Automatically generated two-stage polynomial solvers have gained popularity in computer vision for solving certain geometric problems, because they offer a computationally efficient framework especially when the same solver is used repeatedly in problems, where only the coefficients of the polynomials vary. In that case, a significant part of the computation is performed only once in the offline stage, while the online stage mostly consists of numerical matrix operations. In this project, a novel method for optimal implementation of the solvers as well as a new solver generation strategy are developed by considering the computation required and the stability of the solution. In addition, potential new applications are investigated in computer vision and in some other fields. The aim of the project is also popularizing the use of automatically generated polynomial solvers in particular among those researchers and experts who are not familiar with the methods of algebraic geometry.

2023

2027