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 Video

Updated: Jun 27, 2025

A Methodology for Capturing Joint Visual Attention Using Mobile Eye-Trackers
12:39

A Methodology for Capturing Joint Visual Attention Using Mobile Eye-Trackers

Published on: January 18, 2020

7.6K

MixFormer: End-to-End Tracking With Iterative Mixed Attention.

Yutao Cui, Cheng Jiang, Gangshan Wu

    IEEE Transactions on Pattern Analysis and Machine Intelligence
    |May 7, 2024
    PubMed
    Summary

    MixFormer, a novel transformer-based framework, unifies feature extraction and target integration for visual object tracking. This compact model achieves state-of-the-art performance across seven benchmarks.

    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

    Targeting Lcn2 to Inhibit Myocardial Cell Ferroptosis is a Potential Therapy for Alleviating Septic Cardiomyopathy.

    Inflammation·2025
    Same author

    Development and validation of a screening tool for sepsis without laboratory results in the emergency department: a machine learning study.

    EClinicalMedicine·2025
    Same author

    UPLC-MS metabolite profiling and antioxidant activity of <i>Sanghuangporus sanghuang</i> extract.

    PeerJ·2025
    Same author

    Mechanical properties and mechanisms of soda residue and fly ash stabilized soil.

    Scientific reports·2025
    Same author

    Angelica gigas Nakai (Korean Dang-gui) Root Alcoholic Extracts in Health Promotion and Disease Therapy - active Phytochemicals and In Vivo Molecular Targets.

    Pharmaceutical research·2025
    Same author

    Extension of an ICU-based noninvasive model to predict latent shock in the emergency department: an exploratory study.

    Frontiers in cardiovascular medicine·2025

    Area of Science:

    • Computer Vision
    • Machine Learning
    • Deep Learning

    Background:

    • Traditional visual object tracking relies on multi-stage pipelines.
    • These pipelines involve separate feature extraction, target information integration, and bounding box estimation.
    • Simplifying and unifying these processes is crucial for improved efficiency and performance.

    Purpose of the Study:

    • To introduce MixFormer, a compact tracking framework built upon transformers.
    • To unify feature extraction and target information integration using a novel Mixed Attention Module (MAM).
    • To investigate pre-training strategies and enhance online tracking capabilities.

    Main Methods:

    • Developed the Mixed Attention Module (MAM) for simultaneous feature extraction and target information integration.

    More Related Videos

    Methods to Test Visual Attention Online
    09:44

    Methods to Test Visual Attention Online

    Published on: February 19, 2015

    11.9K
    Measurement of Neurophysiological Signals of Ignoring and Attending Processes in Attention Control
    09:37

    Measurement of Neurophysiological Signals of Ignoring and Attending Processes in Attention Control

    Published on: July 5, 2015

    9.1K

    Related Experiment Videos

    Last Updated: Jun 27, 2025

    A Methodology for Capturing Joint Visual Attention Using Mobile Eye-Trackers
    12:39

    A Methodology for Capturing Joint Visual Attention Using Mobile Eye-Trackers

    Published on: January 18, 2020

    7.6K
    Methods to Test Visual Attention Online
    09:44

    Methods to Test Visual Attention Online

    Published on: February 19, 2015

    11.9K
    Measurement of Neurophysiological Signals of Ignoring and Attending Processes in Attention Control
    09:37

    Measurement of Neurophysiological Signals of Ignoring and Attending Processes in Attention Control

    Published on: July 5, 2015

    9.1K
  • Constructed MixFormer trackers by stacking MAMs, including hierarchical (MixCvT) and non-hierarchical (MixViT) variants.
  • Explored supervised, self-supervised, and Masked Autoencoder (MAE) pre-training, introducing the TrackMAE technique.
  • Designed an asymmetric attention scheme and score prediction module for efficient multi-template online tracking.

    • Main Results:

    • MixFormer trackers achieved state-of-the-art performance on seven challenging benchmarks (LaSOT, TrackingNet, VOT2020, GOT-10k, OTB100, TOTB, UAV123).
    • The MixViT-L model attained high AUC scores: 73.3% on LaSOT, 86.1% on TrackingNet, and 82.8% on TOTB.
    • Investigated and highlighted the distinct behaviors of supervised versus self-supervised pre-training.

    Conclusions:

    • The proposed MixFormer framework effectively unifies key visual object tracking processes.
    • The Mixed Attention Module (MAM) is a core component enabling simultaneous feature extraction and target integration.
    • MixFormer demonstrates superior performance and efficiency, setting new benchmarks in visual object tracking.