BIROn - Birkbeck Institutional Research Online

    Do not lose the details: reinforced representation learning for high performance visual tracking

    Wang, Q. and Zhang, M. and Xing, J. and Gao, J. and Hu, W. and Maybank, Stephen Do not lose the details: reinforced representation learning for high performance visual tracking. 27th International Joint Conference on Artificial Intelligence , (In Press)

    [img]
    Preview
    Text
    VisualTrackingIJCAI2018.pdf - Author's Accepted Manuscript

    Download (1MB) | Preview

    Abstract

    This work presents a novel end-to-end trainable CNN model for high performance visual object tracking. It learns both low-level fine-grained representations and a high-level semantic embedding space in a mutual reinforced way, and a multi-task learning strategy is proposed to perform the correlation analysis on representations from both levels. In particular, a fully convolutional encoder-decoder network is designed to reconstruct the original visual features from the semantic projections to preserve all the geometric information. Moreover, the correlation filter layer working on the fine-grained representations leverages a global context constraint for accurate object appearance modeling. The correlation filter in this layer is updated online efficiently without network fine-tuning. Therefore, the proposed tracker benefits from two complementary effects: the adaptability of the fine-grained correlation analysis and the generalization capability of the semantic embedding. Extensive experimental evaluations on four popular benchmarks demonstrate its state-of-the-art performance.

    Metadata

    Item Type: Article
    Additional Information: IJCAI-ECAI-18, Stockholm, Sweden, July 13-19, 2018
    School: Birkbeck Schools and Departments > School of Business, Economics & Informatics > Computer Science and Information Systems
    Depositing User: Stephen Maybank
    Date Deposited: 21 May 2018 13:47
    Last Modified: 21 May 2018 13:47
    URI: http://eprints.bbk.ac.uk/id/eprint/22481

    Statistics

    Downloads
    Activity Overview
    34Downloads
    116Hits

    Additional statistics are available via IRStats2.

    Archive Staff Only (login required)

    Edit/View Item Edit/View Item