Ensemble Learning for Semantic Segmentation of Ancient Maya Architectures

Thorben Hellweg; Stefen Oehmcke; Ankit Kariryaa; Fabian Gieseke; Christian Igel

doi:10.48550/arXiv.2208.03163

Ensemble Learning for Semantic Segmentation of Ancient Maya Architectures

Thorben Hellweg, Stefen Oehmcke, Ankit Kariryaa, Fabian Gieseke, Christian Igel

Machine Learning

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

16 Downloads (Pure)

Abstract

Deep learning methods hold great promise for the automatic analysis of large-scale remote sensing data in archaeological research. Here, we present a robust approach to locating ancient Maya architectures (buildings, aguadas, and platforms) based on integrated segmentation of satellite imagery and aerial laser scanning data. Deep learning models with different architectures and loss functions were trained and combined to form an ensemble for pixel-wise classification. We applied both training data augmentation as well as test-time augmentation and performed morphological cleaning in the postprocessing phase. Our approach was evaluated in the context of the “Discover the mysteries of the Maya: An Integrated Image Segmentation Challenge” at ECML PKDD 2021 and achieved one of the best results with an average IoU of 0.8183.

Original language	English
Title of host publication	Discover the Mysteries of the Maya : Selected Contributions from the Machine Learning Challenge & the Discovery Challenge Workshop, ECML PKDD 2021
Editors	Dragi Kocev, Nikola Simidjievski, Ana Kostovska, Ivica Dimitrovski, Žiga Kokalj
Number of pages	7
Place of Publication	Ljubljana
Publisher	Jožef Stefan Institute
Publication date	2022
Pages	13-19
Chapter	3
ISBN (Electronic)	978-961-264-228-0
DOIs	https://doi.org/10.48550/arXiv.2208.03163
Publication status	Published - 2022
Event	European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases 2021 - Online, Bilbao, Spain Duration: 13 Sep 2021 → 17 Sep 2021 https://2021.ecmlpkdd.org/index.html

Conference

Conference	European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases 2021
Location	Online
Country/Territory	Spain
City	Bilbao
Period	13/09/2021 → 17/09/2021
Internet address	https://2021.ecmlpkdd.org/index.html

Access to Document

10.48550/arXiv.2208.03163

FulltextFinal published version, 1.03 MB

Cite this

Hellweg, T., Oehmcke, S., Kariryaa, A., Gieseke, F., & Igel, C. (2022). Ensemble Learning for Semantic Segmentation of Ancient Maya Architectures. In D. Kocev, N. Simidjievski, A. Kostovska, I. Dimitrovski, & Ž. Kokalj (Eds.), Discover the Mysteries of the Maya: Selected Contributions from the Machine Learning Challenge & the Discovery Challenge Workshop, ECML PKDD 2021 (pp. 13-19). Jožef Stefan Institute . https://doi.org/10.48550/arXiv.2208.03163

Ensemble Learning for Semantic Segmentation of Ancient Maya Architectures. / Hellweg, Thorben ; Oehmcke, Stefen ; Kariryaa, Ankit ; Gieseke, Fabian ; Igel, Christian.

Discover the Mysteries of the Maya: Selected Contributions from the Machine Learning Challenge & the Discovery Challenge Workshop, ECML PKDD 2021. ed. / Dragi Kocev; Nikola Simidjievski; Ana Kostovska; Ivica Dimitrovski; Žiga Kokalj. Ljubljana : Jožef Stefan Institute , 2022. p. 13-19.

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

Hellweg, T , Oehmcke, S , Kariryaa, A , Gieseke, F & Igel, C 2022, Ensemble Learning for Semantic Segmentation of Ancient Maya Architectures. in D Kocev, N Simidjievski, A Kostovska, I Dimitrovski & Ž Kokalj (eds), Discover the Mysteries of the Maya: Selected Contributions from the Machine Learning Challenge & the Discovery Challenge Workshop, ECML PKDD 2021. Jožef Stefan Institute , Ljubljana, pp. 13-19, European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases 2021, Bilbao, Spain, 13/09/2021. https://doi.org/10.48550/arXiv.2208.03163

Hellweg T , Oehmcke S , Kariryaa A , Gieseke F , Igel C. Ensemble Learning for Semantic Segmentation of Ancient Maya Architectures. In Kocev D, Simidjievski N, Kostovska A, Dimitrovski I, Kokalj Ž, editors, Discover the Mysteries of the Maya: Selected Contributions from the Machine Learning Challenge & the Discovery Challenge Workshop, ECML PKDD 2021. Ljubljana: Jožef Stefan Institute . 2022. p. 13-19 https://doi.org/10.48550/arXiv.2208.03163

