Academic Bibliography

 Search 200 years of publications by Ghent University researchers.
Advanced search
2 files | 21.28 MB
Add to list
  1. Search results
  2. Recovering from catastrophic receptiv...

Recovering from catastrophic receptive field overflow in semantic segmentation of high resolution images : application to seabed characterization

Yoann Arhant (UGent) , Olga Lopera Tellez, Xavier Neyt and Aleksandra Pizurica (UGent)
(2024) IGARSS 2024-2024 IEEE INTERNATIONAL GEOSCIENCE AND REMOTE SENSING SYMPOSIUM, IGARSS 2024. In IEEE International Symposium on Geoscience and Remote Sensing IGARSS p.9561-9565
Author
Organization
Abstract
This paper addresses a critical issue in seabed characterization with deep learning semantic segmentation using high-resolution Synthetic Aperture Sonar (SAS) data, that we call Catastrophic Receptive Field Overflow (CRFO). We propose novel methods, including Mosaic Augmentation and Homogeneous Patch Rejection, to (1) effectively mitigate CRFO and (2) enhance model performance. Through experiments on real-world SAS data, we investigate the origins of CRFO, revealing its dependence on model architectures and data characteristics. The presented solutions exhibit promising results, whether measured in terms of Overall Accuracy or the reliability of models in inference across various image input sizes or aspect ratios, in the face of new proposed metrics. These findings provide valuable insights for addressing CRFO challenges in tasks involving relatively homogeneous datasets.
Keywords
Deep Learning, Semantic Segmentation, Remote Sensing, Seabed Characterization

Downloads

  • Download
    Preprint ARHANT IGARSS 2024.pdf
    • full text (Accepted manuscript)
    • |
    • open access
    • |
    • PDF
    • |
    • 10.37 MB
  • Download
    (...).pdf
    • full text (Published version)
    • |
    • UGent only
    • |
    • PDF
    • |
    • 10.91 MB

Citation

Please use this url to cite or link to this publication:

MLA
Arhant, Yoann, et al. “Recovering from Catastrophic Receptive Field Overflow in Semantic Segmentation of High Resolution Images : Application to Seabed Characterization.” IGARSS 2024-2024 IEEE INTERNATIONAL GEOSCIENCE AND REMOTE SENSING SYMPOSIUM, IGARSS 2024, IEEE, 2024, pp. 9561–65, doi:10.1109/IGARSS53475.2024.10642639.
APA
Arhant, Y., Tellez, O. L., Neyt, X., & Pizurica, A. (2024). Recovering from catastrophic receptive field overflow in semantic segmentation of high resolution images : application to seabed characterization. IGARSS 2024-2024 IEEE INTERNATIONAL GEOSCIENCE AND REMOTE SENSING SYMPOSIUM, IGARSS 2024, 9561–9565. https://doi.org/10.1109/IGARSS53475.2024.10642639
Chicago author-date
Arhant, Yoann, Olga Lopera Tellez, Xavier Neyt, and Aleksandra Pizurica. 2024. “Recovering from Catastrophic Receptive Field Overflow in Semantic Segmentation of High Resolution Images : Application to Seabed Characterization.” In IGARSS 2024-2024 IEEE INTERNATIONAL GEOSCIENCE AND REMOTE SENSING SYMPOSIUM, IGARSS 2024, 9561–65. IEEE. https://doi.org/10.1109/IGARSS53475.2024.10642639.
Chicago author-date (all authors)
Arhant, Yoann, Olga Lopera Tellez, Xavier Neyt, and Aleksandra Pizurica. 2024. “Recovering from Catastrophic Receptive Field Overflow in Semantic Segmentation of High Resolution Images : Application to Seabed Characterization.” In IGARSS 2024-2024 IEEE INTERNATIONAL GEOSCIENCE AND REMOTE SENSING SYMPOSIUM, IGARSS 2024, 9561–9565. IEEE. doi:10.1109/IGARSS53475.2024.10642639.
Vancouver
1.
Arhant Y, Tellez OL, Neyt X, Pizurica A. Recovering from catastrophic receptive field overflow in semantic segmentation of high resolution images : application to seabed characterization. In: IGARSS 2024-2024 IEEE INTERNATIONAL GEOSCIENCE AND REMOTE SENSING SYMPOSIUM, IGARSS 2024. IEEE; 2024. p. 9561–5.
IEEE
[1]
Y. Arhant, O. L. Tellez, X. Neyt, and A. Pizurica, “Recovering from catastrophic receptive field overflow in semantic segmentation of high resolution images : application to seabed characterization,” in IGARSS 2024-2024 IEEE INTERNATIONAL GEOSCIENCE AND REMOTE SENSING SYMPOSIUM, IGARSS 2024, Athens, Greece, 2024, pp. 9561–9565.
@inproceedings{01J7BQVP98P79SRP8HD7WY6FND,
  abstract     = {{This paper addresses a critical issue in seabed characterization with deep learning semantic segmentation using high-resolution Synthetic Aperture Sonar (SAS) data, that we call Catastrophic Receptive Field Overflow (CRFO). We propose novel methods, including Mosaic Augmentation and Homogeneous Patch Rejection, to (1) effectively mitigate CRFO and (2) enhance model performance. Through experiments on real-world SAS data, we investigate the origins of CRFO, revealing its dependence on model architectures and data characteristics. The presented solutions exhibit promising results, whether measured in terms of Overall Accuracy or the reliability of models in inference across various image input sizes or aspect ratios, in the face of new proposed metrics. These findings provide valuable insights for addressing CRFO challenges in tasks involving relatively homogeneous datasets.}},
  author       = {{Arhant, Yoann and Tellez, Olga Lopera and Neyt, Xavier and Pizurica, Aleksandra}},
  booktitle    = {{IGARSS 2024-2024 IEEE INTERNATIONAL GEOSCIENCE AND REMOTE SENSING SYMPOSIUM, IGARSS 2024}},
  isbn         = {{9798350360332}},
  issn         = {{2153-6996}},
  keywords     = {{Deep Learning,Semantic Segmentation,Remote Sensing,Seabed Characterization}},
  language     = {{eng}},
  location     = {{Athens, Greece}},
  pages        = {{9561--9565}},
  publisher    = {{IEEE}},
  title        = {{Recovering from catastrophic receptive field overflow in semantic segmentation of high resolution images : application to seabed characterization}},
  url          = {{http://doi.org/10.1109/IGARSS53475.2024.10642639}},
  year         = {{2024}},
}
Altmetric
View in Altmetric
Web of Science
Times cited: