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

NVS-SQA: Exploring Self-Supervised Quality Representation Learning for Neurally Synthesized Scenes Without

Qiang Qu, Yiran Shen, Xiaoming Chen

    IEEE Transactions on Pattern Analysis and Machine Intelligence
    |October 30, 2025
    PubMed
    Summary

    Related Concept Videos

    Nonconscious Mimicry01:13

    Nonconscious Mimicry

    5.1K
    Nonconscious mimicry occurs when individuals alter their mannerisms to match the behaviors and expressions of those nearby, without intention.
    5.1K
    State Space Representation01:27

    State Space Representation

    515
    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...
    515

    You might also read

    Related Articles

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

    Sort by
    Same author

    Comparative Prognostic Assessment of EKFC Equations for Mortality Prediction: A Population-based Cohort Study.

    Cardiorenal medicine·2026
    Same author

    Pharmacokinetic/pharmacodynamic target attainment of eravacycline in a liver transplant recipient with refractory VRE <i>faecium</i> infection.

    Antimicrobial agents and chemotherapy·2026
    Same author

    Bibliometric Analysis of White Matter Disease and Novel Drug Therapy.

    Current neuropharmacology·2026
    Same author

    Instance-dependent Early Stopping for Adaptive Data Pruning.

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

    Strength-Adaptive Adversarial Training.

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

    Co-Boosting++: Coupled Optimization of Data and Ensemble for One-Shot Federated Learning.

    IEEE transactions on pattern analysis and machine intelligence·2026

    Neural View Synthesis (NVS) quality assessment is improved by NVS-SQA, a novel no-reference method. It learns quality representations via self-supervision, outperforming existing full-reference and no-reference approaches.

    Area of Science:

    • Computer Vision
    • Computer Graphics
    • Machine Learning

    Background:

    • Neural View Synthesis (NVS) methods like NeRF and 3D Gaussian Splatting generate photorealistic scenes.
    • Current quality assessment relies on full-reference metrics (PSNR, SSIM, LPIPS), which struggle with limited reference views in NVS.
    • Acquiring human perceptual labels for NVS is challenging, hindering dataset creation and model generalizability.

    Purpose of the Study:

    • To develop a no-reference Neural Synthesized Scene (NSS) quality assessment method (NVS-SQA).
    • To enable self-supervised learning of quality representations without human labels.
    • To overcome limitations of traditional self-supervised learning in the context of NSS quality assessment.

    Main Methods:

    • NVS-SQA employs self-supervision, heuristic cues, and quality scores as learning objectives.

    Related Experiment Videos

  • A specialized contrastive pair preparation process is utilized for effective learning.
  • The method avoids reliance on human perceptual labels and extensive datasets.

    • Main Results:

    • NVS-SQA significantly outperforms 17 no-reference methods across SRCC, PLCC, and KRCC metrics.
    • NVS-SQA surpasses 16 full-reference methods in performance on all evaluation metrics.
    • The proposed method demonstrates superior effectiveness and efficiency in NSS quality assessment.

    Conclusions:

    • NVS-SQA provides a robust and efficient solution for no-reference NSS quality assessment.
    • The self-supervised approach effectively learns quality representations, addressing data scarcity and label acquisition challenges.
    • NVS-SQA sets a new benchmark for evaluating neurally synthesized scenes.