Zero-shot video retrieval from a query phrase including multiple concepts - Efforts and challenges in trecvid avs task-

Kazuya Ueki, Koji Hirakawa, Kotaro Kikuchi, Tetsunori Kobayashi

Research output: Contribution to journalArticle

Abstract

In this paper, we report our efforts and challenges on the TRECVID ad-hoc video search (AVS) task. The goal of the AVS task it to build a zero-shot video retrieval system using a complicated query phrase. Our system has the following two characteristics. First, we prepared a large number of pre-trained concept classifiers in advance that can detect various kinds of objects, persons, scenes, and actions. This strategy contributes to improve the word coverage rate of keywords in query phrases. Second, we selected additional concept classifiers by natural language processing techniques such as using word similarities or synonyms. We submitted our systems with these two characteristics to the TRECVID AVS task in 2016 and 2017, and one of our systems ranked the highest among all the submitted systems for the second consecutive year.

Original languageEnglish
Pages (from-to)983-990
Number of pages8
JournalSeimitsu Kogaku Kaishi/Journal of the Japan Society for Precision Engineering
Volume84
Issue number12
DOIs
Publication statusPublished - 2018 Jan 1
Externally publishedYes

Fingerprint

Classifiers
Processing

Keywords

  • Ad-hoc video search
  • Convolutional neural network
  • TRECVID
  • Video retrieval
  • Zero-shot learning

ASJC Scopus subject areas

  • Mechanical Engineering

Cite this

Zero-shot video retrieval from a query phrase including multiple concepts - Efforts and challenges in trecvid avs task-. / Ueki, Kazuya; Hirakawa, Koji; Kikuchi, Kotaro; Kobayashi, Tetsunori.

In: Seimitsu Kogaku Kaishi/Journal of the Japan Society for Precision Engineering, Vol. 84, No. 12, 01.01.2018, p. 983-990.

Research output: Contribution to journalArticle

@article{eb08867294ab40e893c5add98f1c77f0,
title = "Zero-shot video retrieval from a query phrase including multiple concepts - Efforts and challenges in trecvid avs task-",
abstract = "In this paper, we report our efforts and challenges on the TRECVID ad-hoc video search (AVS) task. The goal of the AVS task it to build a zero-shot video retrieval system using a complicated query phrase. Our system has the following two characteristics. First, we prepared a large number of pre-trained concept classifiers in advance that can detect various kinds of objects, persons, scenes, and actions. This strategy contributes to improve the word coverage rate of keywords in query phrases. Second, we selected additional concept classifiers by natural language processing techniques such as using word similarities or synonyms. We submitted our systems with these two characteristics to the TRECVID AVS task in 2016 and 2017, and one of our systems ranked the highest among all the submitted systems for the second consecutive year.",
keywords = "Ad-hoc video search, Convolutional neural network, TRECVID, Video retrieval, Zero-shot learning",
author = "Kazuya Ueki and Koji Hirakawa and Kotaro Kikuchi and Tetsunori Kobayashi",
year = "2018",
month = "1",
day = "1",
doi = "10.2493/jjspe.84.983",
language = "English",
volume = "84",
pages = "983--990",
journal = "Seimitsu Kogaku Kaishi/Journal of the Japan Society for Precision Engineering",
issn = "0912-0289",
publisher = "Japan Society for Precision Engineering",
number = "12",

}

TY - JOUR

T1 - Zero-shot video retrieval from a query phrase including multiple concepts - Efforts and challenges in trecvid avs task-

AU - Ueki, Kazuya

AU - Hirakawa, Koji

AU - Kikuchi, Kotaro

AU - Kobayashi, Tetsunori

PY - 2018/1/1

Y1 - 2018/1/1

N2 - In this paper, we report our efforts and challenges on the TRECVID ad-hoc video search (AVS) task. The goal of the AVS task it to build a zero-shot video retrieval system using a complicated query phrase. Our system has the following two characteristics. First, we prepared a large number of pre-trained concept classifiers in advance that can detect various kinds of objects, persons, scenes, and actions. This strategy contributes to improve the word coverage rate of keywords in query phrases. Second, we selected additional concept classifiers by natural language processing techniques such as using word similarities or synonyms. We submitted our systems with these two characteristics to the TRECVID AVS task in 2016 and 2017, and one of our systems ranked the highest among all the submitted systems for the second consecutive year.

AB - In this paper, we report our efforts and challenges on the TRECVID ad-hoc video search (AVS) task. The goal of the AVS task it to build a zero-shot video retrieval system using a complicated query phrase. Our system has the following two characteristics. First, we prepared a large number of pre-trained concept classifiers in advance that can detect various kinds of objects, persons, scenes, and actions. This strategy contributes to improve the word coverage rate of keywords in query phrases. Second, we selected additional concept classifiers by natural language processing techniques such as using word similarities or synonyms. We submitted our systems with these two characteristics to the TRECVID AVS task in 2016 and 2017, and one of our systems ranked the highest among all the submitted systems for the second consecutive year.

KW - Ad-hoc video search

KW - Convolutional neural network

KW - TRECVID

KW - Video retrieval

KW - Zero-shot learning

UR - http://www.scopus.com/inward/record.url?scp=85057768067&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85057768067&partnerID=8YFLogxK

U2 - 10.2493/jjspe.84.983

DO - 10.2493/jjspe.84.983

M3 - Article

AN - SCOPUS:85057768067

VL - 84

SP - 983

EP - 990

JO - Seimitsu Kogaku Kaishi/Journal of the Japan Society for Precision Engineering

JF - Seimitsu Kogaku Kaishi/Journal of the Japan Society for Precision Engineering

SN - 0912-0289

IS - 12

ER -