In the study, the authors address the challenge of accurately modeling 3D geometries in monocular SLAM by introducing quadric representations. They propose a method that reshapes volumetric representations with quadric planes to improve accuracy and efficiency. The approach involves depth correction using quadric assumptions and a novel quadric-decomposed Transformer for mapping. By integrating quadrics into the pipeline, they achieve superior performance over other approaches relying on estimated depth.
The study evaluates the proposed method on various datasets, showcasing improvements in novel-view synthesis, depth estimation, and pose estimation tasks. The results demonstrate the effectiveness of Q-SLAM in enhancing tracking and mapping accuracy through quadric representations.
To Another Language
from source content
arxiv.org
Ключові висновки, отримані з
by Chensheng Pe... о arxiv.org 03-14-2024
https://arxiv.org/pdf/2403.08125.pdfГлибші Запити