roverp: data processing

The data processing tooling for red-rover includes some library components, a number of CLI tools, and a makefile for managing a dockerized cartographer SLAM pipeline.

Tool	Description	Required Extras
roverp anonymize	Anonymize camera data by blurring faces.	anonymize
roverp sensorpose	Get interpolated poses for a specific sensor from a cartographer output.
roverp report	Generate a "speed report" of a dataset.
roverp segment	Run semantic segmentation on collected video data.	segment
make trajectory	Compute poses using cartographer.
make lidar	Compute poses and output globally-aligned point clouds.
make info	Alias for rosbag info.
make validate	Alias for cartographer_rosbag_validate.

Setup

uv + localpip + github

git clone git@github.com:RadarML/red-rover.git
cd roverp
uv sync

pip install "roverp@git+ssh://git@github.com/RadarML/red-rover.git#subdirectory=processing"

Warning

roverp includes a number of extras, many of which are fairly heavy, and most of which you probably don't need. Avoid installing unnecessary dependencies by only installing the required extras and definitely don't uv sync --all-extras!.

Info

If you need to run cartographer SLAM, see the dockerized cartographer setup instructions.