Self Supervised Learning for Few Shot Hyperspectral Image Classification

Nassim Ait Ali Braham; Lichao Mou; Jocelyn Chanussot; Julien Mairal; Xiao Xiang Zhu

doi:10.1109/IGARSS46834.2022.9884494

Self Supervised Learning for Few Shot Hyperspectral Image Classification

Nassim Ait Ali Braham
, Lichao Mou
, Jocelyn Chanussot
, Julien Mairal
, Xiao Xiang Zhu

Chair of Data Science in Earth Observation

Deutsches Zentrum für Luft- und Raumfahrt e.V. (DLR)
Technical University of Munich
University of Grenoble Alpes

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

20 Scopus citations

Abstract

Deep learning has proven to be a very effective approach for Hyperspectral Image (HSI) classification. However, deep neural networks require large annotated datasets to generalize well. This limits the applicability of deep learning for HSI classification, where manually labelling thousands of pixels for every scene is impractical. In this paper, we propose to leverage Self Supervised Learning (SSL) for HSI classification. We show that by pre-training an encoder on unlabeled pixels using Barlow-Twins, a state-of-the-art SSL algorithm, we can obtain accurate models with a handful of labels. Experimental results demonstrate that this approach significantly outperforms vanilla supervised learning.

Original language	English
Title of host publication	IGARSS 2022 - 2022 IEEE International Geoscience and Remote Sensing Symposium - Proceedings
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	267-270
Number of pages	4
ISBN (Electronic)	9781665427920
DOIs	https://doi.org/10.1109/IGARSS46834.2022.9884494
State	Published - 2022
Event	2022 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2022 - Kuala Lumpur, Malaysia Duration: 17 Jul 2022 → 22 Jul 2022

Publication series

Name	International Geoscience and Remote Sensing Symposium (IGARSS)
Volume	2022-July
ISSN (Print)	2153-6996
ISSN (Electronic)	2153-7003

Conference

Conference	2022 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2022
Country/Territory	Malaysia
City	Kuala Lumpur
Period	17/07/22 → 22/07/22

Keywords

Few Shot Classification
Hyperspectral Image Classification
Self Supervised Learning

Access to Document

10.1109/IGARSS46834.2022.9884494

Cite this

Braham, N. A. A., Mou, L., Chanussot, J., Mairal, J., & Zhu, X. X. (2022). Self Supervised Learning for Few Shot Hyperspectral Image Classification. In IGARSS 2022 - 2022 IEEE International Geoscience and Remote Sensing Symposium - Proceedings (pp. 267-270). (International Geoscience and Remote Sensing Symposium (IGARSS) ; Vol. 2022-July). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/IGARSS46834.2022.9884494

Braham, Nassim Ait Ali ; Mou, Lichao ; Chanussot, Jocelyn et al. / Self Supervised Learning for Few Shot Hyperspectral Image Classification. IGARSS 2022 - 2022 IEEE International Geoscience and Remote Sensing Symposium - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2022. pp. 267-270 (International Geoscience and Remote Sensing Symposium (IGARSS) ).

@inproceedings{e4665f8f14774b9aa95c9f972a2f9ef8,

title = "Self Supervised Learning for Few Shot Hyperspectral Image Classification",

abstract = "Deep learning has proven to be a very effective approach for Hyperspectral Image (HSI) classification. However, deep neural networks require large annotated datasets to generalize well. This limits the applicability of deep learning for HSI classification, where manually labelling thousands of pixels for every scene is impractical. In this paper, we propose to leverage Self Supervised Learning (SSL) for HSI classification. We show that by pre-training an encoder on unlabeled pixels using Barlow-Twins, a state-of-the-art SSL algorithm, we can obtain accurate models with a handful of labels. Experimental results demonstrate that this approach significantly outperforms vanilla supervised learning.",

keywords = "Few Shot Classification, Hyperspectral Image Classification, Self Supervised Learning",

author = "Braham, \{Nassim Ait Ali\} and Lichao Mou and Jocelyn Chanussot and Julien Mairal and Zhu, \{Xiao Xiang\}",

note = "Publisher Copyright: {\textcopyright} 2022 IEEE.; 2022 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2022 ; Conference date: 17-07-2022 Through 22-07-2022",

year = "2022",

doi = "10.1109/IGARSS46834.2022.9884494",

language = "English",

series = "International Geoscience and Remote Sensing Symposium (IGARSS) ",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "267--270",

booktitle = "IGARSS 2022 - 2022 IEEE International Geoscience and Remote Sensing Symposium - Proceedings",

