Refer to the Spot-Compose GitHub repository for details on Docker container setup.
Activate the open3d_gpu Conda environment:
conda activate open3d_gpuPrepare the data by running the following script:
python utils/prepare_om3_input.pyRun the following command to start the OpenMask3D Docker container:
docker run -p 5001:5001 --gpus all --name openmask3d_container -it craiden/openmask:v1.0Transfer the required data and configuration files to the container:
docker cp /home/cvg-robotics/project9_ws/SpotMap/data/data_om3 openmask3d_container:/home/openmask/resources
docker cp /home/cvg-robotics/project9_ws/SpotMap/configs/run_openmask3d_spotmap.sh openmask3d_container:/home/openmask/Execute the segmentation script within the container:
cd openmask
bash run_openmask3d_spotmap.shRetrieve the generated masks and features from the container to the local host:
docker cp openmask3d_container:/home/openmask/outputs/data_om3/mask_features/2025-06-23/experiment_0/clip_features.npy /home/cvg-robotics/project9_ws/SpotMap/data/data_om3/openmask3d/
docker cp openmask3d_container:/home/openmask/outputs/data_om3/masks/scene_MASKS.pt /home/cvg-robotics/project9_ws/SpotMap/data/data_om3/openmask3d/Activate the previously installed open3d_gpu Conda environment:
conda activate open3d_gpuEnsure the script is executable and run it:
chmod +x run.sh
bash run.shRun the interaction module with the following example command:
python scene_graph/update_scene_graph.py --pointcloud /home/cvg-robotics/project9_ws/SpotMap/scene_graph/scene.ply --labels_csv /home/cvg-robotics/project9_ws/SpotMap/scene_graph/color_label_mapping.csv --actions translate left 20 "White plastic bottle"