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: Jul 9, 2026

End-To-End Deep Neural Network for Salient Object Detection in Complex Environments
03:31

End-To-End Deep Neural Network for Salient Object Detection in Complex Environments

Published on: December 15, 2023

Boosting deep detector efficiency and robustness through detection discriminant reorganization and compression.

Jung Im Choi1, Qizhen Lan2, Qing Tian2

  • 1Dept. of Computer Science, Bowling Green State University, Ohio, USA.

Neural Networks : the Official Journal of the International Neural Network Society
|July 7, 2026
PubMed
Summary

Related Concept Videos

Difference from Background: Limit of Detection01:05

Difference from Background: Limit of Detection

The limit of detection (LOD) is the smallest amount of analyte that can be distinguished from the background noise. The LOD value corresponds to the concentration at which the analyte signal is three times larger than the standard deviation of the blank signal. Below this value, the analyte signal cannot be differentiated from the background noise. It is calculated by dividing the calibration slope by 3 times the standard deviation of the blank signals.
The LOD indicates the presence or absence...

You might also read

Related Articles

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

Sort by
Same author

<i>PPAN</i> modulates mouse male germ cell development via maintaining nucleolar homeostasis.

Genes & diseases·2024
Same author

Effect of mixed probiotics on pulmonary flora in patients with mechanical ventilation: an exploratory randomized intervention study.

European journal of medical research·2024
Same author

A staged approach using machine learning and uncertainty quantification to predict the risk of hip fracture.

Bone reports·2024
Same author

Clinical value of peripheral blood miR-21 and miR-486 combined with CT forearly cancer diagnosis in pulmonary nodulessmoking.

Journal of cardiothoracic surgery·2024
Same author

Application Value and Safety Analysis of Warfarin, Rivaroxaban, and Dabigatran Ester in Elderly Patients With Atrial Fibrillation.

Clinical cardiology·2024
Same author

An atlas of genetic effects on the monocyte methylome across European and African populations.

medRxiv : the preprint server for health sciences·2024

We introduce a new framework, proactive deep detection discriminant analysis (D3A), to make deep visual detectors more efficient and robust against adversarial attacks. This method improves model compactness and reduces latency for better real-world deployment.

Area of Science:

  • Computer Vision
  • Deep Learning
  • Machine Learning

Background:

  • Deep neural networks (DNNs) face challenges in resource-constrained and safety-critical applications due to large model sizes and susceptibility to adversarial attacks.
  • Existing research often addresses efficiency or adversarial robustness separately, with limited focus on detection tasks, particularly concerning simultaneous improvements.

Purpose of the Study:

  • To propose a novel framework, proactive deep detection discriminant analysis (D3A), that enhances both inference efficiency and adversarial robustness of deep visual detectors.
  • To address the limitations of current methods by focusing on detection tasks and tackling efficiency and robustness concurrently during training.

Main Methods:

  • Developed the D3A framework, a training-time approach that actively extracts and condenses detection utility into fewer dimensions for structured pruning.
Keywords:
Adversarial robustnessDeep detection discriminant analysisDeep visual detector compression

Related Experiment Videos

Last Updated: Jul 9, 2026

End-To-End Deep Neural Network for Salient Object Detection in Complex Environments
03:31

End-To-End Deep Neural Network for Salient Object Detection in Complex Environments

Published on: December 15, 2023

  • Aligned condensed dimensions with latent filters to remove vulnerable components and mitigate misalignment between classification and localization subtasks.
  • Integrated D3A with adversarial training strategies to improve detector robustness.

    • Main Results:

    • Demonstrated improved inference efficiency (model compactness, reduced latency) and enhanced adversarial robustness across various visual detectors.
    • Achieved superior performance compared to state-of-the-art compression methods on KITTI and COCO datasets.
    • The compressed YOLOX-S detector on KITTI showed a 1.9% mAP increase on clean images with 20% fewer parameters, alongside robustness gains of up to 11.4% and 24.8% under standard and adversarial training.

    Conclusions:

    • The D3A framework effectively boosts inference efficiency and adversarial robustness in deep visual detectors without compromising performance.
    • The training-time approach offers a proactive solution to model size and vulnerability issues, particularly beneficial for complex detection tasks.
    • D3A provides a viable method for developing more deployable and secure deep learning models for computer vision applications.