Home > Research > Publications & Outputs > Boundary Aware U-Net for Glacier Segmentation

Research

Links

Text available via DOI:

View graph of relations

Boundary Aware U-Net for Glacier Segmentation

Research output: Contribution to Journal/Magazine › Journal article › peer-review

Published

Standard

Boundary Aware U-Net for Glacier Segmentation. / Aryal, Bibek; Miles, Katie E.; Zesati, Sergio A. Vargas et al.
In: Proceedings of the Northern Lights Deep Learning Workshop, Vol. 4, 23.01.2023.

Research output: Contribution to Journal/Magazine › Journal article › peer-review

Harvard

Aryal, B, Miles, KE, Zesati, SAV & Fuentes, O 2023, 'Boundary Aware U-Net for Glacier Segmentation', Proceedings of the Northern Lights Deep Learning Workshop, vol. 4. https://doi.org/10.7557/18.6789

APA

Aryal, B., Miles, K. E., Zesati, S. A. V., & Fuentes, O. (2023). Boundary Aware U-Net for Glacier Segmentation. Proceedings of the Northern Lights Deep Learning Workshop, 4. https://doi.org/10.7557/18.6789

Vancouver

Aryal B, Miles KE, Zesati SAV, Fuentes O. Boundary Aware U-Net for Glacier Segmentation. Proceedings of the Northern Lights Deep Learning Workshop. 2023 Jan 23;4. doi: 10.7557/18.6789

Author

Aryal, Bibek ; Miles, Katie E. ; Zesati, Sergio A. Vargas et al. / Boundary Aware U-Net for Glacier Segmentation. In: Proceedings of the Northern Lights Deep Learning Workshop. 2023 ; Vol. 4.

Bibtex

@article{603579bdbc034f96ae131dadf9834058,
title = "Boundary Aware U-Net for Glacier Segmentation",
abstract = "Large-scale study of glaciers improves our understanding of global glacier change and is imperative for monitoring the ecological environment, preventing disasters, and studying the effects of global climate change. Glaciers in the Hindu Kush Himalaya (HKH) are particularly interesting as the HKH is one of the world's most sensitive regions for climate change. In this work, we: (1) propose a modified version of the U-Net for large-scale, spatially non-overlapping, clean glacial ice, and debris-covered glacial ice segmentation; (2) introduce a novel self-learning boundary-aware loss to improve debris-covered glacial ice segmentation performance; and (3) propose a feature-wise saliency score to understand the contribution of each feature in the multispectral Landsat 7 imagery for glacier mapping. Our results show that the debris-covered glacial ice segmentation model trained using self-learning boundary-aware loss outperformed the model trained using dice loss. Furthermore, we conclude that red, shortwave infrared, and near-infrared bands have the highest contribution toward debris-covered glacial ice segmentation from Landsat 7 images.",
author = "Bibek Aryal and Miles, {Katie E.} and Zesati, {Sergio A. Vargas} and Olac Fuentes",
year = "2023",
month = jan,
day = "23",
doi = "10.7557/18.6789",
language = "English",
volume = "4",
journal = "Proceedings of the Northern Lights Deep Learning Workshop",
issn = "2703-6928",
publisher = "UiT The Arctic University of Norway",

}

RIS

TY - JOUR

T1 - Boundary Aware U-Net for Glacier Segmentation

AU - Aryal, Bibek

AU - Miles, Katie E.

AU - Zesati, Sergio A. Vargas

AU - Fuentes, Olac

PY - 2023/1/23

Y1 - 2023/1/23

N2 - Large-scale study of glaciers improves our understanding of global glacier change and is imperative for monitoring the ecological environment, preventing disasters, and studying the effects of global climate change. Glaciers in the Hindu Kush Himalaya (HKH) are particularly interesting as the HKH is one of the world's most sensitive regions for climate change. In this work, we: (1) propose a modified version of the U-Net for large-scale, spatially non-overlapping, clean glacial ice, and debris-covered glacial ice segmentation; (2) introduce a novel self-learning boundary-aware loss to improve debris-covered glacial ice segmentation performance; and (3) propose a feature-wise saliency score to understand the contribution of each feature in the multispectral Landsat 7 imagery for glacier mapping. Our results show that the debris-covered glacial ice segmentation model trained using self-learning boundary-aware loss outperformed the model trained using dice loss. Furthermore, we conclude that red, shortwave infrared, and near-infrared bands have the highest contribution toward debris-covered glacial ice segmentation from Landsat 7 images.

AB - Large-scale study of glaciers improves our understanding of global glacier change and is imperative for monitoring the ecological environment, preventing disasters, and studying the effects of global climate change. Glaciers in the Hindu Kush Himalaya (HKH) are particularly interesting as the HKH is one of the world's most sensitive regions for climate change. In this work, we: (1) propose a modified version of the U-Net for large-scale, spatially non-overlapping, clean glacial ice, and debris-covered glacial ice segmentation; (2) introduce a novel self-learning boundary-aware loss to improve debris-covered glacial ice segmentation performance; and (3) propose a feature-wise saliency score to understand the contribution of each feature in the multispectral Landsat 7 imagery for glacier mapping. Our results show that the debris-covered glacial ice segmentation model trained using self-learning boundary-aware loss outperformed the model trained using dice loss. Furthermore, we conclude that red, shortwave infrared, and near-infrared bands have the highest contribution toward debris-covered glacial ice segmentation from Landsat 7 images.

U2 - 10.7557/18.6789

DO - 10.7557/18.6789

M3 - Journal article

VL - 4

JO - Proceedings of the Northern Lights Deep Learning Workshop

JF - Proceedings of the Northern Lights Deep Learning Workshop

SN - 2703-6928

ER -