openvslam

Supports monocular, stereo, and RGB-D cameras through a unified codebase, enabling adaptation to diverse visual sensor setups without major code changes.

Allows saving and reloading 3D maps for persistent localization across sessions, which is crucial for applications like AR/VR and robotics that need long-term spatial memory.

Features a modular architecture that lets users easily replace components like feature extractors and camera models, facilitating research and customization for specific needs.

Compatible with the Robot Operating System, streamlining deployment in robotics applications and enabling seamless integration with existing ROS-based pipelines.