HED-UNet: A multi-scale framework for simultaneous segmentation and edge detection
Segmentation models for remote sensing imagery are usually trained on the segmentation task alone. However, for many applications, the class boundaries carry semantic value. To account for this, we propose a new approach that unites both tasks within a single deep learning model. The proposed networ...
Published in: | 2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS |
---|---|
Main Authors: | , , , , |
Format: | Conference Object |
Language: | English |
Published: |
2021
|
Subjects: | |
Online Access: | https://elib.dlr.de/143088/ https://elib.dlr.de/143088/1/hed_unet-igarss.pdf |
id |
ftdlr:oai:elib.dlr.de:143088 |
---|---|
record_format |
openpolar |
spelling |
ftdlr:oai:elib.dlr.de:143088 2024-05-19T07:30:19+00:00 HED-UNet: A multi-scale framework for simultaneous segmentation and edge detection Heidler, Konrad Mou, LiChao Baumhoer, Celia Dietz, Andreas Zhu, Xiao Xiang 2021-07 application/pdf https://elib.dlr.de/143088/ https://elib.dlr.de/143088/1/hed_unet-igarss.pdf en eng https://elib.dlr.de/143088/1/hed_unet-igarss.pdf Heidler, Konrad und Mou, LiChao und Baumhoer, Celia und Dietz, Andreas und Zhu, Xiao Xiang (2021) HED-UNet: A multi-scale framework for simultaneous segmentation and edge detection. In: International Geoscience and Remote Sensing Symposium (IGARSS), Seiten 1-4. IGARSS 2021, 2021-07-12 - 2021-07-16, Brussels, Belgium. doi:10.1109/IGARSS47720.2021.9553585 <https://doi.org/10.1109/IGARSS47720.2021.9553585>. EO Data Science Dynamik der Landoberfläche Konferenzbeitrag PeerReviewed 2021 ftdlr https://doi.org/10.1109/IGARSS47720.2021.9553585 2024-04-25T00:56:38Z Segmentation models for remote sensing imagery are usually trained on the segmentation task alone. However, for many applications, the class boundaries carry semantic value. To account for this, we propose a new approach that unites both tasks within a single deep learning model. The proposed network architecture follows the successful encoder-decoder approach, and is improved by employing deep supervision at multiple resolution levels, as well as merging these resolution levels into a final prediction using a hierarchical attention mechanism. This framework is trained to detect the coastline in Sentinel-1 images of the Antarctic coastline. Its performance is then compared to conventional single-task approaches, and shown to outperform these methods. The code is available at https://github.com/khdlr/HED-UNet Conference Object Antarc* Antarctic German Aerospace Center: elib - DLR electronic library 2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS 3037 3040 |
institution |
Open Polar |
collection |
German Aerospace Center: elib - DLR electronic library |
op_collection_id |
ftdlr |
language |
English |
topic |
EO Data Science Dynamik der Landoberfläche |
spellingShingle |
EO Data Science Dynamik der Landoberfläche Heidler, Konrad Mou, LiChao Baumhoer, Celia Dietz, Andreas Zhu, Xiao Xiang HED-UNet: A multi-scale framework for simultaneous segmentation and edge detection |
topic_facet |
EO Data Science Dynamik der Landoberfläche |
description |
Segmentation models for remote sensing imagery are usually trained on the segmentation task alone. However, for many applications, the class boundaries carry semantic value. To account for this, we propose a new approach that unites both tasks within a single deep learning model. The proposed network architecture follows the successful encoder-decoder approach, and is improved by employing deep supervision at multiple resolution levels, as well as merging these resolution levels into a final prediction using a hierarchical attention mechanism. This framework is trained to detect the coastline in Sentinel-1 images of the Antarctic coastline. Its performance is then compared to conventional single-task approaches, and shown to outperform these methods. The code is available at https://github.com/khdlr/HED-UNet |
format |
Conference Object |
author |
Heidler, Konrad Mou, LiChao Baumhoer, Celia Dietz, Andreas Zhu, Xiao Xiang |
author_facet |
Heidler, Konrad Mou, LiChao Baumhoer, Celia Dietz, Andreas Zhu, Xiao Xiang |
author_sort |
Heidler, Konrad |
title |
HED-UNet: A multi-scale framework for simultaneous segmentation and edge detection |
title_short |
HED-UNet: A multi-scale framework for simultaneous segmentation and edge detection |
title_full |
HED-UNet: A multi-scale framework for simultaneous segmentation and edge detection |
title_fullStr |
HED-UNet: A multi-scale framework for simultaneous segmentation and edge detection |
title_full_unstemmed |
HED-UNet: A multi-scale framework for simultaneous segmentation and edge detection |
title_sort |
hed-unet: a multi-scale framework for simultaneous segmentation and edge detection |
publishDate |
2021 |
url |
https://elib.dlr.de/143088/ https://elib.dlr.de/143088/1/hed_unet-igarss.pdf |
genre |
Antarc* Antarctic |
genre_facet |
Antarc* Antarctic |
op_relation |
https://elib.dlr.de/143088/1/hed_unet-igarss.pdf Heidler, Konrad und Mou, LiChao und Baumhoer, Celia und Dietz, Andreas und Zhu, Xiao Xiang (2021) HED-UNet: A multi-scale framework for simultaneous segmentation and edge detection. In: International Geoscience and Remote Sensing Symposium (IGARSS), Seiten 1-4. IGARSS 2021, 2021-07-12 - 2021-07-16, Brussels, Belgium. doi:10.1109/IGARSS47720.2021.9553585 <https://doi.org/10.1109/IGARSS47720.2021.9553585>. |
op_doi |
https://doi.org/10.1109/IGARSS47720.2021.9553585 |
container_title |
2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS |
container_start_page |
3037 |
op_container_end_page |
3040 |
_version_ |
1799485294065483776 |