Towards Robust Contrail Detection by Mitigating Label Bias via a Probabilistic Deep Learning Model: A Preliminary Study

Yejun Lee; Eun-Kyeong Kim; Jaejun Yoo

doi:10.1145/3589132.3628364

Towards Robust Contrail Detection by Mitigating Label Bias via a Probabilistic Deep Learning Model: A Preliminary Study

Yejun Lee, Eun-Kyeong Kim, Jaejun Yoo

Urban Development & Mobility

Résultats de recherche: Le chapitre dans un livre, un rapport, une anthologie ou une collection › Conference contribution › Revue par des pairs

Résumé

Contrails, formed by jet flights, alter Earth's energy balance, prompting research into monitoring contrails and developing satellite-based automated contrail detection. This demand has advanced deep learning (DL)-based techniques. However, training DL algorithms to detect contrails has limitations: class imbalance, labeling difficulty, and a lack of reliable labeled datasets. We propose a probabilistic DL approach using P-UNet to alleviate label bias in contrail detection. By observing model outputs based on two labeled datasets, OpenContrails and MIT-Contrails, we found the probabilistic approach robust against potentially biased labels.

langue originale	Anglais
titre	SIGSPATIAL '23: 31st ACM International Conference on Advances in Geographic Information Systems
Sous-titre	Hamburg, Germany, November 13-16, 2023
rédacteurs en chef	Maria Luisa Damiani, Matthias Renz, Ahmed Eldawy, Peer Kröger, Mario A. Nascimento
Lieu de publication	New-York
Editeur	Association for Computing Machinery (ACM)
Pages	1-2
Nombre de pages	2
ISBN (imprimé)	979-8-4007-0168-9
Les DOIs	https://doi.org/10.1145/3589132.3628364
état	Publié - 22 déc. 2023

Accès au document

10.1145/3589132.3628364

Autres fichiers et liens

COVID-Contrail: Changes in US Contrail Outbreaks before and during the COVID-19 Pandemic
Kim, E. & Lee, Y.
1/09/22 → 31/08/24
Projet: Recherche

Contient cette citation

Lee, Y., Kim, E.-K., & Yoo, J. (2023). Towards Robust Contrail Detection by Mitigating Label Bias via a Probabilistic Deep Learning Model: A Preliminary Study. Dans M. L. Damiani, M. Renz, A. Eldawy, P. Kröger, & M. A. Nascimento (eds.), SIGSPATIAL '23: 31st ACM International Conference on Advances in Geographic Information Systems : Hamburg, Germany, November 13-16, 2023 (p. 1-2). Article 3 Association for Computing Machinery (ACM). https://doi.org/10.1145/3589132.3628364

Lee, Yejun ; Kim, Eun-Kyeong ; Yoo, Jaejun. / Towards Robust Contrail Detection by Mitigating Label Bias via a Probabilistic Deep Learning Model: A Preliminary Study. SIGSPATIAL '23: 31st ACM International Conference on Advances in Geographic Information Systems : Hamburg, Germany, November 13-16, 2023. Editeur / Maria Luisa Damiani ; Matthias Renz ; Ahmed Eldawy ; Peer Kröger ; Mario A. Nascimento. New-York : Association for Computing Machinery (ACM), 2023. p. 1-2

@inproceedings{58fce69121e841e594993f05efeafd8d,

title = "Towards Robust Contrail Detection by Mitigating Label Bias via a Probabilistic Deep Learning Model: A Preliminary Study",

abstract = "Contrails, formed by jet flights, alter Earth's energy balance, prompting research into monitoring contrails and developing satellite-based automated contrail detection. This demand has advanced deep learning (DL)-based techniques. However, training DL algorithms to detect contrails has limitations: class imbalance, labeling difficulty, and a lack of reliable labeled datasets. We propose a probabilistic DL approach using P-UNet to alleviate label bias in contrail detection. By observing model outputs based on two labeled datasets, OpenContrails and MIT-Contrails, we found the probabilistic approach robust against potentially biased labels.",

keywords = "Computing methodologies, Artificial intelligence, Mathematics of computing, Probability and statistics, Neural networks",

author = "Yejun Lee and Eun-Kyeong Kim and Jaejun Yoo",

year = "2023",

month = dec,

day = "22",

doi = "10.1145/3589132.3628364",

language = "English",

isbn = "979-8-4007-0168-9",

pages = "1--2",

editor = "Damiani, {Maria Luisa } and Renz, {Matthias } and Eldawy, {Ahmed } and Kr{\"o}ger, {Peer } and Nascimento, {Mario A. }",

