This paper presents ORB-SLAM3, an open-source library for visual, visual-inertial, and multi-map SLAM using monocular, stereo, and RGB-D cameras with pin-hole and fisheye models. The system introduces a tightly-integrated visual-inertial SLAM based on Maximum-a-Posteriori (MAP) estimation, achieving robust real-time performance with two to ten times higher accuracy than previous methods. A multi-map system with improved place recognition enables long-term data association, allowing seamless map merging and relocalization even in poor visual conditions. ORB-SLAM3 also supports long-term data association through place recognition, enabling accurate map merging and relocalization. The system is evaluated on various datasets, showing superior accuracy compared to existing systems, particularly in challenging scenarios like the TUM VI benchmark. ORB-SLAM3 is also compared with other systems, demonstrating its robustness and accuracy. The paper also discusses the system's architecture, including tracking, mapping, and map merging, and presents experimental results showing its effectiveness in various scenarios. The system is open-source and available for public use.This paper presents ORB-SLAM3, an open-source library for visual, visual-inertial, and multi-map SLAM using monocular, stereo, and RGB-D cameras with pin-hole and fisheye models. The system introduces a tightly-integrated visual-inertial SLAM based on Maximum-a-Posteriori (MAP) estimation, achieving robust real-time performance with two to ten times higher accuracy than previous methods. A multi-map system with improved place recognition enables long-term data association, allowing seamless map merging and relocalization even in poor visual conditions. ORB-SLAM3 also supports long-term data association through place recognition, enabling accurate map merging and relocalization. The system is evaluated on various datasets, showing superior accuracy compared to existing systems, particularly in challenging scenarios like the TUM VI benchmark. ORB-SLAM3 is also compared with other systems, demonstrating its robustness and accuracy. The paper also discusses the system's architecture, including tracking, mapping, and map merging, and presents experimental results showing its effectiveness in various scenarios. The system is open-source and available for public use.