|Description:||Minimal solvers for camera geometry compute geometric relations between images (relative camera pose) and between an image and a scene (absolute camera pose). Minimal solvers are a core component of many 3D computer vision algorithms, such as Structure-from-Motion, Simultaneous Localization and Mapping, and Visual Localization. As such, they are used in applications such as self-driving cars and Augmented Reality.
Camera geometry problems lead to systems of polynomial equations that need to be solved efficiently. Recent advances in solving polynomial systems allow to (semi-)automatically generate such solvers. The goal of this project is to revisit existing camera geometry problems and use state-of-the-art methods for generating efficient solvers for these problems. This includes tuning the solvers, evaluating their efficiency, numerical stability and robustness, as well as cataloging the resulting solvers together with their source code (e.g., on a website). Such a catalogue has the potential to be used by researchers and practitioners alike to search for the best available implementation for camera geometry problems.