header. Check them to see if the point stamps are configured right.