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

Fast asymmetric learning for cascade face detection.

Jianxin Wu1, S Charles Brubaker, Matthew D Mullin

  • 1School of Interactive Computing, College of Computing, Georgia Institute of Technology, Atlanta, GA 30332-0760, USA. wujx@cc.gatech.edu

IEEE Transactions on Pattern Analysis and Machine Intelligence
|January 16, 2008
PubMed
Summary
This summary is machine-generated.

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

<i>LsToll</i> Gene Mediates Antibacterial Immunity and Developmental Regulation in <i>Loxostege sticticalis</i>.

Insects·2026
Same author

Long-Term Outcomes From a Decade of Donation After Circulatory Death Heart Transplantation in Australia.

JACC. Heart failure·2026
Same author

Rapid Determination of Several Biogenic Amines in Cold-Chain Fish Samples by Portable Ion Trap Mass Spectrometry with Nano-Electrospray Ionization.

Foods (Basel, Switzerland)·2026
Same author

Aerosol Inhalation of a Recombinant H7N9 Hemagglutinin Antigen Elicits Systemic and Mucosal Immune Responses in Mice.

Viruses·2026
Same author

Localized hybridization chain reaction on self-assembled DNA nanospheres overcomes the kinetics-stability-complexity trilemma in miRNA sensing.

Biosensors & bioelectronics·2026
Same author

Gastrointestinal dysfunction in patients with Mowat-Wilson syndrome is associated with feeding difficulties and altered plasma neurotransmitters.

Frontiers in pediatrics·2026
Same journal

HardFlow: Hard-Constrained Sampling for Flow-Matching Models Via Trajectory Optimization.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Industrial Brain: Self-Evolving Neuro-Symbolic Autonomy with Causal Resilience for Cyber-Physical Systems.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Adaptive Hardness-Driven Dictionary Distillation for Incomplete Streaming View Clustering.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Mixture of Global and Local Experts with Diffusion Transformer for Controllable Face Generation.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Task-KV: Task-aware KV Cache Optimization via Semantic Differentiation of Attention Heads.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Achieving Text-based Person Retrieval with Any Granularity.

IEEE transactions on pattern analysis and machine intelligence·2026
See all related articles

This study introduces a new method for designing cascade face detectors by decoupling feature selection and classifier design. This approach, using the Linear Asymmetric Classifier (LAC), significantly improves face detection performance and reduces training time.

Area of Science:

  • Computer Vision
  • Machine Learning
  • Pattern Recognition

Background:

  • Cascade face detectors rely on sequential node classifiers.
  • Existing methods often combine feature selection and ensemble classifier formation.
  • Asymmetric learning goals in face detection present significant challenges.

Purpose of the Study:

  • To propose a novel approach for designing node classifiers in cascade face detectors.
  • To address the difficulties arising from asymmetric learning goals in face detection.
  • To improve the overall performance of ensemble classifiers in face detection systems.

Main Methods:

  • Categorization of asymmetries in the learning goal for face detection.
  • Development of the Forward Feature Selection (FFS) algorithm.

Related Experiment Videos

  • Introduction of a fast pre-computing strategy for AdaBoost.
  • Design of the Linear Asymmetric Classifier (LAC) as a constrained optimization problem.

    • Main Results:

    • FFS and fast AdaBoost reduce training time by approximately 100x and 50x, respectively.
    • LAC explicitly handles asymmetric learning goals.
    • Experimental results demonstrate improved ensemble classifier performance using LAC.

    Conclusions:

    • Decoupling feature selection and classifier design offers advantages in cascade face detection.
    • The proposed FFS algorithm and fast AdaBoost significantly accelerate training.
    • The LAC classifier effectively addresses asymmetric learning goals, leading to enhanced face detection accuracy.