}

Braham, NAA, Mou, L, Chanussot, J, Mairal, J & Zhu, XX 2022, Self Supervised Learning for Few Shot Hyperspectral Image Classification. in IGARSS 2022 - 2022 IEEE International Geoscience and Remote Sensing Symposium - Proceedings. International Geoscience and Remote Sensing Symposium (IGARSS) , vol. 2022-July, Institute of Electrical and Electronics Engineers Inc., pp. 267-270, 2022 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2022, Kuala Lumpur, Malaysia, 17/07/22. https://doi.org/10.1109/IGARSS46834.2022.9884494

Self Supervised Learning for Few Shot Hyperspectral Image Classification. / Braham, Nassim Ait Ali; Mou, Lichao; Chanussot, Jocelyn et al.
IGARSS 2022 - 2022 IEEE International Geoscience and Remote Sensing Symposium - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2022. p. 267-270 (International Geoscience and Remote Sensing Symposium (IGARSS) ; Vol. 2022-July).

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

TY - GEN

T1 - Self Supervised Learning for Few Shot Hyperspectral Image Classification

AU - Braham, Nassim Ait Ali

AU - Mou, Lichao

AU - Chanussot, Jocelyn

AU - Mairal, Julien

AU - Zhu, Xiao Xiang

PY - 2022

Y1 - 2022

N2 - Deep learning has proven to be a very effective approach for Hyperspectral Image (HSI) classification. However, deep neural networks require large annotated datasets to generalize well. This limits the applicability of deep learning for HSI classification, where manually labelling thousands of pixels for every scene is impractical. In this paper, we propose to leverage Self Supervised Learning (SSL) for HSI classification. We show that by pre-training an encoder on unlabeled pixels using Barlow-Twins, a state-of-the-art SSL algorithm, we can obtain accurate models with a handful of labels. Experimental results demonstrate that this approach significantly outperforms vanilla supervised learning.

AB - Deep learning has proven to be a very effective approach for Hyperspectral Image (HSI) classification. However, deep neural networks require large annotated datasets to generalize well. This limits the applicability of deep learning for HSI classification, where manually labelling thousands of pixels for every scene is impractical. In this paper, we propose to leverage Self Supervised Learning (SSL) for HSI classification. We show that by pre-training an encoder on unlabeled pixels using Barlow-Twins, a state-of-the-art SSL algorithm, we can obtain accurate models with a handful of labels. Experimental results demonstrate that this approach significantly outperforms vanilla supervised learning.

KW - Few Shot Classification

KW - Hyperspectral Image Classification

KW - Self Supervised Learning

UR - https://www.scopus.com/pages/publications/85133826858

U2 - 10.1109/IGARSS46834.2022.9884494

DO - 10.1109/IGARSS46834.2022.9884494

M3 - Conference contribution

AN - SCOPUS:85133826858

T3 - International Geoscience and Remote Sensing Symposium (IGARSS)

SP - 267

EP - 270

BT - IGARSS 2022 - 2022 IEEE International Geoscience and Remote Sensing Symposium - Proceedings

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2022 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2022

Y2 - 17 July 2022 through 22 July 2022

ER -

Braham NAA, Mou L, Chanussot J, Mairal J, Zhu XX. Self Supervised Learning for Few Shot Hyperspectral Image Classification. In IGARSS 2022 - 2022 IEEE International Geoscience and Remote Sensing Symposium - Proceedings. Institute of Electrical and Electronics Engineers Inc. 2022. p. 267-270. (International Geoscience and Remote Sensing Symposium (IGARSS) ). doi: 10.1109/IGARSS46834.2022.9884494

Self Supervised Learning for Few Shot Hyperspectral Image Classification

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this