Monocular SLAM

Monocular SLAM (Simultaneous Localization and Mapping) aims to reconstruct 3D environments and simultaneously track a camera's position using only a single camera, a challenging problem due to inherent scale ambiguity. Current research focuses on improving accuracy and efficiency through novel approaches like Gaussian splatting, deep learning-based feature extraction and tracking (e.g., using deep keypoints or patch-based methods), and incorporating additional sensor data (e.g., inertial measurements or LiDAR). These advancements are driving progress in applications such as augmented reality, robotics, and medical imaging, particularly in scenarios where deploying multiple cameras or other sensors is impractical or impossible.