Re-Identification of Giant Sunfish using Keypoint Matching

Malte Pedersen; Joakim Bruslund Haurum; Thomas B. Moeslund; Marianne Nyegaard

doi:10.7557/18.6234

Re-Identification of Giant Sunfish using Keypoint Matching

Malte Pedersen^*, Joakim Bruslund Haurum, Thomas B. Moeslund, Marianne Nyegaard

^*Kontaktforfatter

Publikation: Bidrag til tidsskrift › Konferenceartikel i tidsskrift › Forskning › peer review

74 Downloads (Pure)

Abstract

We present the first work where re-identification ofthe Giant Sunfish (Mola alexandrini) is automated using computer vision and deep learning. We propose a pipeline that scores an mAP of 60.34% on a full rank of the novel TinyMola dataset which includes 31 IDs and 91 images. The method requires no domain-adaptation or training which makes it especially suited for low-budget or volunteer-based projects, like Match My Mola, as part of a human-in-the-loop model.

The pipeline includes segmentation, keypoint detection and description, keypoint matching, and ranking. The choice of feature descriptor has the largest impact on the performance and we show that the deep learning based SuperPoint descriptor greatly outperforms handcrafted descriptors like SIFT and RootSIFT independent of the segmentation level and matching method. Combining SuperPoint and the graph neural network based SuperGlue matching method produces the best results.

Originalsprog	Engelsk
Tidsskrift	Proceedings of the Northern Lights Deep Learning Workshop
Vol/bind	3
Antal sider	9
ISSN	2703-6928
DOI	https://doi.org/10.7557/18.6234
Status	Udgivet - 28 mar. 2022
Begivenhed	Northern Lights Deep Learning Workshop - Tromsø, Norge Varighed: 10 jan. 2022 → 12 jan. 2022 https://www.nldl.org/

Workshop

Workshop	Northern Lights Deep Learning Workshop
Land/Område	Norge
By	Tromsø
Periode	10/01/2022 → 12/01/2022
Internetadresse	https://www.nldl.org/

FN’s Verdensmål

Denne publikation bidrager til følgende verdensmål

Adgang til dokumentet

10.7557/18.6234Licens: CC BY 4.0

Open Access articleForlagets udgivne version, 3,23 MBLicens: CC BY 4.0

AUB Link

Søg efter materialet i Aalborg Universitetsbiblioteks søgemaskine

Citationsformater

@inproceedings{0ddf9b8a21c34e33a1e4530a557b3d6b,

title = "Re-Identification of Giant Sunfish using Keypoint Matching",

abstract = "We present the first work where re-identification ofthe Giant Sunfish (Mola alexandrini) is automated using computer vision and deep learning. We propose a pipeline that scores an mAP of 60.34% on a full rank of the novel TinyMola dataset which includes 31 IDs and 91 images. The method requires no domain-adaptation or training which makes it especially suited for low-budget or volunteer-based projects, like Match My Mola, as part of a human-in-the-loop model.The pipeline includes segmentation, keypoint detection and description, keypoint matching, and ranking. The choice of feature descriptor has the largest impact on the performance and we show that the deep learning based SuperPoint descriptor greatly outperforms handcrafted descriptors like SIFT and RootSIFT independent of the segmentation level and matching method. Combining SuperPoint and the graph neural network based SuperGlue matching method produces the best results.",

author = "Malte Pedersen and Haurum, {Joakim Bruslund} and Moeslund, {Thomas B.} and Marianne Nyegaard",

note = "This work has been funded by the Independent ResearchFund Denmark under case number 9131-00128B.; Northern Lights Deep Learning Workshop ; Conference date: 10-01-2022 Through 12-01-2022",

year = "2022",

month = mar,

day = "28",

doi = "10.7557/18.6234",

language = "English",

volume = "3",

journal = "Proceedings of the Northern Lights Deep Learning Workshop",

issn = "2703-6928",

url = "https://www.nldl.org/",

}

TY - GEN

T1 - Re-Identification of Giant Sunfish using Keypoint Matching

AU - Pedersen, Malte

AU - Haurum, Joakim Bruslund

AU - Moeslund, Thomas B.

AU - Nyegaard, Marianne

N1 - This work has been funded by the Independent ResearchFund Denmark under case number 9131-00128B.

PY - 2022/3/28

Y1 - 2022/3/28

N2 - We present the first work where re-identification ofthe Giant Sunfish (Mola alexandrini) is automated using computer vision and deep learning. We propose a pipeline that scores an mAP of 60.34% on a full rank of the novel TinyMola dataset which includes 31 IDs and 91 images. The method requires no domain-adaptation or training which makes it especially suited for low-budget or volunteer-based projects, like Match My Mola, as part of a human-in-the-loop model.The pipeline includes segmentation, keypoint detection and description, keypoint matching, and ranking. The choice of feature descriptor has the largest impact on the performance and we show that the deep learning based SuperPoint descriptor greatly outperforms handcrafted descriptors like SIFT and RootSIFT independent of the segmentation level and matching method. Combining SuperPoint and the graph neural network based SuperGlue matching method produces the best results.

AB - We present the first work where re-identification ofthe Giant Sunfish (Mola alexandrini) is automated using computer vision and deep learning. We propose a pipeline that scores an mAP of 60.34% on a full rank of the novel TinyMola dataset which includes 31 IDs and 91 images. The method requires no domain-adaptation or training which makes it especially suited for low-budget or volunteer-based projects, like Match My Mola, as part of a human-in-the-loop model.The pipeline includes segmentation, keypoint detection and description, keypoint matching, and ranking. The choice of feature descriptor has the largest impact on the performance and we show that the deep learning based SuperPoint descriptor greatly outperforms handcrafted descriptors like SIFT and RootSIFT independent of the segmentation level and matching method. Combining SuperPoint and the graph neural network based SuperGlue matching method produces the best results.

U2 - 10.7557/18.6234

DO - 10.7557/18.6234

M3 - Conference article in Journal

SN - 2703-6928

VL - 3

JO - Proceedings of the Northern Lights Deep Learning Workshop

JF - Proceedings of the Northern Lights Deep Learning Workshop

T2 - Northern Lights Deep Learning Workshop

Y2 - 10 January 2022 through 12 January 2022

ER -

Re-Identification of Giant Sunfish using Keypoint Matching

Abstract

Workshop

FN’s Verdensmål

Adgang til dokumentet

AUB Link

Fingeraftryk

Citationsformater