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: May 10, 2026

Photorealistic Learned Landscapes for Augmented Reality
06:54

Photorealistic Learned Landscapes for Augmented Reality

Published on: June 27, 2025

Efficient Large-Scale Scene Reconstruction via Semantic-Aware Hybrid Representation.

Husen Li, Jingyu Lin, Yudong Guo

    IEEE Transactions on Visualization and Computer Graphics
    |May 8, 2026
    PubMed
    Summary

    This study introduces a novel hybrid 3D scene representation, combining meshes and 3D Gaussians (3DGS), to efficiently reconstruct large-scale environments. This approach significantly reduces storage and speeds up rendering while maintaining high visual quality.

    Related Concept Videos

    State Space Representation01:27

    State Space Representation

    The frequency-domain technique, commonly used in analyzing and designing feedback control systems, is effective for linear, time-invariant systems. However, it falls short when dealing with nonlinear, time-varying, and multiple-input multiple-output systems. The time-domain or state-space approach addresses these limitations by utilizing state variables to construct simultaneous, first-order differential equations, known as state equations, for an nth-order system.
    Consider an RLC circuit, a...

    You might also read

    Related Articles

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

    Sort by
    Same author

    Dielectric Energy Storage Performance of Reductive Polyaniline/Polyethylenimine All-Organic Composite Films with Tunable Molecular Weight and Chain Structure.

    Polymers·2026
    Same author

    DTX3L Inhibits the EMT, Metastasis, and Stem-Like Features of Gastric Cancer Through Promoting GSK-3β Dependent SNAI1 Decay.

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

    Proteome-wide Mendelian randomization and colocalization analyses identify potential biomarkers for schizophrenia.

    Frontiers in psychiatry·2026
    Same author

    Characterization of pulmonary tuberculosis in high-altitude region using nanopore targeted sequencing.

    BMC microbiology·2026
    Same author

    Bruceine A Ameliorates Osteoporosis Through Suppression of RACK1-Mediated Osteoclastogenesis.

    FASEB journal : official publication of the Federation of American Societies for Experimental Biology·2025
    Same author

    Trait-based divergence in free-living and particle-attached bacterial succession during algal biomass degradation.

    Environmental research·2025

    Area of Science:

    • Computer Vision
    • Computer Graphics
    • 3D Reconstruction

    Background:

    • Reconstructing large-scale 3D scenes faces challenges in balancing photorealism, real-time rendering, and storage efficiency.
    • Current 3D Gaussian Splatting (3DGS) methods achieve high fidelity but struggle with excessive primitives for large scenes, increasing costs.

    Purpose of the Study:

    • To develop an efficient hybrid representation for large-scale 3D scene reconstruction.
    • To overcome the storage and rendering limitations of existing 3DGS techniques.

    Main Methods:

    • Introduced a semantic-guided hybrid representation unifying textured meshes and 3D Gaussians in a differentiable framework.
    • Developed a semantic-guided adaptive modeling pipeline using multi-view segmentation for scene partitioning and Gaussian pruning.

    Related Experiment Videos

    Last Updated: May 10, 2026

    Photorealistic Learned Landscapes for Augmented Reality
    06:54

    Photorealistic Learned Landscapes for Augmented Reality

    Published on: June 27, 2025

  • Implemented a CUDA-based hybrid renderer for joint optimization and occlusion handling.
  • Proposed a mesh-guided sampling strategy to add Gaussians for fine detail recovery.

    • Main Results:

    • Significantly reduced storage requirements compared to traditional methods.
    • Accelerated rendering performance substantially.
    • Maintained or improved visual quality on diverse large-scale datasets.

    Conclusions:

    • The proposed hybrid representation offers an efficient solution for large-scale 3D scene reconstruction.
    • This method effectively balances quality, speed, and storage for complex environments.