Similar-Video Retrieval via Learned Exemplars and Time-Warped Alignment

Teruki Horie, Masafumi Moriwaki, Ryota Yokote, Shota Ninomiya, Akihiro Shikano, Yasuo Matsuyama

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

    Abstract

    New learning algorithms and systems for retrieving similar videos are presented. Each query is a video itself. For each video, a set of exemplars is machine-learned by new algorithms. Two methods were tried. The first and main one is the time-bound affinity propagation. The second is the harmonic competition which approximates the first. In the similar-video retrieval, the number of exemplar frames is variable according to the length and contents of videos. Therefore, each exemplar possesses responsible frames. By considering this property, we give a novel similarity measure which contains the Levenshtein distance (L-distance) as its special case. This new measure, the M-distance, is applicable to both of global and local alignments for exemplars. Experimental results in view of precision-recall curves show creditable scores in the region of interest.

    Original languageEnglish
    Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
    PublisherSpringer Verlag
    Pages85-94
    Number of pages10
    Volume8836
    ISBN (Print)9783319126425
    Publication statusPublished - 2014
    Event21st International Conference on Neural Information Processing, ICONIP 2014 - Kuching
    Duration: 2014 Nov 32014 Nov 6

    Publication series

    NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
    Volume8836
    ISSN (Print)03029743
    ISSN (Electronic)16113349

    Other

    Other21st International Conference on Neural Information Processing, ICONIP 2014
    CityKuching
    Period14/11/314/11/6

    Fingerprint

    Video Retrieval
    Learning algorithms
    Learning systems
    Alignment
    Learning Systems
    Region of Interest
    Similarity Measure
    Affine transformation
    Learning Algorithm
    Harmonic
    Query
    Propagation
    Curve
    Experimental Results

    Keywords

    • Exemplar
    • M-distance
    • Numerical label
    • Similar-video retrieval
    • Time-bound affinity propagation

    ASJC Scopus subject areas

    • Computer Science(all)
    • Theoretical Computer Science

    Cite this

    Horie, T., Moriwaki, M., Yokote, R., Ninomiya, S., Shikano, A., & Matsuyama, Y. (2014). Similar-Video Retrieval via Learned Exemplars and Time-Warped Alignment. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 8836, pp. 85-94). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8836). Springer Verlag.

    Similar-Video Retrieval via Learned Exemplars and Time-Warped Alignment. / Horie, Teruki; Moriwaki, Masafumi; Yokote, Ryota; Ninomiya, Shota; Shikano, Akihiro; Matsuyama, Yasuo.

    Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 8836 Springer Verlag, 2014. p. 85-94 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8836).

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

    Horie, T, Moriwaki, M, Yokote, R, Ninomiya, S, Shikano, A & Matsuyama, Y 2014, Similar-Video Retrieval via Learned Exemplars and Time-Warped Alignment. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). vol. 8836, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 8836, Springer Verlag, pp. 85-94, 21st International Conference on Neural Information Processing, ICONIP 2014, Kuching, 14/11/3.
    Horie T, Moriwaki M, Yokote R, Ninomiya S, Shikano A, Matsuyama Y. Similar-Video Retrieval via Learned Exemplars and Time-Warped Alignment. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 8836. Springer Verlag. 2014. p. 85-94. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
    Horie, Teruki ; Moriwaki, Masafumi ; Yokote, Ryota ; Ninomiya, Shota ; Shikano, Akihiro ; Matsuyama, Yasuo. / Similar-Video Retrieval via Learned Exemplars and Time-Warped Alignment. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 8836 Springer Verlag, 2014. pp. 85-94 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
    @inproceedings{30536d4e3e1c43f99cfa286172172f58,
    title = "Similar-Video Retrieval via Learned Exemplars and Time-Warped Alignment",
    abstract = "New learning algorithms and systems for retrieving similar videos are presented. Each query is a video itself. For each video, a set of exemplars is machine-learned by new algorithms. Two methods were tried. The first and main one is the time-bound affinity propagation. The second is the harmonic competition which approximates the first. In the similar-video retrieval, the number of exemplar frames is variable according to the length and contents of videos. Therefore, each exemplar possesses responsible frames. By considering this property, we give a novel similarity measure which contains the Levenshtein distance (L-distance) as its special case. This new measure, the M-distance, is applicable to both of global and local alignments for exemplars. Experimental results in view of precision-recall curves show creditable scores in the region of interest.",
    keywords = "Exemplar, M-distance, Numerical label, Similar-video retrieval, Time-bound affinity propagation",
    author = "Teruki Horie and Masafumi Moriwaki and Ryota Yokote and Shota Ninomiya and Akihiro Shikano and Yasuo Matsuyama",
    year = "2014",
    language = "English",
    isbn = "9783319126425",
    volume = "8836",
    series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
    publisher = "Springer Verlag",
    pages = "85--94",
    booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

    }

    TY - GEN

    T1 - Similar-Video Retrieval via Learned Exemplars and Time-Warped Alignment

    AU - Horie, Teruki

    AU - Moriwaki, Masafumi

    AU - Yokote, Ryota

    AU - Ninomiya, Shota

    AU - Shikano, Akihiro

    AU - Matsuyama, Yasuo

    PY - 2014

    Y1 - 2014

    N2 - New learning algorithms and systems for retrieving similar videos are presented. Each query is a video itself. For each video, a set of exemplars is machine-learned by new algorithms. Two methods were tried. The first and main one is the time-bound affinity propagation. The second is the harmonic competition which approximates the first. In the similar-video retrieval, the number of exemplar frames is variable according to the length and contents of videos. Therefore, each exemplar possesses responsible frames. By considering this property, we give a novel similarity measure which contains the Levenshtein distance (L-distance) as its special case. This new measure, the M-distance, is applicable to both of global and local alignments for exemplars. Experimental results in view of precision-recall curves show creditable scores in the region of interest.

    AB - New learning algorithms and systems for retrieving similar videos are presented. Each query is a video itself. For each video, a set of exemplars is machine-learned by new algorithms. Two methods were tried. The first and main one is the time-bound affinity propagation. The second is the harmonic competition which approximates the first. In the similar-video retrieval, the number of exemplar frames is variable according to the length and contents of videos. Therefore, each exemplar possesses responsible frames. By considering this property, we give a novel similarity measure which contains the Levenshtein distance (L-distance) as its special case. This new measure, the M-distance, is applicable to both of global and local alignments for exemplars. Experimental results in view of precision-recall curves show creditable scores in the region of interest.

    KW - Exemplar

    KW - M-distance

    KW - Numerical label

    KW - Similar-video retrieval

    KW - Time-bound affinity propagation

    UR - http://www.scopus.com/inward/record.url?scp=84910019740&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=84910019740&partnerID=8YFLogxK

    M3 - Conference contribution

    SN - 9783319126425

    VL - 8836

    T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

    SP - 85

    EP - 94

    BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

    PB - Springer Verlag

    ER -