SpotMap

Kerem Kılıç¹, Vikram Iyer¹, Roham Zendehdel Nobari¹, Yagiz Devre^2
1ETH Zürich, , ²Princeton

---

SpotMap is a complete end-to-end pipeline for robotic scene understanding and interactive mapping. It integrates on-board SLAM, volumetric 3D reconstruction, instance segmentation, dynamic scene graph generation, and affordance prediction — giving mobile robots the capability to build detailed maps and understand their surroundings with minimal human supervision.

---

📂 Repository Structure

SpotMap/
├── src/          # Core source code: modules of the pipeline
├── configs/      # Configuration files for SLAM and OpenMask3D
├── data/         # Original and intermediate datasets
├── scene_graph/  # Current scene graph representation
├── docs/         # Documentation, figures, and additional resources

📂 Source code

src/
├── data_acquisition/                
├── mapping_and_reconstruction/
├── segmentation_and_scene_graph/   
└── re_scan/    

---

🚀 Setup

Each of the core modules in src/ has its own README with all the installation, configuration, and usage instructions:

• Data Acquisition
  See src/data_acquisition/README.md for how to extract and preprocess raw RGB-D frames and ROS bag files.

• Mapping & Reconstruction
  See src/mapping_and_reconstruction/README.md for how to configure and run SLAM, TSDF fusion, denoising, and mesh/point-cloud export.

• Segmentation & Scene Graph
  See src/segmentation_and_scene_graph/README.md for how to perform semantic segmentation, build the scene graph, and visualize the results.

  📂 Dataset

  A sample dataset captured by Spot is already available at src/data_acquisition/ros2/rgbd_dataset. This dataset can be used to test and validate the pipeline components.

📄 Citation

If you find this work useful, please cite our paper:

WORK IN PROGRESS

---

⭐ Support

If you find SpotMAP helpful, please ⭐ star our GitHub repository to support the project!

📧 Contact

For questions or issues, please open a GitHub Issue

---

📜 License

This project is licensed under the MIT License — see LICENSE for details.