A negative sample image selection method referring to semantic hierarchical structure for image annotation

Shan Bin Chan, Hayato Yamana, Shin'Ichi Satoh

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

    Abstract

    When SVM is adopted for image annotation, we know that high quality sample images will improve image recognition accuracy. Images with the same visual/semantic features are adopted as positive sample images, and images with different visual/semantic features are adopted as negative sample images. But it is labor intensive in high quality sample images selection, especially when collecting by visual features. Most researchers randomly choose positive and negative sample images for classifier training. In many applications, adopting different negative sample image datasets will vary annotation accuracy. In this research, we will discuss the accuracy between different negative sample images dataset collected by semantic features. We adopted Image Net as image dataset in this study, and we adopted Word Net for building semantic hierarchical tree. Semantic hierarchical structure tree is adopted to calculate the distance between each node. Then we adopt this distance relationship to prepare positive and negative sample images. We prepare one baseline method and suggest six different negative sample images selection methods for experiment. The binary SVM classifier training and prediction is implemented to compare the accuracy and Mean Reciprocal Rank (MRR) between baseline and each proposed method. Our results show that if we select uniform amount of negative sample images in each distance in the semantic hierarchical tree, we will achieve highest accuracy.

    Original languageEnglish
    Title of host publicationProceedings - 2013 International Conference on Signal-Image Technology and Internet-Based Systems, SITIS 2013
    Pages162-167
    Number of pages6
    DOIs
    Publication statusPublished - 2013
    Event2013 9th International Conference on Signal-Image Technology and Internet-Based Systems, SITIS 2013 - Kyoto
    Duration: 2013 Dec 22013 Dec 5

    Other

    Other2013 9th International Conference on Signal-Image Technology and Internet-Based Systems, SITIS 2013
    CityKyoto
    Period13/12/213/12/5

    Fingerprint

    Semantics
    Classifiers
    Image recognition
    Personnel
    Experiments

    Keywords

    • Image Annotation
    • ImageNet
    • Machine Learning
    • Negative Sample Selection
    • SVM
    • WordNet

    ASJC Scopus subject areas

    • Computer Networks and Communications
    • Signal Processing

    Cite this

    Chan, S. B., Yamana, H., & Satoh, SI. (2013). A negative sample image selection method referring to semantic hierarchical structure for image annotation. In Proceedings - 2013 International Conference on Signal-Image Technology and Internet-Based Systems, SITIS 2013 (pp. 162-167). [6727186] https://doi.org/10.1109/SITIS.2013.37

    A negative sample image selection method referring to semantic hierarchical structure for image annotation. / Chan, Shan Bin; Yamana, Hayato; Satoh, Shin'Ichi.

    Proceedings - 2013 International Conference on Signal-Image Technology and Internet-Based Systems, SITIS 2013. 2013. p. 162-167 6727186.

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

    Chan, SB, Yamana, H & Satoh, SI 2013, A negative sample image selection method referring to semantic hierarchical structure for image annotation. in Proceedings - 2013 International Conference on Signal-Image Technology and Internet-Based Systems, SITIS 2013., 6727186, pp. 162-167, 2013 9th International Conference on Signal-Image Technology and Internet-Based Systems, SITIS 2013, Kyoto, 13/12/2. https://doi.org/10.1109/SITIS.2013.37
    Chan SB, Yamana H, Satoh SI. A negative sample image selection method referring to semantic hierarchical structure for image annotation. In Proceedings - 2013 International Conference on Signal-Image Technology and Internet-Based Systems, SITIS 2013. 2013. p. 162-167. 6727186 https://doi.org/10.1109/SITIS.2013.37
    Chan, Shan Bin ; Yamana, Hayato ; Satoh, Shin'Ichi. / A negative sample image selection method referring to semantic hierarchical structure for image annotation. Proceedings - 2013 International Conference on Signal-Image Technology and Internet-Based Systems, SITIS 2013. 2013. pp. 162-167
    @inproceedings{cd4a31e8745c490da9b436c7c1587b72,
    title = "A negative sample image selection method referring to semantic hierarchical structure for image annotation",
    abstract = "When SVM is adopted for image annotation, we know that high quality sample images will improve image recognition accuracy. Images with the same visual/semantic features are adopted as positive sample images, and images with different visual/semantic features are adopted as negative sample images. But it is labor intensive in high quality sample images selection, especially when collecting by visual features. Most researchers randomly choose positive and negative sample images for classifier training. In many applications, adopting different negative sample image datasets will vary annotation accuracy. In this research, we will discuss the accuracy between different negative sample images dataset collected by semantic features. We adopted Image Net as image dataset in this study, and we adopted Word Net for building semantic hierarchical tree. Semantic hierarchical structure tree is adopted to calculate the distance between each node. Then we adopt this distance relationship to prepare positive and negative sample images. We prepare one baseline method and suggest six different negative sample images selection methods for experiment. The binary SVM classifier training and prediction is implemented to compare the accuracy and Mean Reciprocal Rank (MRR) between baseline and each proposed method. Our results show that if we select uniform amount of negative sample images in each distance in the semantic hierarchical tree, we will achieve highest accuracy.",
    keywords = "Image Annotation, ImageNet, Machine Learning, Negative Sample Selection, SVM, WordNet",
    author = "Chan, {Shan Bin} and Hayato Yamana and Shin'Ichi Satoh",
    year = "2013",
    doi = "10.1109/SITIS.2013.37",
    language = "English",
    isbn = "9781479932115",
    pages = "162--167",
    booktitle = "Proceedings - 2013 International Conference on Signal-Image Technology and Internet-Based Systems, SITIS 2013",

