2020 IEEE International Symposium on Mixed and Augmented Reality (ISMAR) 2020
DOI: 10.1109/ismar50242.2020.00036
|View full text |Cite
|
Sign up to set email alerts
|

Learning Bipartite Graph Matching for Robust Visual Localization

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

0
4
0

Year Published

2021
2021
2024
2024

Publication Types

Select...
3
2

Relationship

0
5

Authors

Journals

citations
Cited by 5 publications
(4 citation statements)
references
References 36 publications
0
4
0
Order By: Relevance
“…[35] uses deep architecture to improve performance. SuperGlue [22], parallel to our prior work [23], designs a graph neural network for feature matching and uses self-attention and cross-attention to aggregate global context to achieve robust matching. [36] adopts a coarse-to-fine manner to accelerate feature matching.…”
Section: Related Workmentioning
confidence: 99%
See 3 more Smart Citations
“…[35] uses deep architecture to improve performance. SuperGlue [22], parallel to our prior work [23], designs a graph neural network for feature matching and uses self-attention and cross-attention to aggregate global context to achieve robust matching. [36] adopts a coarse-to-fine manner to accelerate feature matching.…”
Section: Related Workmentioning
confidence: 99%
“…Negative Sample Mining. Different from the previous work [23] where all negative edges are generated in a completely random manner, we instead perform kNN ratio matching online with a fixed number of randomly selected features from both 2D and 3D feature sets to generate positive and negative edge samples during training. The negative samples generated in this way tend to be harder than that of random selection while being consistent with the inference stage in terms of data distribution.…”
Section: Learning From Sfm Modelmentioning
confidence: 99%
See 2 more Smart Citations