stella-cv/stella_vslam

stella_vslam – A Versatile Visual SLAM Framework

What it is

• An open‑source library that estimates a camera’s pose and builds a 3‑D map from video streams. It works with monocular, stereo, and RGB‑D cameras.
• It is a community‑maintained fork of the original OpenVSLAM project, rewritten from scratch but keeping the same high‑level design.

Why it matters

• Visual SLAM (Simultaneous Localization and Mapping) is a core technology for robotics, AR/VR, autonomous drones, and any system that needs to understand its surroundings using only cameras.
• stella_vslam is deliberately modular and supports many camera models (perspective, fisheye, equirectangular, etc.), making it easy to adapt to unconventional optics such as 360° cameras.

Key features (as described in the README)

• Multi‑camera support – interchangeable camera models; users can add new ones (e.g., dual‑fisheye, catadioptric) with minimal code changes.
• Map persistence – built maps can be saved to disk and later re‑loaded for fast relocalization of new image streams.
• Modular architecture – core functions are split into well‑defined components with clean APIs, facilitating experimentation and integration.
• Example code & tutorials – a set of *.cc snippets, a “Simple Tutorial”, and full documentation help newcomers get a working system quickly.
• ROS wrapper – a separate repository (stella_vslam_ros) provides ROS integration for robot developers.
• Ongoing development – roadmap includes refactoring, algorithmic tweaks, marker integration, extra camera models, Python bindings, and IMU fusion.

Typical use cases

• Building a map of an indoor or outdoor environment from a handheld or robot‑mounted camera.
• Real‑time pose tracking for AR/VR headsets or handheld devices.
• Relocalizing a robot in a previously mapped area using only visual input.
• Research on SLAM algorithms, camera model extensions, or sensor fusion (future IMU support).

Getting started

1. Installation – Follow the step‑by‑step guide in the online docs (https://stella‑cv.readthedocs.io/en/latest/installation.html). Pre‑built Docker images are also provided.
2. Run the examples – The example/ directory contains ready‑to‑compile programs; the docs walk you through a “Simple Tutorial”.
3. Data – Download a sample ORB vocabulary file and example datasets (links in the docs) to test the system.
4. ROS integration – Clone stella_vslam_ros if you need ROS topics/services.

License

• Distributed under the permissive 2‑clause BSD license. Some bundled third‑party code uses MIT, BSD‑3, or LGPLv3+ (e.g., g2o).

Project health

• Actively maintained (as of the latest README) with a public issue tracker and GitHub Discussions for support.
• The developers are focusing on refactoring, adding tests, and expanding functionality (e.g., Python bindings, IMU integration).

Citation If you use the library in academic work, cite the original OpenVSLAM paper (MM ’19) as shown in the README.

---

All information above is taken directly from the repository’s README and linked documentation.