A methodology for underwater video classifier design and comparison on limited datasets

David B. Redpath, Katia Lebart, Chris Smith

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

This paper presents an experimental protocol developed for the design, performance estimation and comparison of underwater video classifer systems. Such systems have to be designed using application data that is small, sparse and extremely variable. The proposed protocol uses outlier rejection, data pairing, Bootstrap performance estimation and hypothesis testing to achieve a robust performance estimate and comparison between classifer designs. The protocol is demonstrated and assessed on an application experiment. The application involves the design of a classification system for the automated detection of trawling marks from mission video. Two systems are proposed using selective and geometric feature types and an ensemble classifier. The protocol robustly identifies differences between the two proposed system designs using error and discrimination rates. Overall the geometric feature system is chosen as the final system. The protocol was also compared with other performance estimates and found to have the closest match to actual test data performance. © 2005 IEEE.

Original languageEnglish
Title of host publicationOceans 2005 - Europe
Pages1084-1089
Number of pages6
Volume2
DOIs
Publication statusPublished - 2005
EventOceans 2005 - Europe - Brest, France
Duration: 20 Jun 200523 Jun 2005

Conference

ConferenceOceans 2005 - Europe
CountryFrance
CityBrest
Period20/06/0523/06/05

Fingerprint

Classifiers
Systems analysis
Testing
Experiments

Cite this

Redpath, D. B., Lebart, K., & Smith, C. (2005). A methodology for underwater video classifier design and comparison on limited datasets. In Oceans 2005 - Europe (Vol. 2, pp. 1084-1089) https://doi.org/10.1109/OCEANSE.2005.1513209
Redpath, David B. ; Lebart, Katia ; Smith, Chris. / A methodology for underwater video classifier design and comparison on limited datasets. Oceans 2005 - Europe. Vol. 2 2005. pp. 1084-1089
@inproceedings{aa067b2acabe481d820cd4b469e1336e,
title = "A methodology for underwater video classifier design and comparison on limited datasets",
abstract = "This paper presents an experimental protocol developed for the design, performance estimation and comparison of underwater video classifer systems. Such systems have to be designed using application data that is small, sparse and extremely variable. The proposed protocol uses outlier rejection, data pairing, Bootstrap performance estimation and hypothesis testing to achieve a robust performance estimate and comparison between classifer designs. The protocol is demonstrated and assessed on an application experiment. The application involves the design of a classification system for the automated detection of trawling marks from mission video. Two systems are proposed using selective and geometric feature types and an ensemble classifier. The protocol robustly identifies differences between the two proposed system designs using error and discrimination rates. Overall the geometric feature system is chosen as the final system. The protocol was also compared with other performance estimates and found to have the closest match to actual test data performance. {\circledC} 2005 IEEE.",
author = "Redpath, {David B.} and Katia Lebart and Chris Smith",
year = "2005",
doi = "10.1109/OCEANSE.2005.1513209",
language = "English",
isbn = "0780391039",
volume = "2",
pages = "1084--1089",
booktitle = "Oceans 2005 - Europe",

}

Redpath, DB, Lebart, K & Smith, C 2005, A methodology for underwater video classifier design and comparison on limited datasets. in Oceans 2005 - Europe. vol. 2, pp. 1084-1089, Oceans 2005 - Europe, Brest, France, 20/06/05. https://doi.org/10.1109/OCEANSE.2005.1513209

A methodology for underwater video classifier design and comparison on limited datasets. / Redpath, David B.; Lebart, Katia; Smith, Chris.

Oceans 2005 - Europe. Vol. 2 2005. p. 1084-1089.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - A methodology for underwater video classifier design and comparison on limited datasets

AU - Redpath, David B.

AU - Lebart, Katia

AU - Smith, Chris

PY - 2005

Y1 - 2005

N2 - This paper presents an experimental protocol developed for the design, performance estimation and comparison of underwater video classifer systems. Such systems have to be designed using application data that is small, sparse and extremely variable. The proposed protocol uses outlier rejection, data pairing, Bootstrap performance estimation and hypothesis testing to achieve a robust performance estimate and comparison between classifer designs. The protocol is demonstrated and assessed on an application experiment. The application involves the design of a classification system for the automated detection of trawling marks from mission video. Two systems are proposed using selective and geometric feature types and an ensemble classifier. The protocol robustly identifies differences between the two proposed system designs using error and discrimination rates. Overall the geometric feature system is chosen as the final system. The protocol was also compared with other performance estimates and found to have the closest match to actual test data performance. © 2005 IEEE.

AB - This paper presents an experimental protocol developed for the design, performance estimation and comparison of underwater video classifer systems. Such systems have to be designed using application data that is small, sparse and extremely variable. The proposed protocol uses outlier rejection, data pairing, Bootstrap performance estimation and hypothesis testing to achieve a robust performance estimate and comparison between classifer designs. The protocol is demonstrated and assessed on an application experiment. The application involves the design of a classification system for the automated detection of trawling marks from mission video. Two systems are proposed using selective and geometric feature types and an ensemble classifier. The protocol robustly identifies differences between the two proposed system designs using error and discrimination rates. Overall the geometric feature system is chosen as the final system. The protocol was also compared with other performance estimates and found to have the closest match to actual test data performance. © 2005 IEEE.

UR - http://www.scopus.com/inward/record.url?scp=33746874513&partnerID=8YFLogxK

U2 - 10.1109/OCEANSE.2005.1513209

DO - 10.1109/OCEANSE.2005.1513209

M3 - Conference contribution

SN - 0780391039

SN - 9780780391031

VL - 2

SP - 1084

EP - 1089

BT - Oceans 2005 - Europe

ER -