BrackishMOT: The Brackish Multi-Object Tracking Dataset

Malte Pedersen; Daniel Lehotský; Ivan Adriyanov Nikolov; Thomas B. Moeslund

doi:10.1007/978-3-031-31435-3_2

BrackishMOT: The Brackish Multi-Object Tracking Dataset

Malte Pedersen, Daniel Lehotský, Ivan Adriyanov Nikolov, Thomas B. Moeslund

Research output: Contribution to book/anthology/report/conference proceeding › Article in proceeding › Research › peer-review

6 Citations (Scopus)

7 Downloads (Pure)

Abstract

There exist no publicly available annotated underwater multi-object tracking (MOT) datasets captured in turbid environments. To remedy this we propose the BrackishMOT dataset with focus on tracking schools of small fish, which is a notoriously difficultMOT task. BrackishMOT consists of 98 sequences captured in the wild. Alongside the novel dataset, we present baseline results by training a state-of-the-art tracker. Additionally, we propose a framework for creating
synthetic sequences in order to expand the dataset. The framework consists of animated fish models and realistic underwater environments. We analyse the effects of including synthetic data during training and show that a combination of real and synthetic underwater training data can enhance tracking performance. Links to code and data can be found at https://www.vap.aau.dk/brackishmot.

Original language	English
Title of host publication	Image Analysis : 22nd Scandinavian Conference, SCIA 2023, Sirkka, Finland, April 18–21, 2023, Proceedings, Part I.
Editors	Rikke Gade, Michael Felsberg, Joni-Kristian Kämäräinen
Number of pages	17
Publisher	Springer
Publication date	Apr 2023
Pages	17-33
ISBN (Print)	978-3-031-31434-6
ISBN (Electronic)	978-3-031-31435-3
DOIs	https://doi.org/10.1007/978-3-031-31435-3_2
Publication status	Published - Apr 2023
Event	Scandinavian Conference on Image Analysis - Levi, Finland Duration: 18 Apr 2023 → 21 Apr 2023

Conference

Conference	Scandinavian Conference on Image Analysis
Country/Territory	Finland
City	Levi
Period	18/04/2023 → 21/04/2023

Series	Lecture Notes in Computer Science
Volume	LNCS 13885
ISSN	0302-9743

Keywords

Dataset
Fish
Multi-Object Tracking
Synthetic data
Underwater

Access to Document

10.1007/978-3-031-31435-3_2

Accepted Author ManuscriptAccepted author manuscript, 6.55 MB

https://arxiv.org/pdf/2302.10645.pdf

AUB Link

Search for the material in Aalborg University Library's search engine

Cite this

Pedersen, Malte ; Lehotský, Daniel ; Nikolov, Ivan Adriyanov et al. / BrackishMOT : The Brackish Multi-Object Tracking Dataset. Image Analysis: 22nd Scandinavian Conference, SCIA 2023, Sirkka, Finland, April 18–21, 2023, Proceedings, Part I.. editor / Rikke Gade ; Michael Felsberg ; Joni-Kristian Kämäräinen. Springer, 2023. pp. 17-33 (Lecture Notes in Computer Science, Vol. LNCS 13885).

@inproceedings{a05663d2711e4c55be36de977a17bb6b,

title = "BrackishMOT: The Brackish Multi-Object Tracking Dataset",

abstract = "There exist no publicly available annotated underwater multi-object tracking (MOT) datasets captured in turbid environments. To remedy this we propose the BrackishMOT dataset with focus on tracking schools of small fish, which is a notoriously difficultMOT task. BrackishMOT consists of 98 sequences captured in the wild. Alongside the novel dataset, we present baseline results by training a state-of-the-art tracker. Additionally, we propose a framework for creatingsynthetic sequences in order to expand the dataset. The framework consists of animated fish models and realistic underwater environments. We analyse the effects of including synthetic data during training and show that a combination of real and synthetic underwater training data can enhance tracking performance. Links to code and data can be found at https://www.vap.aau.dk/brackishmot.",

keywords = "Dataset, Fish, Multi-Object Tracking, Synthetic data, Underwater",

author = "Malte Pedersen and Daniel Lehotsk{\'y} and Nikolov, {Ivan Adriyanov} and Moeslund, {Thomas B.}",

year = "2023",

month = apr,

doi = "10.1007/978-3-031-31435-3_2",

language = "English",

isbn = "978-3-031-31434-6",

series = "Lecture Notes in Computer Science",

publisher = "Springer",

pages = "17--33",

editor = "Rikke Gade and Michael Felsberg and Joni-Kristian K{\"a}m{\"a}r{\"a}inen",

booktitle = "Image Analysis",

address = "Germany",

note = "Scandinavian Conference on Image Analysis ; Conference date: 18-04-2023 Through 21-04-2023",