booktitle = "SIGSPATIAL '23: 31st ACM International Conference on Advances in Geographic Information Systems",

publisher = "Association for Computing Machinery (ACM)",

address = "United States",

}

Lee, Y, Kim, E-K & Yoo, J 2023, Towards Robust Contrail Detection by Mitigating Label Bias via a Probabilistic Deep Learning Model: A Preliminary Study. Dans ML Damiani, M Renz, A Eldawy, P Kröger & MA Nascimento (eds), SIGSPATIAL '23: 31st ACM International Conference on Advances in Geographic Information Systems : Hamburg, Germany, November 13-16, 2023., 3, Association for Computing Machinery (ACM), New-York, p. 1-2. https://doi.org/10.1145/3589132.3628364

Towards Robust Contrail Detection by Mitigating Label Bias via a Probabilistic Deep Learning Model: A Preliminary Study. / Lee, Yejun; Kim, Eun-Kyeong; Yoo, Jaejun.
SIGSPATIAL '23: 31st ACM International Conference on Advances in Geographic Information Systems : Hamburg, Germany, November 13-16, 2023. Ed. / Maria Luisa Damiani; Matthias Renz; Ahmed Eldawy; Peer Kröger; Mario A. Nascimento. New-York: Association for Computing Machinery (ACM), 2023. p. 1-2 3.

Résultats de recherche: Le chapitre dans un livre, un rapport, une anthologie ou une collection › Conference contribution › Revue par des pairs

TY - GEN

T1 - Towards Robust Contrail Detection by Mitigating Label Bias via a Probabilistic Deep Learning Model: A Preliminary Study

AU - Lee, Yejun

AU - Kim, Eun-Kyeong

AU - Yoo, Jaejun

PY - 2023/12/22

Y1 - 2023/12/22

N2 - Contrails, formed by jet flights, alter Earth's energy balance, prompting research into monitoring contrails and developing satellite-based automated contrail detection. This demand has advanced deep learning (DL)-based techniques. However, training DL algorithms to detect contrails has limitations: class imbalance, labeling difficulty, and a lack of reliable labeled datasets. We propose a probabilistic DL approach using P-UNet to alleviate label bias in contrail detection. By observing model outputs based on two labeled datasets, OpenContrails and MIT-Contrails, we found the probabilistic approach robust against potentially biased labels.

AB - Contrails, formed by jet flights, alter Earth's energy balance, prompting research into monitoring contrails and developing satellite-based automated contrail detection. This demand has advanced deep learning (DL)-based techniques. However, training DL algorithms to detect contrails has limitations: class imbalance, labeling difficulty, and a lack of reliable labeled datasets. We propose a probabilistic DL approach using P-UNet to alleviate label bias in contrail detection. By observing model outputs based on two labeled datasets, OpenContrails and MIT-Contrails, we found the probabilistic approach robust against potentially biased labels.

KW - Computing methodologies

KW - Artificial intelligence

KW - Mathematics of computing

KW - Probability and statistics

KW - Neural networks

UR - https://doi.org/10.1145/3589132.3628364

UR - https://dl.acm.org/doi/epdf/10.1145/3589132

U2 - 10.1145/3589132.3628364

DO - 10.1145/3589132.3628364

M3 - Conference contribution

SN - 979-8-4007-0168-9

SP - 1

EP - 2

BT - SIGSPATIAL '23: 31st ACM International Conference on Advances in Geographic Information Systems

A2 - Damiani, Maria Luisa

A2 - Renz, Matthias

A2 - Eldawy, Ahmed

A2 - Kröger, Peer

A2 - Nascimento, Mario A.

PB - Association for Computing Machinery (ACM)

CY - New-York

ER -

Lee Y, Kim EK, Yoo J. Towards Robust Contrail Detection by Mitigating Label Bias via a Probabilistic Deep Learning Model: A Preliminary Study. Dans Damiani ML, Renz M, Eldawy A, Kröger P, Nascimento MA, rédacteurs en chef, SIGSPATIAL '23: 31st ACM International Conference on Advances in Geographic Information Systems : Hamburg, Germany, November 13-16, 2023. New-York: Association for Computing Machinery (ACM). 2023. p. 1-2. 3 doi: 10.1145/3589132.3628364

Towards Robust Contrail Detection by Mitigating Label Bias via a Probabilistic Deep Learning Model: A Preliminary Study

Résumé

Accès au document

Autres fichiers et liens

Projets

COVID-Contrail: Changes in US Contrail Outbreaks before and during the COVID-19 Pandemic

Contient cette citation