Multi-layer feature extractions for image classification - Knowledge from deep CNNs

Kazuya Ueki, Tetsunori Kobayashi

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

    Abstract

    Recently, there has been considerable research into the application of deep learning to image recognition. Notably, deep convolutional neural networks (CNNs) have achieved excellent performance in a number of image classification tasks, compared with conventional methods based on techniques such as Bag-of-Features (BoF) using local descriptors. In this paper, to cultivate a better understanding of the structure of CNN, we focus on the characteristics of deep CNNs, and adapt them to SIFT+BoF-based methods to improve the classification accuracy. We introduce the multi-layer structure of CNNs into the classification pipeline of the BoF framework, and conduct experiments to confirm the effectiveness of this approach using a fine-grained visual categorization dataset. The results show that the average classification rate is improved from 52.4% to 69.8%.

    Original languageEnglish
    Title of host publication2015 22nd International Conference on Systems, Signals and Image Processing - Proceedings of IWSSIP 2015
    PublisherInstitute of Electrical and Electronics Engineers Inc.
    Pages9-12
    Number of pages4
    ISBN (Print)9781467383530
    DOIs
    Publication statusPublished - 2015 Oct 30
    Event22nd International Conference on Systems, Signals and Image Processing, IWSSIP 2015 - London, United Kingdom
    Duration: 2015 Sep 102015 Sep 12

    Other

    Other22nd International Conference on Systems, Signals and Image Processing, IWSSIP 2015
    CountryUnited Kingdom
    CityLondon
    Period15/9/1015/9/12

    Fingerprint

    Image classification
    Feature extraction
    Neural networks
    Image recognition
    Pipelines
    Experiments

    Keywords

    • Bag-of-Features
    • Deep learning
    • Feature extraction
    • Fine-grained visual categorization
    • Generic object recognition

    ASJC Scopus subject areas

    • Computer Networks and Communications
    • Computer Vision and Pattern Recognition
    • Signal Processing

    Cite this

    Ueki, K., & Kobayashi, T. (2015). Multi-layer feature extractions for image classification - Knowledge from deep CNNs. In 2015 22nd International Conference on Systems, Signals and Image Processing - Proceedings of IWSSIP 2015 (pp. 9-12). [7313925] Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/IWSSIP.2015.7313925

    Multi-layer feature extractions for image classification - Knowledge from deep CNNs. / Ueki, Kazuya; Kobayashi, Tetsunori.

    2015 22nd International Conference on Systems, Signals and Image Processing - Proceedings of IWSSIP 2015. Institute of Electrical and Electronics Engineers Inc., 2015. p. 9-12 7313925.

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

    Ueki, K & Kobayashi, T 2015, Multi-layer feature extractions for image classification - Knowledge from deep CNNs. in 2015 22nd International Conference on Systems, Signals and Image Processing - Proceedings of IWSSIP 2015., 7313925, Institute of Electrical and Electronics Engineers Inc., pp. 9-12, 22nd International Conference on Systems, Signals and Image Processing, IWSSIP 2015, London, United Kingdom, 15/9/10. https://doi.org/10.1109/IWSSIP.2015.7313925
    Ueki K, Kobayashi T. Multi-layer feature extractions for image classification - Knowledge from deep CNNs. In 2015 22nd International Conference on Systems, Signals and Image Processing - Proceedings of IWSSIP 2015. Institute of Electrical and Electronics Engineers Inc. 2015. p. 9-12. 7313925 https://doi.org/10.1109/IWSSIP.2015.7313925
    Ueki, Kazuya ; Kobayashi, Tetsunori. / Multi-layer feature extractions for image classification - Knowledge from deep CNNs. 2015 22nd International Conference on Systems, Signals and Image Processing - Proceedings of IWSSIP 2015. Institute of Electrical and Electronics Engineers Inc., 2015. pp. 9-12
    @inproceedings{5ebb651d3d0645c39374af51f8298059,
    title = "Multi-layer feature extractions for image classification - Knowledge from deep CNNs",
    abstract = "Recently, there has been considerable research into the application of deep learning to image recognition. Notably, deep convolutional neural networks (CNNs) have achieved excellent performance in a number of image classification tasks, compared with conventional methods based on techniques such as Bag-of-Features (BoF) using local descriptors. In this paper, to cultivate a better understanding of the structure of CNN, we focus on the characteristics of deep CNNs, and adapt them to SIFT+BoF-based methods to improve the classification accuracy. We introduce the multi-layer structure of CNNs into the classification pipeline of the BoF framework, and conduct experiments to confirm the effectiveness of this approach using a fine-grained visual categorization dataset. The results show that the average classification rate is improved from 52.4{\%} to 69.8{\%}.",
    keywords = "Bag-of-Features, Deep learning, Feature extraction, Fine-grained visual categorization, Generic object recognition",
    author = "Kazuya Ueki and Tetsunori Kobayashi",
    year = "2015",
    month = "10",
    day = "30",
    doi = "10.1109/IWSSIP.2015.7313925",
    language = "English",
    isbn = "9781467383530",
    pages = "9--12",
    booktitle = "2015 22nd International Conference on Systems, Signals and Image Processing - Proceedings of IWSSIP 2015",
    publisher = "Institute of Electrical and Electronics Engineers Inc.",

    }

    TY - GEN

    T1 - Multi-layer feature extractions for image classification - Knowledge from deep CNNs

    AU - Ueki, Kazuya

    AU - Kobayashi, Tetsunori

    PY - 2015/10/30

    Y1 - 2015/10/30

    N2 - Recently, there has been considerable research into the application of deep learning to image recognition. Notably, deep convolutional neural networks (CNNs) have achieved excellent performance in a number of image classification tasks, compared with conventional methods based on techniques such as Bag-of-Features (BoF) using local descriptors. In this paper, to cultivate a better understanding of the structure of CNN, we focus on the characteristics of deep CNNs, and adapt them to SIFT+BoF-based methods to improve the classification accuracy. We introduce the multi-layer structure of CNNs into the classification pipeline of the BoF framework, and conduct experiments to confirm the effectiveness of this approach using a fine-grained visual categorization dataset. The results show that the average classification rate is improved from 52.4% to 69.8%.

    AB - Recently, there has been considerable research into the application of deep learning to image recognition. Notably, deep convolutional neural networks (CNNs) have achieved excellent performance in a number of image classification tasks, compared with conventional methods based on techniques such as Bag-of-Features (BoF) using local descriptors. In this paper, to cultivate a better understanding of the structure of CNN, we focus on the characteristics of deep CNNs, and adapt them to SIFT+BoF-based methods to improve the classification accuracy. We introduce the multi-layer structure of CNNs into the classification pipeline of the BoF framework, and conduct experiments to confirm the effectiveness of this approach using a fine-grained visual categorization dataset. The results show that the average classification rate is improved from 52.4% to 69.8%.

    KW - Bag-of-Features

    KW - Deep learning

    KW - Feature extraction

    KW - Fine-grained visual categorization

    KW - Generic object recognition

    UR - http://www.scopus.com/inward/record.url?scp=84961718052&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=84961718052&partnerID=8YFLogxK

    U2 - 10.1109/IWSSIP.2015.7313925

    DO - 10.1109/IWSSIP.2015.7313925

    M3 - Conference contribution

    SN - 9781467383530

    SP - 9

    EP - 12

    BT - 2015 22nd International Conference on Systems, Signals and Image Processing - Proceedings of IWSSIP 2015

    PB - Institute of Electrical and Electronics Engineers Inc.

    ER -