Seeing the bigger picture: Enabling large context windows in neural networks by combining multiple zoom levels

When adopting deep learning methods for remote sensing applications, the data usually needs to be cut into patches due to hardware limitations. Clearly, this practice discards a lot of contextual information as the model's information is limited to imagery from the given patch. We propose a mem...

Full description

Bibliographic Details
Published in:2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS
Main Authors: Heidler, Konrad, Mou, LiChao, Zhu, Xiao Xiang
Format: Conference Object
Language:English
Published: 2021
Subjects:
EO Data Science
Sea ice
Online Access:https://elib.dlr.de/143090/
https://elib.dlr.de/143090/1/igarss_zoomnn_heidler_final.pdf
id ftdlr:oai:elib.dlr.de:143090
record_format openpolar
spelling ftdlr:oai:elib.dlr.de:143090 2024-05-19T07:48:19+00:00 Seeing the bigger picture: Enabling large context windows in neural networks by combining multiple zoom levels Heidler, Konrad Mou, LiChao Zhu, Xiao Xiang 2021-07 application/pdf https://elib.dlr.de/143090/ https://elib.dlr.de/143090/1/igarss_zoomnn_heidler_final.pdf en eng https://elib.dlr.de/143090/1/igarss_zoomnn_heidler_final.pdf Heidler, Konrad und Mou, LiChao und Zhu, Xiao Xiang (2021) Seeing the bigger picture: Enabling large context windows in neural networks by combining multiple zoom levels. In: International Geoscience and Remote Sensing Symposium (IGARSS), Seiten 1-4. IGARSS 2021, 2021-07-12 - 2021-07-16, Brussels, Belgium. doi:10.1109/IGARSS47720.2021.9554434 <https://doi.org/10.1109/IGARSS47720.2021.9554434>. EO Data Science Konferenzbeitrag PeerReviewed 2021 ftdlr https://doi.org/10.1109/IGARSS47720.2021.9554434 2024-04-25T00:56:38Z When adopting deep learning methods for remote sensing applications, the data usually needs to be cut into patches due to hardware limitations. Clearly, this practice discards a lot of contextual information as the model's information is limited to imagery from the given patch. We propose a memory-efficient way around this limitation by using multiple patches of varying spatial extents on different resolution levels. Finally, this new approach is evaluated for the task of automated sea ice charting, where the added contextual information is shown to be beneficial to model performance. Conference Object Sea ice German Aerospace Center: elib - DLR electronic library 2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS 3033 3036
institution Open Polar
collection German Aerospace Center: elib - DLR electronic library
op_collection_id ftdlr
language English
topic EO Data Science
spellingShingle EO Data Science
Heidler, Konrad
Mou, LiChao
Zhu, Xiao Xiang
Seeing the bigger picture: Enabling large context windows in neural networks by combining multiple zoom levels
topic_facet EO Data Science
description When adopting deep learning methods for remote sensing applications, the data usually needs to be cut into patches due to hardware limitations. Clearly, this practice discards a lot of contextual information as the model's information is limited to imagery from the given patch. We propose a memory-efficient way around this limitation by using multiple patches of varying spatial extents on different resolution levels. Finally, this new approach is evaluated for the task of automated sea ice charting, where the added contextual information is shown to be beneficial to model performance.
format Conference Object
author Heidler, Konrad
Mou, LiChao
Zhu, Xiao Xiang
author_facet Heidler, Konrad
Mou, LiChao
Zhu, Xiao Xiang
author_sort Heidler, Konrad
title Seeing the bigger picture: Enabling large context windows in neural networks by combining multiple zoom levels
title_short Seeing the bigger picture: Enabling large context windows in neural networks by combining multiple zoom levels
title_full Seeing the bigger picture: Enabling large context windows in neural networks by combining multiple zoom levels
title_fullStr Seeing the bigger picture: Enabling large context windows in neural networks by combining multiple zoom levels
title_full_unstemmed Seeing the bigger picture: Enabling large context windows in neural networks by combining multiple zoom levels
title_sort seeing the bigger picture: enabling large context windows in neural networks by combining multiple zoom levels
publishDate 2021
url https://elib.dlr.de/143090/
https://elib.dlr.de/143090/1/igarss_zoomnn_heidler_final.pdf
genre Sea ice
genre_facet Sea ice
op_relation https://elib.dlr.de/143090/1/igarss_zoomnn_heidler_final.pdf
Heidler, Konrad und Mou, LiChao und Zhu, Xiao Xiang (2021) Seeing the bigger picture: Enabling large context windows in neural networks by combining multiple zoom levels. In: International Geoscience and Remote Sensing Symposium (IGARSS), Seiten 1-4. IGARSS 2021, 2021-07-12 - 2021-07-16, Brussels, Belgium. doi:10.1109/IGARSS47720.2021.9554434 <https://doi.org/10.1109/IGARSS47720.2021.9554434>.
op_doi https://doi.org/10.1109/IGARSS47720.2021.9554434
container_title 2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS
container_start_page 3033
op_container_end_page 3036
_version_ 1799488874637950976

Similar Items