OSM-BKI

Getting Started

1. Download Example Data

pip install gdown
python example_data/download_data.py --output-dir /mnt/datasets

Use --output-dir (or -o) to specify where data is extracted. Defaults to example_data/ if omitted.

2. Run OSM-BKI (ROS2 Humble)

./docker/run.sh osm-bki /mnt/datasets

Once inside the container, the workspace is built automatically. Launch the example:

ros2 launch osm_bki mcd_example_launch.py

To skip the C++ build (e.g. if already built):

./docker/run.sh osm-bki /mnt/datasets --skip-build

3. Run Baselines (ROS1 Noetic)

Baseline methods (SEE-CSOM, EBS) have their own Docker setup under baselines/docker/:

cd baselines
./docker/run.sh see-csom /mnt/datasets
./docker/run.sh ebs /mnt/datasets

Once inside, the workspace is patched and built automatically. For SEE-CSOM, run the toy example:

roslaunch see_csom toy_example_node.launch

To skip the build step:

./docker/run.sh see-csom /mnt/datasets --skip-build

Docker Script Usage

OSM-BKI (docker/run.sh):

./docker/run.sh <container> [data_dir] [--skip-build]

Containers:
  osm-bki    - ROS2 Humble (OSM-BKI). Requires data_dir.

Options:
  --skip-build  Skip building C++ code inside the container

Baselines (baselines/docker/run.sh):

./docker/run.sh <baseline> [data_dir] [--skip-build]

Baselines:
  see-csom   - ROS1 Noetic (SEE-CSOM)
  ebs        - ROS1 Noetic (EBS) with CUDA + PyTorch

Options:
  --skip-build  Skip building code inside the container

Citation

OSM-BKI is an ongoing research project, but the results have been exiting and we want to extend that excitement to all who are interested. If would like to reference our project in your work, you can use the bibtex below:

Acknowledgment

We would like to extend our graditude to Professor Lu Gan for her coadvisement throughout this project and for the foundation on which this work has been built. To cite her original work, you can use the bibtex below:

@ARTICLE{gan2019bayesian,
author={L. {Gan} and R. {Zhang} and J. W. {Grizzle} and R. M. {Eustice} and M. {Ghaffari}},
journal={IEEE Robotics and Automation Letters},
title={Bayesian Spatial Kernel Smoothing for Scalable Dense Semantic Mapping},
year={2020},
volume={5},
number={2},
pages={790-797},
keywords={Mapping;semantic scene understanding;range sensing;RGB-D perception},
doi={10.1109/LRA.2020.2965390},
ISSN={2377-3774},
month={April},}

as well as visit the repository for S-BKI:

https://github.com/ganlumomo/BKISemanticMapping