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

Evaluating 'Graphical Perception' with CNNs.

Daniel Haehn, James Tompkin, Hanspeter Pfister

    IEEE Transactions on Visualization and Computer Graphics
    |August 24, 2018
    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

    OMAMA-DB: the Oregon-Massachusetts Mammography Database.

    Journal of medical imaging (Bellingham, Wash.)·2026
    Same author

    Machine learning for N-dimensional spatial reasoning tasks on the web.

    Frontiers in bioinformatics·2026
    Same author

    Clocks and Dominoes: Timing Mechanisms of Embryogenesis.

    bioRxiv : the preprint server for biology·2026
    Same author

    Task-Specific Directions: Definition, Exploration, and Utilization in Parameter Efficient Fine-Tuning.

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

    Explainable Machine-Learning Model to Classify Culprit Calcified Carotid Plaque in Embolic Stroke of Undetermined Source.

    Journal of neuroimaging : official journal of the American Society of Neuroimaging·2026
    Same author

    SmartEM: machine learning-guided electron microscopy.

    Nature methods·2025
    Same journal

    MesoSplats: Texture Synthesis with Gaussian Splatting.

    IEEE transactions on visualization and computer graphics·2026
    Same journal

    GLLA: A Unified Force-Directed Graph Layout Framework Supporting Local Adjustments.

    IEEE transactions on visualization and computer graphics·2026
    Same journal

    Multi-Perception Crowd: Learning to combine entity and implicit perception for diverse crowd simulation.

    IEEE transactions on visualization and computer graphics·2026
    Same journal

    Hiding in Plain Sight: Camouflaging Real-world Objects.

    IEEE transactions on visualization and computer graphics·2026
    Same journal

    RTF2Mesh: Restricted Tangent Face Based Mesh Compression With Neural Displacement Fields.

    IEEE transactions on visualization and computer graphics·2026
    Same journal

    Practical Occluder Generation for Mobile Games.

    IEEE transactions on visualization and computer graphics·2026
    See all related articles

    Convolutional neural networks (CNNs) show varied performance on graphical perception tasks, sometimes matching human efficiency. However, CNNs are not yet a reliable model for understanding human graphical perception in data visualization.

    Area of Science:

    • Computer Vision
    • Data Visualization
    • Human-Computer Interaction

    Background:

    • Convolutional neural networks (CNNs) excel at image-based computer vision tasks.
    • The application of CNNs to graphical perception tasks, crucial for data visualization, remains underexplored.
    • Understanding CNNs' performance in visualization is key to advancing AI in data analysis.

    Purpose of the Study:

    • To evaluate the graphical perceptual capabilities of CNNs on visualization tasks.
    • To compare CNN performance against human perception baselines established by Cleveland and McGill.
    • To identify how and why CNNs succeed or fail in interpreting data visualizations.

    Main Methods:

    • Reproduced Cleveland and McGill's 1984 experiments on visual encoding perception.

    Related Experiment Videos

  • Tested four CNN architectures on five distinct elementary perceptual tasks.
  • Established new human performance baselines for comparison.

    • Main Results:

    • CNNs demonstrated performance comparable to or exceeding human baselines in specific, limited scenarios.
    • Significant discrepancies were observed between CNN performance and human graphical perception.
    • CNNs' ability to generalize across different visual encodings and tasks was inconsistent.

    Conclusions:

    • Current CNNs are not suitable models for human graphical perception.
    • Further research is needed to bridge the gap between CNN capabilities and human visual processing in data visualization.
    • Findings provide insights into CNN limitations and potential for future AI development in visualization.