Hellweg, Thorben ; Oehmcke, Stefen ; Kariryaa, Ankit ; Gieseke, Fabian ; Igel, Christian. / Ensemble Learning for Semantic Segmentation of Ancient Maya Architectures. Discover the Mysteries of the Maya: Selected Contributions from the Machine Learning Challenge & the Discovery Challenge Workshop, ECML PKDD 2021. editor / Dragi Kocev ; Nikola Simidjievski ; Ana Kostovska ; Ivica Dimitrovski ; Žiga Kokalj. Ljubljana : Jožef Stefan Institute , 2022. pp. 13-19

@inproceedings{25c2043fd99344cf9b045b926cf004a3,

title = "Ensemble Learning for Semantic Segmentation of Ancient Maya Architectures",

abstract = "Deep learning methods hold great promise for the automatic analysis of large-scale remote sensing data in archaeological research. Here, we present a robust approach to locating ancient Maya architectures (buildings, aguadas, and platforms) based on integrated segmentation of satellite imagery and aerial laser scanning data. Deep learning models with different architectures and loss functions were trained and combined to form an ensemble for pixel-wise classification. We applied both training data augmentation as well as test-time augmentation and performed morphological cleaning in the postprocessing phase. Our approach was evaluated in the context of the “Discover the mysteries of the Maya: An Integrated Image Segmentation Challenge” at ECML PKDD 2021 and achieved one of the best results with an average IoU of 0.8183.",

author = "Thorben Hellweg and Stefen Oehmcke and Ankit Kariryaa and Fabian Gieseke and Christian Igel",

year = "2022",

doi = "10.48550/arXiv.2208.03163",

language = "English",

pages = "13--19",

editor = "Dragi Kocev and Nikola Simidjievski and Ana Kostovska and Ivica Dimitrovski and {\v Z}iga Kokalj",

booktitle = "Discover the Mysteries of the Maya",

publisher = "Jo{\v z}ef Stefan Institute ",

note = " European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases 2021, ECML PKDD 2021 ; Conference date: 13-09-2021 Through 17-09-2021",

url = "https://2021.ecmlpkdd.org/index.html",

}

TY - GEN

T1 - Ensemble Learning for Semantic Segmentation of Ancient Maya Architectures

AU - Hellweg, Thorben

AU - Oehmcke, Stefen

AU - Kariryaa, Ankit

AU - Gieseke, Fabian

AU - Igel, Christian

PY - 2022

Y1 - 2022

N2 - Deep learning methods hold great promise for the automatic analysis of large-scale remote sensing data in archaeological research. Here, we present a robust approach to locating ancient Maya architectures (buildings, aguadas, and platforms) based on integrated segmentation of satellite imagery and aerial laser scanning data. Deep learning models with different architectures and loss functions were trained and combined to form an ensemble for pixel-wise classification. We applied both training data augmentation as well as test-time augmentation and performed morphological cleaning in the postprocessing phase. Our approach was evaluated in the context of the “Discover the mysteries of the Maya: An Integrated Image Segmentation Challenge” at ECML PKDD 2021 and achieved one of the best results with an average IoU of 0.8183.

AB - Deep learning methods hold great promise for the automatic analysis of large-scale remote sensing data in archaeological research. Here, we present a robust approach to locating ancient Maya architectures (buildings, aguadas, and platforms) based on integrated segmentation of satellite imagery and aerial laser scanning data. Deep learning models with different architectures and loss functions were trained and combined to form an ensemble for pixel-wise classification. We applied both training data augmentation as well as test-time augmentation and performed morphological cleaning in the postprocessing phase. Our approach was evaluated in the context of the “Discover the mysteries of the Maya: An Integrated Image Segmentation Challenge” at ECML PKDD 2021 and achieved one of the best results with an average IoU of 0.8183.

U2 - 10.48550/arXiv.2208.03163

DO - 10.48550/arXiv.2208.03163

M3 - Article in proceedings

SP - 13

EP - 19

BT - Discover the Mysteries of the Maya

A2 - Kocev, Dragi

A2 - Simidjievski, Nikola

A2 - Kostovska, Ana

A2 - Dimitrovski, Ivica

A2 - Kokalj, Žiga

PB - Jožef Stefan Institute

CY - Ljubljana

T2 - European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases 2021

Y2 - 13 September 2021 through 17 September 2021

ER -