Eurographics Digital Library
Not a member yet
17121 research outputs found
Sort by
Necessary but not Sufficient: Limitations of Projection Quality Metrics
No full textHigh-dimensional data analysis often uses dimensionality reduction (DR, also called projection) to map data patterns to human-digestible visual patterns in a 2D scatterplot. Yet, DR methods may fail to show true data patterns and/or create visual patterns that do not represent any data patterns. Projection Quality Metrics (PQMs) are used as objective measures to gauge the above process: the higher a projection's scores in PQMs, the more it is deemed faithful to the data it represents. We show that, while PQMs can be used as exclusion criteria - low values usually mean poor projections - the converse does not always hold. For this, we develop a technique to automatically generate projections that score similar or even higher PQM values than projections created by well-known techniques, but show different, often confusing, visual patterns. Our results show that accepted PQMs cannot be used as an exclusive way to tell whether a projection yields accurate and interpretable visual patterns - in this sense, PQMs play a role akin to that of summary statistics in exploratory data analysis. We also show that not all studied metrics can be fooled equally well, suggesting a ranking of metrics in their ability to reliably capture quality.Computer Graphics ForumDimensionality Reduction and High-Dimensional Dat
Nodes, Edges, and Artistic Wedges: A Survey on Network Visualization in Art History
No full textArt history traditionally relies on qualitative methods. However, the increasing availability of digitized archives has opened new possibilities for research by integrating visual analytics. This survey presents a comprehensive review of the intersection between art history and visual analytics, focusing on network visualization and how it supports researchers in analyzing and understanding complex art historical relationships through nodes (e.g., artists, artworks, institutions) and edges (the relationships between them). We explore how these approaches enable dynamic analysis, offering novel perspectives on artistic influence, stylistic evolution, and social interactions within the art world. Through this, we also examine wedges, a metaphor for the friction often present in art history between individuals and institutions. These tensions, which have historically played a pivotal role in shaping artistic movements, are now better understood through the lens of network visualization, revealing how conflicts and power dynamics influenced the development of art. Through a hierarchical categorization of the literature, we outline saturated problems and research areas as well as ongoing challenges in art historical research. Furthermore, we highlight the potential of visual analytics to bridge the gap between traditional qualitative research and modern computational analysis, offering interactive exploration, temporal analysis, and complex network visualization. We provide a structured foundation for future research in art history, emphasizing the value of network visualization in enriching the understanding of art history.Computer Graphics ForumDomain-Specific Visualization Application
Lifelike Motions for Robotic Characters
No full textHumanoids have made significant advances in recent years. Nonetheless, the motions they perform often remain rigid, mechanical, and lack the diversity and expressiveness of human motion. This stands in stark contrast to physics-based simulated characters, which are capable of performing agile and lifelike motions in fully simulated environments. Such characters typically leverage reinforcement learning in combination with motion capture data to learn how to move like humans. However, their success is closely tied to unrealistic modeling assumptions such as simplified dynamics, overpowered actuators, or noise-free sensing. While these assumptions enable efficient and stable training, they hinder the transfer to the real world. In the real world, there are no shortcuts. To achieve more dynamic motions for humanoids, physically accurate simulation and robust learning methods are essential. This requires rethinking many components along the pipeline, starting from the simulators and how to account for sim-to-real gaps, up to questions about how to represent, track, and generate motions for humanoids. In this dissertation, we present several contributions in this direction and bring more lifelike motions to robotic characters. First, we present a learning-based modular simulation augmentation to reduce the sim-to-real gap. Our method can generalize across robot configurations and helps to better estimate the state of the robot. In a second contribution, we propose a novel architecture for encoding motions as a trajectory in latent space. The architecture overcomes the need for absolute positional encoding, leading to better reconstruction quality of various sequential data types. In a third contribution, we show how a pretrained latent space can be leveraged to train more accurate and robust control policies using reinforcement learning. Our two-stage method transfers to the real world and brings dynamic dancing motions to a humanoid robot. Our last contribution physically aligns kinematic motion generators with the capabilities of the character and its control policy. This allows for a more successful transfer of generated motions to the real world. The methods and concepts introduced in this dissertation make robots move more lifelike and reduce the gap to simulated characters. We hope they will inspire future research and bring more believable robots into our world.EG Graphics Dissertation Onlin
