Efficient ATR Using Contrastive Learning

Attiano Purpura-Pontoniere; Abdullah Al Zubaer Imran; Tarun Bhattacharya

doi:10.1117/12.2616822

Efficient ATR Using Contrastive Learning

Attiano Purpura-Pontoniere, Abdullah Al Zubaer Imran, Tarun Bhattacharya

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

1 Scopus citations

Abstract

Automatic Target Recognition (ATR) is a valuable application of computer vision that traditionally requires copious and tedious labeling through supervised learning. This research explored if ATR can be performed on satellite imagery at a comparable accuracy to a fully supervised baseline model with a considerably smaller subset of data labelled, on the order of 10%, using a recently developed semi-supervised technique, contrastive learning. Supervised contrastive loss was explored and compared to traditional cross entropy loss. Supervised contrastive loss was found to perform significantly better with a subset of the data labelled on the XView dataset, a publicly available dataset of satellite imagery captured with .3 meter ground sampling. The caveats when nothing and everything is labelled were additionally explored. [1].

Original language	English
Title of host publication	Automatic Target Recognition XXXII
Editors	Riad I. Hammoud, Timothy L. Overman, Abhijit Mahalanobis, Kristen Jaskie
ISBN (Electronic)	9781510650688
DOIs	https://doi.org/10.1117/12.2616822
State	Published - 2022
Event	Automatic Target Recognition XXXII 2022 - Virtual, Online Duration: Jun 6 2022 → Jun 12 2022

Publication series

Name	Proceedings of SPIE - The International Society for Optical Engineering
Volume	12096
ISSN (Print)	0277-786X
ISSN (Electronic)	1996-756X

Conference

Conference	Automatic Target Recognition XXXII 2022
City	Virtual, Online
Period	6/6/22 → 6/12/22

Bibliographical note

Publisher Copyright:
© 2022 SPIE. All rights reserved.

Keywords

ATR
Classification
Computer Vision
Contrastive Learning
Satellite Imagery
Self-Supervised Learning
Semi-Supervised Learning

ASJC Scopus subject areas

Electronic, Optical and Magnetic Materials
Condensed Matter Physics
Computer Science Applications
Applied Mathematics
Electrical and Electronic Engineering

Access to Document

10.1117/12.2616822

Cite this

@inproceedings{29e620ca551b4e01860c5ce239a54618,

title = "Efficient ATR Using Contrastive Learning",

abstract = "Automatic Target Recognition (ATR) is a valuable application of computer vision that traditionally requires copious and tedious labeling through supervised learning. This research explored if ATR can be performed on satellite imagery at a comparable accuracy to a fully supervised baseline model with a considerably smaller subset of data labelled, on the order of 10%, using a recently developed semi-supervised technique, contrastive learning. Supervised contrastive loss was explored and compared to traditional cross entropy loss. Supervised contrastive loss was found to perform significantly better with a subset of the data labelled on the XView dataset, a publicly available dataset of satellite imagery captured with .3 meter ground sampling. The caveats when nothing and everything is labelled were additionally explored. [1].",

keywords = "ATR, Classification, Computer Vision, Contrastive Learning, Satellite Imagery, Self-Supervised Learning, Semi-Supervised Learning",

author = "Attiano Purpura-Pontoniere and Imran, {Abdullah Al Zubaer} and Tarun Bhattacharya",

year = "2022",

doi = "10.1117/12.2616822",

language = "English",

series = "Proceedings of SPIE - The International Society for Optical Engineering",

editor = "Hammoud, {Riad I.} and Overman, {Timothy L.} and Abhijit Mahalanobis and Kristen Jaskie",

booktitle = "Automatic Target Recognition XXXII",

}

Purpura-Pontoniere, A, Imran, AAZ & Bhattacharya, T 2022, Efficient ATR Using Contrastive Learning. in RI Hammoud, TL Overman, A Mahalanobis & K Jaskie (eds), Automatic Target Recognition XXXII., 120960H, Proceedings of SPIE - The International Society for Optical Engineering, vol. 12096, Automatic Target Recognition XXXII 2022, Virtual, Online, 6/6/22. https://doi.org/10.1117/12.2616822

Efficient ATR Using Contrastive Learning. / Purpura-Pontoniere, Attiano; Imran, Abdullah Al Zubaer; Bhattacharya, Tarun.
Automatic Target Recognition XXXII. ed. / Riad I. Hammoud; Timothy L. Overman; Abhijit Mahalanobis; Kristen Jaskie. 2022. 120960H (Proceedings of SPIE - The International Society for Optical Engineering; Vol. 12096).

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

TY - GEN

T1 - Efficient ATR Using Contrastive Learning

AU - Purpura-Pontoniere, Attiano

AU - Imran, Abdullah Al Zubaer

AU - Bhattacharya, Tarun

PY - 2022

Y1 - 2022

N2 - Automatic Target Recognition (ATR) is a valuable application of computer vision that traditionally requires copious and tedious labeling through supervised learning. This research explored if ATR can be performed on satellite imagery at a comparable accuracy to a fully supervised baseline model with a considerably smaller subset of data labelled, on the order of 10%, using a recently developed semi-supervised technique, contrastive learning. Supervised contrastive loss was explored and compared to traditional cross entropy loss. Supervised contrastive loss was found to perform significantly better with a subset of the data labelled on the XView dataset, a publicly available dataset of satellite imagery captured with .3 meter ground sampling. The caveats when nothing and everything is labelled were additionally explored. [1].

AB - Automatic Target Recognition (ATR) is a valuable application of computer vision that traditionally requires copious and tedious labeling through supervised learning. This research explored if ATR can be performed on satellite imagery at a comparable accuracy to a fully supervised baseline model with a considerably smaller subset of data labelled, on the order of 10%, using a recently developed semi-supervised technique, contrastive learning. Supervised contrastive loss was explored and compared to traditional cross entropy loss. Supervised contrastive loss was found to perform significantly better with a subset of the data labelled on the XView dataset, a publicly available dataset of satellite imagery captured with .3 meter ground sampling. The caveats when nothing and everything is labelled were additionally explored. [1].

KW - ATR

KW - Classification

KW - Computer Vision

KW - Contrastive Learning

KW - Satellite Imagery

KW - Self-Supervised Learning

KW - Semi-Supervised Learning

UR - http://www.scopus.com/inward/record.url?scp=85133613503&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85133613503&partnerID=8YFLogxK

U2 - 10.1117/12.2616822

DO - 10.1117/12.2616822

M3 - Conference contribution

AN - SCOPUS:85133613503

T3 - Proceedings of SPIE - The International Society for Optical Engineering

BT - Automatic Target Recognition XXXII

A2 - Hammoud, Riad I.

A2 - Overman, Timothy L.

A2 - Mahalanobis, Abhijit

A2 - Jaskie, Kristen

T2 - Automatic Target Recognition XXXII 2022

Y2 - 6 June 2022 through 12 June 2022

ER -

Efficient ATR Using Contrastive Learning

Abstract

Publication series

Conference

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this