    }

    TY - GEN

    T1 - A negative sample image selection method referring to semantic hierarchical structure for image annotation

    AU - Chan, Shan Bin

    AU - Yamana, Hayato

    AU - Satoh, Shin'Ichi

    PY - 2013

    Y1 - 2013

    N2 - When SVM is adopted for image annotation, we know that high quality sample images will improve image recognition accuracy. Images with the same visual/semantic features are adopted as positive sample images, and images with different visual/semantic features are adopted as negative sample images. But it is labor intensive in high quality sample images selection, especially when collecting by visual features. Most researchers randomly choose positive and negative sample images for classifier training. In many applications, adopting different negative sample image datasets will vary annotation accuracy. In this research, we will discuss the accuracy between different negative sample images dataset collected by semantic features. We adopted Image Net as image dataset in this study, and we adopted Word Net for building semantic hierarchical tree. Semantic hierarchical structure tree is adopted to calculate the distance between each node. Then we adopt this distance relationship to prepare positive and negative sample images. We prepare one baseline method and suggest six different negative sample images selection methods for experiment. The binary SVM classifier training and prediction is implemented to compare the accuracy and Mean Reciprocal Rank (MRR) between baseline and each proposed method. Our results show that if we select uniform amount of negative sample images in each distance in the semantic hierarchical tree, we will achieve highest accuracy.

    AB - When SVM is adopted for image annotation, we know that high quality sample images will improve image recognition accuracy. Images with the same visual/semantic features are adopted as positive sample images, and images with different visual/semantic features are adopted as negative sample images. But it is labor intensive in high quality sample images selection, especially when collecting by visual features. Most researchers randomly choose positive and negative sample images for classifier training. In many applications, adopting different negative sample image datasets will vary annotation accuracy. In this research, we will discuss the accuracy between different negative sample images dataset collected by semantic features. We adopted Image Net as image dataset in this study, and we adopted Word Net for building semantic hierarchical tree. Semantic hierarchical structure tree is adopted to calculate the distance between each node. Then we adopt this distance relationship to prepare positive and negative sample images. We prepare one baseline method and suggest six different negative sample images selection methods for experiment. The binary SVM classifier training and prediction is implemented to compare the accuracy and Mean Reciprocal Rank (MRR) between baseline and each proposed method. Our results show that if we select uniform amount of negative sample images in each distance in the semantic hierarchical tree, we will achieve highest accuracy.

    KW - Image Annotation

    KW - ImageNet

    KW - Machine Learning

    KW - Negative Sample Selection

    KW - SVM

    KW - WordNet

    UR - http://www.scopus.com/inward/record.url?scp=84894197935&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=84894197935&partnerID=8YFLogxK

    U2 - 10.1109/SITIS.2013.37

    DO - 10.1109/SITIS.2013.37

    M3 - Conference contribution

    SN - 9781479932115

    SP - 162

    EP - 167

    BT - Proceedings - 2013 International Conference on Signal-Image Technology and Internet-Based Systems, SITIS 2013

    ER -