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

In-vehicle low-light face image enhancement with physical-semantic constrained diffusion and gated selective

Pancheng Zhang1, Zhe Chen1, Yihui Hu1

  • 1School of Information Engineering, Chang'an University, Xi'an, 710064, Shaanxi, China.

Neural Networks : the Official Journal of the International Neural Network Society
|June 1, 2026
PubMed
Summary

Related Concept Videos

Light Acquisition02:16

Light Acquisition

In order to produce glucose, plants need to capture sufficient light energy. Many modern plants have evolved leaves specialized for light acquisition. Leaves can be only millimeters in width or tens of meters wide, depending on the environment. Due to competition for sunlight, evolution has driven the evolution of increasingly larger leaves and taller plants, to avoid shading by their neighbors with contaminant elaboration of root architecture and mechanisms to transport water and nutrients.

You might also read

Related Articles

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

Sort by
Same author

PDA@MnO<sub>x</sub> nanozyme-engineered mesenchymal stem cells for severe acute pancreatitis therapy via modulation of ROS-inflammation-microbiota homeostasis.

Journal of nanobiotechnology·2026
Same author

Development of a nursing-sensitive quality evaluation indicator system for perioperative care in breast reconstruction surgery: A Delphi study.

Asia-Pacific journal of oncology nursing·2026
Same author

Design and Evaluation of Universal Multi-Epitope Vaccines Against Vibrio Infections Using a Single/Multi-Target Multi-Derived Epitope Strategy.

Journal of fish diseases·2026
Same author

N-doping enhanced catalytic mechanism of Co-loaded biochar cathode for degrading organic phosphorus flame retardants in heterogeneous electro-Fenton system.

Journal of colloid and interface science·2025
Same author

Noncontact ultrasonic materials identification based on improved frequency responses.

Ultrasonics·2025
Same author

Axial enhancement on non-radicals generation in PMS activation process mediated with Co single-atom catalyst encapsulated nanoparticles.

Journal of hazardous materials·2025

This study introduces STDN-MD, a novel framework for driver fatigue monitoring in nighttime conditions. It enhances visual fidelity and machine perception accuracy by synergistically restoring physical and semantic information, overcoming limitations of existing methods.

Area of Science:

  • Computer Vision
  • Artificial Intelligence
  • Automotive Safety

Background:

  • Nighttime driver fatigue monitoring is challenged by non-uniform illumination and sensor noise.
  • Existing algorithms struggle with feature smoothing and semantic hallucinations.
  • A need exists for robust algorithms addressing these specific nighttime challenges.

Purpose of the Study:

  • To propose STDN-MD, a synergistic restoration framework for nighttime driver fatigue monitoring.
  • To improve visual fidelity and machine perception accuracy in challenging low-light conditions.
  • To overcome limitations of discriminative and generative models in driver monitoring.

Main Methods:

  • Developed STDN-MD, a physical-semantic synergistic restoration framework based on Retinex theory.
Keywords:
Diffusion modelsDriver monitoring systemLow-light image enhancementPhysical-semantic synergyState space models

Related Experiment Videos

  • Introduced a Gated Recurrent 2D Selective Scan (GR-SS2D) module to mitigate noise propagation.
  • Utilized face parsing masks as empirical priors in a conditional diffusion process for reflectance restoration.

    • Main Results:

    • STDN-MD achieved an LPIPS of 0.068 on the YaWDD-Dark benchmark, a 52.8% reduction in perceptual error.
    • The method significantly improved visual perception and preserved key fatigue features.
    • Demonstrated superior performance in both visual fidelity and machine perception accuracy.

    Conclusions:

    • STDN-MD offers a robust solution for nighttime driver fatigue monitoring.
    • The framework effectively addresses challenges of illumination and noise, enhancing safety.
    • Achieved dual breakthroughs in visual fidelity and perception accuracy for intelligent cockpits.