Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Real-time object tracking via online discriminative feature selection.

Kaihua Zhang, Lei Zhang, Ming-Hsuan Yang

    IEEE Transactions on Image Processing : a Publication of the IEEE Signal Processing Society
    |August 20, 2013
    PubMed
    Summary

    This study introduces a new supervised learning method for object tracking that effectively handles visual drift by using prior instance information. The approach is more efficient and robust than existing multiple instance learning (MIL) trackers.

    Related Concept Videos

    You might also read

    Related Articles

    Articles linked to this work by shared authors, journal, and citation graph.

    Sort by
    Same author

    Fatty Acid Binding Protein 5 Mediates Astrocytic Pyroptosis and Neuroinflammation in Epilepsy via cGAS/STING Pathway.

    Advanced science (Weinheim, Baden-Wurttemberg, Germany)·2026
    Same author

    Wavelength-Selective Thermal Regulators with Dual-Band Radiative Dissipation for Multispectral Camouflage.

    ACS applied materials & interfaces·2026
    Same author

    Frequency-Aware Refinement Network with Multi-Scale Fusion for Remote Sensing Change Detection.

    Sensors (Basel, Switzerland)·2026
    Same author

    Osteosarcoma pulmonary metastasis: the hemodynamic mechanical-filtration hypothesis for subpleural predominance.

    Frontiers in oncology·2026
    Same author

    Altered functional connectivity of emotional circuits and default mode network in postpartum women: a resting-state functional magnetic resonance imaging study.

    BMC psychology·2026
    Same author

    Spatial-Frequency Decoupling Alignment Encoding for Remote Sensing Change Detection.

    Sensors (Basel, Switzerland)·2026

    Area of Science:

    • Computer Vision
    • Machine Learning
    • Artificial Intelligence

    Background:

    • Tracking-by-detection algorithms often struggle with visual drift due to noisy samples.
    • Multiple Instance Learning (MIL) has been used to address this, but can be complex.
    • Prior instance label information is often underutilized.

    Purpose of the Study:

    • To develop a more effective and efficient method for handling visual drift in object tracking.
    • To leverage prior instance information within a supervised learning framework.
    • To improve the robustness of tracking algorithms.

    Main Methods:

    • Developed a novel formulation integrating prior instance labels and current frame tracking results.
    • Implemented an online discriminative feature selection algorithm.

    Related Experiment Videos

  • Optimized the objective function using steepest ascent for positive and descent for negative samples.

    • Main Results:

    • The proposed supervised learning approach effectively handles visual drift.
    • The algorithm demonstrates superior robustness and efficiency compared to existing MIL trackers.
    • Experimental evaluations on challenging sequences validate the proposed method's performance.

    Conclusions:

    • Integrating prior instance information into supervised learning offers a simpler and more effective solution for visual drift than MIL.
    • The developed online discriminative feature selection enhances tracker robustness and efficiency.
    • The proposed algorithm represents a significant advancement in object tracking technology.