}

Pedersen, M , Lehotský, D , Nikolov, IA & Moeslund, TB 2023, BrackishMOT: The Brackish Multi-Object Tracking Dataset. in R Gade, M Felsberg & J-K Kämäräinen (eds), Image Analysis: 22nd Scandinavian Conference, SCIA 2023, Sirkka, Finland, April 18–21, 2023, Proceedings, Part I.. Springer, Lecture Notes in Computer Science, vol. LNCS 13885, pp. 17-33, Scandinavian Conference on Image Analysis, Levi, Finland, 18/04/2023. https://doi.org/10.1007/978-3-031-31435-3_2

BrackishMOT: The Brackish Multi-Object Tracking Dataset. / Pedersen, Malte ; Lehotský, Daniel ; Nikolov, Ivan Adriyanov et al.
Image Analysis: 22nd Scandinavian Conference, SCIA 2023, Sirkka, Finland, April 18–21, 2023, Proceedings, Part I.. ed. / Rikke Gade; Michael Felsberg; Joni-Kristian Kämäräinen. Springer, 2023. p. 17-33 (Lecture Notes in Computer Science, Vol. LNCS 13885).

Research output: Contribution to book/anthology/report/conference proceeding › Article in proceeding › Research › peer-review

TY - GEN

T1 - BrackishMOT

T2 - Scandinavian Conference on Image Analysis

AU - Pedersen, Malte

AU - Lehotský, Daniel

AU - Nikolov, Ivan Adriyanov

AU - Moeslund, Thomas B.

PY - 2023/4

Y1 - 2023/4

N2 - There exist no publicly available annotated underwater multi-object tracking (MOT) datasets captured in turbid environments. To remedy this we propose the BrackishMOT dataset with focus on tracking schools of small fish, which is a notoriously difficultMOT task. BrackishMOT consists of 98 sequences captured in the wild. Alongside the novel dataset, we present baseline results by training a state-of-the-art tracker. Additionally, we propose a framework for creatingsynthetic sequences in order to expand the dataset. The framework consists of animated fish models and realistic underwater environments. We analyse the effects of including synthetic data during training and show that a combination of real and synthetic underwater training data can enhance tracking performance. Links to code and data can be found at https://www.vap.aau.dk/brackishmot.

AB - There exist no publicly available annotated underwater multi-object tracking (MOT) datasets captured in turbid environments. To remedy this we propose the BrackishMOT dataset with focus on tracking schools of small fish, which is a notoriously difficultMOT task. BrackishMOT consists of 98 sequences captured in the wild. Alongside the novel dataset, we present baseline results by training a state-of-the-art tracker. Additionally, we propose a framework for creatingsynthetic sequences in order to expand the dataset. The framework consists of animated fish models and realistic underwater environments. We analyse the effects of including synthetic data during training and show that a combination of real and synthetic underwater training data can enhance tracking performance. Links to code and data can be found at https://www.vap.aau.dk/brackishmot.

KW - Dataset

KW - Fish

KW - Multi-Object Tracking

KW - Synthetic data

KW - Underwater

UR - http://www.scopus.com/inward/record.url?scp=85161468000&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-31435-3_2

DO - 10.1007/978-3-031-31435-3_2

M3 - Article in proceeding

SN - 978-3-031-31434-6

T3 - Lecture Notes in Computer Science

SP - 17

EP - 33

BT - Image Analysis

A2 - Gade, Rikke

A2 - Felsberg, Michael

A2 - Kämäräinen, Joni-Kristian

PB - Springer

Y2 - 18 April 2023 through 21 April 2023

ER -

Pedersen M , Lehotský D , Nikolov IA , Moeslund TB. BrackishMOT: The Brackish Multi-Object Tracking Dataset. In Gade R, Felsberg M, Kämäräinen JK, editors, Image Analysis: 22nd Scandinavian Conference, SCIA 2023, Sirkka, Finland, April 18–21, 2023, Proceedings, Part I.. Springer. 2023. p. 17-33. (Lecture Notes in Computer Science, Vol. LNCS 13885). doi: 10.1007/978-3-031-31435-3_2

BrackishMOT: The Brackish Multi-Object Tracking Dataset

Abstract

Conference

Keywords

Access to Document

AUB Link

Other files and links

Fingerprint

Marine Analytics using Computer Vision

BrackishMOT

Cite this

BrackishMOT: The Brackish Multi-Object Tracking Dataset

Abstract

Conference

Keywords

Access to Document

AUB Link

Other files and links

Fingerprint

Projects

Marine Analytics using Computer Vision

Datasets

BrackishMOT

Cite this