Differences
This shows you the differences between two versions of the page.
Both sides previous revision Previous revision Next revision | Previous revision Next revision Both sides next revision | ||
members:stumberg [2020/06/02 13:05] Lukas von Stumberg |
members:stumberg [2020/12/12 15:47] Lukas von Stumberg |
||
---|---|---|---|
Line 6: | Line 6: | ||
Find me on [[https:// | Find me on [[https:// | ||
- | ===== Research Interests | + | ==== Research Interests ==== |
My research interests are visual and visual-inertial SLAM, robotics, and 3D reconstruction. I like to apply both traditional methods and deep learning. | My research interests are visual and visual-inertial SLAM, robotics, and 3D reconstruction. I like to apply both traditional methods and deep learning. | ||
- | ===== Selected Projects ===== | + | ==== Selected Projects ==== |
+ | === LM-Reloc: Levenberg-Marquardt Based Direct Visual Relocalization === | ||
+ | As a follow-up of our previous work GN-Net, LM-Reloc boosts the performance of direct image alignment for relocalization using a novel loss formulation, | ||
+ | < | ||
+ | <iframe width=" | ||
+ | </ | ||
=== GN-Net: The Gauss-Newton Loss for Multi-Weather Relocalization === | === GN-Net: The Gauss-Newton Loss for Multi-Weather Relocalization === | ||
Using our novel Gauss-Newton loss we improve the robustness of direct methods against strong illumination changes and bad initializations. This allows us to accurately relocalize between different weathers. [[: | Using our novel Gauss-Newton loss we improve the robustness of direct methods against strong illumination changes and bad initializations. This allows us to accurately relocalize between different weathers. [[: |