HIERARCHICAL PROBABILISTIC EMBEDDINGS FOR MULTI-VIEW IMAGE CLASSIFICATION

Benjamin Brodie; Subash Khanal; Muhammad Usman Rafique; Connor Greenwell; Nathan Jacobs

doi:10.1109/IGARSS47720.2021.9554405

HIERARCHICAL PROBABILISTIC EMBEDDINGS FOR MULTI-VIEW IMAGE CLASSIFICATION

Benjamin Brodie, Subash Khanal, Muhammad Usman Rafique, Connor Greenwell, Nathan Jacobs

Computer Science

Research output: Contribution to conference › Paper › peer-review

1 Scopus citations

Abstract

We address the task of image classification, when the available spectral bands can vary from image to image. We propose a model that learns to represent uncertainty over latent features in a way that is conditioned on the available bands. We expect that images with fewer bands will generally be more difficult to classify and hence have higher uncertainty. We compare two strategies for training such a model, one which uses explicit hierarchical constraints and one which relies on implicit constraints. We evaluate both using RGB and multispectral imagery from the EuroSat dataset and find that the hierarchical approach improves the compatibility of the resulting distributions without sacrificing accuracy.

Original language	English
Pages	4011-4014
Number of pages	4
DOIs	https://doi.org/10.1109/IGARSS47720.2021.9554405
State	Published - 2021
Event	2021 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2021 - Brussels, Belgium Duration: Jul 12 2021 → Jul 16 2021

Conference

Conference	2021 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2021
Country/Territory	Belgium
City	Brussels
Period	7/12/21 → 7/16/21

Bibliographical note

Publisher Copyright:
© 2021 IEEE

ASJC Scopus subject areas

Computer Science Applications
General Earth and Planetary Sciences

Access to Document

10.1109/IGARSS47720.2021.9554405

Cite this

@conference{051b1038684f42918811992241450f50,

title = "HIERARCHICAL PROBABILISTIC EMBEDDINGS FOR MULTI-VIEW IMAGE CLASSIFICATION",

abstract = "We address the task of image classification, when the available spectral bands can vary from image to image. We propose a model that learns to represent uncertainty over latent features in a way that is conditioned on the available bands. We expect that images with fewer bands will generally be more difficult to classify and hence have higher uncertainty. We compare two strategies for training such a model, one which uses explicit hierarchical constraints and one which relies on implicit constraints. We evaluate both using RGB and multispectral imagery from the EuroSat dataset and find that the hierarchical approach improves the compatibility of the resulting distributions without sacrificing accuracy.",

author = "Benjamin Brodie and Subash Khanal and Rafique, {Muhammad Usman} and Connor Greenwell and Nathan Jacobs",

note = "Publisher Copyright: {\textcopyright} 2021 IEEE; 2021 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2021 ; Conference date: 12-07-2021 Through 16-07-2021",

year = "2021",

doi = "10.1109/IGARSS47720.2021.9554405",

language = "English",

pages = "4011--4014",

}

TY - CONF

T1 - HIERARCHICAL PROBABILISTIC EMBEDDINGS FOR MULTI-VIEW IMAGE CLASSIFICATION

AU - Brodie, Benjamin

AU - Khanal, Subash

AU - Rafique, Muhammad Usman

AU - Greenwell, Connor

AU - Jacobs, Nathan

PY - 2021

Y1 - 2021

N2 - We address the task of image classification, when the available spectral bands can vary from image to image. We propose a model that learns to represent uncertainty over latent features in a way that is conditioned on the available bands. We expect that images with fewer bands will generally be more difficult to classify and hence have higher uncertainty. We compare two strategies for training such a model, one which uses explicit hierarchical constraints and one which relies on implicit constraints. We evaluate both using RGB and multispectral imagery from the EuroSat dataset and find that the hierarchical approach improves the compatibility of the resulting distributions without sacrificing accuracy.

AB - We address the task of image classification, when the available spectral bands can vary from image to image. We propose a model that learns to represent uncertainty over latent features in a way that is conditioned on the available bands. We expect that images with fewer bands will generally be more difficult to classify and hence have higher uncertainty. We compare two strategies for training such a model, one which uses explicit hierarchical constraints and one which relies on implicit constraints. We evaluate both using RGB and multispectral imagery from the EuroSat dataset and find that the hierarchical approach improves the compatibility of the resulting distributions without sacrificing accuracy.

UR - http://www.scopus.com/inward/record.url?scp=85130069989&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85130069989&partnerID=8YFLogxK

U2 - 10.1109/IGARSS47720.2021.9554405

DO - 10.1109/IGARSS47720.2021.9554405

M3 - Paper

AN - SCOPUS:85130069989

SP - 4011

EP - 4014

T2 - 2021 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2021

Y2 - 12 July 2021 through 16 July 2021

ER -

HIERARCHICAL PROBABILISTIC EMBEDDINGS FOR MULTI-VIEW IMAGE CLASSIFICATION

Abstract

Conference

Bibliographical note

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this