Multi-Session SLAM with Differentiable Wide-Baseline Pose Optimization
Abstract
We introduce a new system for Multi-Session SLAM which tracks camera motion across multiple disjoint videos under a single global reference. Our approach couples the prediction of optical flow with solver layers to estimate camera pose. The backbone is trained end-to-end using a novel differentiable solver for wide-baseline two-view pose. The full system can connect disjoint sequences perform visual odometry and global optimization. Compared to existing approaches our design is accurate and robust to catastrophic failures.
Cite
Text
Lipson and Deng. "Multi-Session SLAM with Differentiable Wide-Baseline Pose Optimization." Conference on Computer Vision and Pattern Recognition, 2024. doi:10.1109/CVPR52733.2024.01856Markdown
[Lipson and Deng. "Multi-Session SLAM with Differentiable Wide-Baseline Pose Optimization." Conference on Computer Vision and Pattern Recognition, 2024.](https://mlanthology.org/cvpr/2024/lipson2024cvpr-multisession/) doi:10.1109/CVPR52733.2024.01856BibTeX
@inproceedings{lipson2024cvpr-multisession,
title = {{Multi-Session SLAM with Differentiable Wide-Baseline Pose Optimization}},
author = {Lipson, Lahav and Deng, Jia},
booktitle = {Conference on Computer Vision and Pattern Recognition},
year = {2024},
pages = {19626-19635},
doi = {10.1109/CVPR52733.2024.01856},
url = {https://mlanthology.org/cvpr/2024/lipson2024cvpr-multisession/}
}