2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) 2020
DOI: 10.1109/cvprw50498.2020.00515
|View full text |Cite
|
Sign up to set email alerts
|

Match or No Match: Keypoint Filtering based on Matching Probability

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
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 30 publications
0
4
0
Order By: Relevance
“…Papadaki and Hansch [22] introduced an algorithm that predicts the probability that each point will be successfully matched and filters the detected keypoints before attempting to match. Key.Net [23] proposed a hybrid method to detect key points by combining manual and learned CNN filters in a shallow multi-scale architecture.…”
Section: Other Methodsmentioning
confidence: 99%
“…Papadaki and Hansch [22] introduced an algorithm that predicts the probability that each point will be successfully matched and filters the detected keypoints before attempting to match. Key.Net [23] proposed a hybrid method to detect key points by combining manual and learned CNN filters in a shallow multi-scale architecture.…”
Section: Other Methodsmentioning
confidence: 99%
“…In the literature, there are many works where authors use various methods to get rid of keypoints and their descriptors that do not carry useful information. [8][9][10] employ filtration to address the issue of repetitive patterns when different keypoints possess quite similar descriptors due to periodicity or object similarity within the image. Very close descriptors corresponding to different keypoints can lead to mismatches during the descriptor matching process.…”
Section: Related Workmentioning
confidence: 99%
“…In the subsequent stage, some confusion index is computed for all remaining keypoints, based on which a decision is made whether to retain or delete the point. [10] train Random Forest to classify descriptors into matchable/non-matchable in tasks where images of building facades act as data. Such images usually contain repetitive patterns in the architecture of buildings and plants.…”
Section: Related Workmentioning
confidence: 99%
“…Issues to be considered are also the exploitation of more sophisticated key point detectors, based on machine and deep learning, to avoid matches that are resulting in highly noisy sparse 3D point clouds. This also concerns the underwater imagery being aected by caustics where a solution could be a way to learn the detectors to avoid caustics' pixels similar to [Papadaki 2020]. However, this would not improve the MVS step.…”
Section: Future Workmentioning
confidence: 99%