Breaking the Single-Stage Barrier: Synergistic Data-Model Adaptation at Test-Time for Medical Image Segmentation
No full textDomain shift, predominantly caused by variations in medical imaging across different institutions, often leads to a decline in the accuracy of medical image segmentation models. While Test-Time Adaptation (TTA) holds promise to address this issue, existing methods exhibit significant limitations: model adaptation is prone to error accumulation and catastrophic forgetting in continuous domain learning. Meanwhile, data adaptation struggles to achieve deep latent alignment due to the inaccessibility of source domain data. To address these challenges, we propose Synergistic Data-Model Adaptation (SDMA), which innovatively leverages Batch Normalization (BN) layers as a bidirectional bridge to enable a two-stage joint adaptation process. In the data adaptation stage, domain-aware prompts dynamically adjust the BN statistics of incoming test data, achieving low-level distribution alignment in the Fourier space. In the model adaptation stage, we dynamically optimize the BN affine parameters based on strong-weak data augmentation and entropy minimization, enabling adaptation to high-level semantic features. Experiments conducted on five retinal fundus image datasets from various medical institutions demonstrate that our method achieves an average Dice improvement of 1.23% over previous state-of-the-art (SOTA) methods, establishing a new SOTA performance.Pacific Graphics Conference Papers, Posters, and DemosDetecting & Estimating from image
Exploring the Use of Auditory Feedback as a Guide for 3D Drawing in Extended Reality
No full text3D drawing (or sketching) in Extended Reality (XR) is more difficult to master than 2D drawing. It is performed in mid-air and does not benefit from a physical surface, has more degrees of freedom, and is often dependent on Head-Mounted Displays that contain inaccurate stereoscopic rendering (due to the vergence-accommodation conflict). Many approaches were explored in the scientific literature in order to compensate for these issues, such as visual guides, the use of haptic feedback, or beautifying techniques. However, very few focus on the use of auditory feedback as a possible guide for increasing spatial awareness, and thus accuracy, while drawing inside a 3D space, although sound has proven to be a useful additional feedback in several 3D interactive contexts. In this paper, we explored several auditory feedback (based on the depth of the controller and the proximity to other strokes in the canvas) to try and see if it improves the accuracy of drawings in XR. To do so, we conducted an experiment with 21 participants, mostly novices in drawing. The results show that the use of the proposed auditory feedback do not have an influence on the accuracy of the drawings, but do have an influence on the participants' perceived accuracy and the confidence in their ability to perform the task.ICAT-EGVE 2025 - International Conference on Artificial Reality and Telexistence and Eurographics Symposium on Virtual EnvironmentsSoun
Computer Graphics Instructors' Intentions for Using Generative AI for Teaching
No full textBackground: Generative AI has significant potential to support learning processes, such as generating personalized content matching individual student needs. It also has the potential to support teaching processes by assisting instructors in generating content, assessing students, or supporting practice. This study investigates how computer graphics instructors have used generative AI or are planning to use generative AI to support their teaching. We implemented an anonymous online survey based on the Unified Theory of Acceptance and Use of Technology (UTAUT) methodology and distributed it among Eurographics members. The research questions were: (1) What are computer graphics instructors' ways of integrating generative AI for teaching and learning purposes? (2) What are the influencing factors computer graphics instructors have considered for integrating generative AI for teaching and learning purposes? Results: Between October 2024 and January 2025, we received 12 responses. Findings suggest that while some instructors have integrated generative AI into some aspects of their teaching, others have not and are hesitant to adopt them in the future, particularly as related to generating content for creating assignments such as lecture notes, summaries, teaching examples, etc., and supporting their assessment processes such as providing feedback, evaluating assignments, or grading exams. However, instructors were more open to using generative AI to support their teaching practices, particularly as related to pedagogy, such as providing students with interactive practice problems and supporting their creative content generation. Conclusion: Findings from the study identified the level of acceptance among computer graphics instructors, primarily full professors, and their experiences and intentions for using generative AI. To get a better understanding of the adoption of generative AI in the field of computer graphics education, we would like to invite the community to share their experiences and future intentions via the survey, which will remain open for additional input.Eurographics 2025 - Education PapersEducation
Importance Sampling of BCSDF Derivatives
No full textDifferentiable rendering requires the development of importance sampling for derivative functions with respect to the parameters. While importance sampling for Bidirectional Reflectance Distribution Function derivative has been proposed in recent years, no methods have been introduced for the derivatives of Bidirectional Curve Scattering Distribution Function (BCSDF). To bridge this gap, we propose an importance sampling method for the derivatives of the BCSDF using positivization [BXB∗24]. Our BCSDF derivative importance sampling method achieves up to 94% reduction in RMSE for eqaul-time rendering.Eurographics 2025 - Short PapersShort Paper
