Graph-CoVis: GNN-Based Multi-View Panorama Global Pose Estimation

Negar Nejatishahidin, Will Hutchcroft, Manjunath Narayana, Ivaylo Boyadzhiev, Yuguang Li, Naji Khosravan, Jana Košecká, Sing Bing Kang; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2023, pp. 6459-6468

Abstract

In this paper, we address the problem of wide-baseline camera pose estimation from a group of 360deg panoramas under upright-camera assumption. Recent work has demonstrated the merit of deep-learning for end-to-end direct relative pose regression in 360deg panorama pairs. To exploit the benefits of multi-view logic in a learning-based framework, we introduce Graph-CoVis, which nontrivially extends CoVisPose from relative two-view to global multi-view spherical camera pose estimation. Graph-CoVis is a novel Graph Neural Network based architecture that jointly learns the co-visible structure and global motion in an end-to-end and fully-supervised approach. Using the ZInD dataset, which features real homes presenting wide-baselines, occlusion, and limited visual overlap, we show that our model performs competitively to state-of-the-art approaches

Related Material

[pdf] [supp]
[bibtex]
@InProceedings{Nejatishahidin_2023_CVPR, author = {Nejatishahidin, Negar and Hutchcroft, Will and Narayana, Manjunath and Boyadzhiev, Ivaylo and Li, Yuguang and Khosravan, Naji and Ko\v{s}eck\'a, Jana and Kang, Sing Bing}, title = {Graph-CoVis: GNN-Based Multi-View Panorama Global Pose Estimation}, booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops}, month = {June}, year = {2023}, pages = {6459-6468} }