Preventive and planned conservation: an algorithm for the analysis and evaluation of degradation phenomena in Cultural Heritage
No full textThis research presents an algorithm for the assessment of deterioration risks and the planning of preventive conservation actions. The proposed system is based on systematic collection and analysis of key parameters, including the type of support, conservation history, restoration techniques, environmental conditions and other factors that may influence the material stability of Cultural Heritage over time. The methodological approach involves several phases: data collection, risk factor analysis, assignment of reference values to each significant variable and graphic visualization of conservation status and associated risk level through chromatic gradients. The algorithm enables the calculation of an overall risk score through a weighted evaluation system that integrates mathematical equations with a scoring approach reflecting the relative importance of each parameter. Currently implemented as an Excel®-based application, the system is designed to be accessible to restorers and conservation professionals in interoperability with integrated advanced digital ecosystems/Information systems (BIM/HBIM-CAD and VPL). Visual parametric representation facilitates communication of complex information. As a dynamic and scenario-driven tool, the algorithm requires regular updates to maintain its effectiveness across varying environmental and conservation contexts. The algorithm has been tested on a sample of artworks restored and monitored over the past five years, including significant works by Titian, Lorenzo Lotto and Luca Giordano. These case studies, museum collections, historic interiors and open-air monuments, have allowed for the verification of the system's ability to detect vulnerabilities and support tailored preventive conservation strategies. All data, analyses, and results are systematically recorded and managed within a dedicated information framework to ensure traceability and continuity of conservation management. The experimental application has confirmed the validity of the structured data-driven approach in identifying critical issues, prioritizing conservation actions and supporting informed decision-making.Digital HeritageInstrumental and Computational Approaches for CH Conservation and Restauratio
Characterization of Games Technologies for Learning in the Context of Intangible Cultural Heritage
No full textThis paper explores the application of Game Technologies for Learning (GT4L), which includes Serious Games, Gamification, and Game-Based Learning in the context of Intangible Cultural Heritage (ICH). According to the United Nations and the Sustainable Development Goals (SDGs), in particular, target 4 of SDG 11, this study emphasizes the importance of preserving ICH, such as oral traditions, performing arts, and traditional craftsmanship, in an increasingly globalized world. A systematic literature review of 34 articles was conducted to characterize the use of GT4L in ICH. The review aimed to identify the ICH domains where GT4L is applied, the types of GT4L used, the target audiences, the technological aspects considered, the innovations integrated, and how learning styles and player typologies are incorporated. The findings indicate that GT4L is utilized across several ICH domains, including oral traditions, performing arts, social practices, and traditional craftsmanship. Most games are Serious Games in the genres of simulation of arts and sports, adventure, treasure, puzzle, trivia, and role-playing. The target audiences range from children to heritage professionals. The technologies employed include mobile applications, immersive virtual and augmented reality experiences, body-tracking, and 3D environments, all of which incorporate user-centered and participatory design methodologies. However, several innovative aspects are absent that could enhance the impact of these applications. These include the incorporation of accessibility criteria to ensure equitable access to games for all players, the implementation of recommender systems to guide choices based on user profiles, as well as considerations for different learning styles and player types, among others.Digital HeritageModern Technologies for Serious Gaming in Cultural Heritag
Reliable Visual Analytics with Dimensionality Reduction: Quality Evaluation and Interpretation of Projections
No full textDimensionality reduction (DR) is widely used for visual analytics, but the insights obtained from these visualizations may often be unreliable. For example, DR projections distort the intrinsic structure of high-dimensional data in ways that may not be obvious at first glance, potentially leading analysts to inaccurate interpretations. Even reliable visual patterns may be hard to interpret regarding what exactly they convey about the underlying data, due to the often severe compression from hundreds (or thousands) of dimensions down to the visual space. In this tutorial, we discuss how to enhance the reliability of visual analytics with DR by focusing on two perspectives: quality evaluations and interpretations. While the former helps users identify or create projections with fewer distortions, the latter provides a reliable method for deriving insights from those projections. By combining lecture and coding exercises, we expect our tutorial to provide a grounded basis for audiences to use DR in a more reliable mannerEuroVis 2025 - Panels and TutorialsTutorial