Eurographics Digital Library
Not a member yet
17121 research outputs found
Sort by
KIN-FDNet：Dual-Branch KAN-INN Decomposition Network for Multi-Modality Image Fusion
No full textMulti-modality image fusion (MMIF) aims to integrate information from different source images to preserve the complementary information of each modality, such as feature highlights and texture details. However, current fusion methods fail to effectively address the inter-modality interference and feature redundancy issues. To address this issue, we propose an end-to-end dualbranch KAN-INN decomposition network (KIN-FDNet) with an effective feature decoupling mechanism for separating shared and specific features. It first employs a gated attention-based Transformer module for cross-modal shallow feature extraction. Then, we embed KAN into the Transformer architecture to extract low-frequency global features and solve the problem of low parameter efficiency in multi-branch models. Meanwhile, an invertible neural network (INN) processes high-frequency local information to preserve fine-grained modality-specific details. In addition, we design a dual-frequency cross-fusion module to promote information interaction between low and high frequencies to obtain high-quality fused images. Extensive experiments on visible infrared (VIF) and medical image fusion (MIF) tasks demonstrate the superior performance and generalization ability of our KIN-FDNet framework.Pacific Graphics Conference Papers, Posters, and DemosMulti-Modalit
A Semi-Implicit SPH Method for Compressible and Incompressible Flows with Improved Convergence
No full textIn simulating fluids using position-based dynamics, the accuracy and robustness depend on numerous numerical parameters, including the time step size, iteration count, and particle size, among others. This complexity can lead to unpredictable control of simulation behaviors. In this paper, we first reformulate the problem of enforcing fluid compressibility/incompressibility into an nonlinear optimization problem, and then introduce a semi-implicit successive substitution method (SISSM) to solve the nonlinear optimization problem by adjusting particle positions in parallel. In contrast to calculating an intermediate variable, such as pressure, to enforce fluid incompressibility within the position-based dynamics (PBD) framework, the proposed semiimplicit approach eliminates the necessity of such calculations. Instead, it directly employs successive substitution of particle positions to correct density errors. This method exhibits reduced dependency to numerical parameters, such as particle size and time step variations, and improves consistency and stability in simulating fluids that range from highly compressible to nearly incompressible. We validates the effectiveness of applying a variety of different techniques in accelerating the convergence rate.Computer Graphics ForumSimulating Complex Systems: Turbulent, Crowded, and Shattered44
VV-Studio: Volumetric-Video News and Storytelling on Mobile XR
No full textVolumetric video (VV) supports spatially immersive, human-centered storytelling by capturing 3D performances that convey gaze, posture, and subtle gestures. Mobile deployment remains constrained by large assets, limited bandwidth, and fragmented toolchains. VV-Studio is introduced as a lightweight XR demonstrator for commodity Android phones that enables browsing of VV assets, AR anchoring, and touch-based playback control. A complementary 60-120 s micro-podcast is generated via a Python backend that aggregates real-time weather and news, synthesizes speech (TTS), and routes audio to a neutral ''broadcaster'' avatar. The contribution is a pragmatic, reproducible mobile pipeline that couples VV placement with automated audio briefings, establishing a baseline for immersive journalism and education. The paper details architecture and methods, situates the work within the literature on VV capture/streaming and AR interaction.ICAT-EGVE 2025 - International Conference on Artificial Reality and Telexistence and Eurographics Symposium on Virtual Environments - Posters and DemosPosters and Demo
Visualizing Interaction Effects for Combinatorial Cost-Benefit Analysis
No full textInteraction effects occur when the combined impact of multiple actions differs from the sum of their individual effects. This creates challenges for scenarios that require analyzing how different combinations of actions affect an outcome of interest (i.e., combinatorial cost-benefit analysis). Visualization techniques support interpretation, but most existing approaches rely on multiseries line charts (interaction plots), which are widely used but do not explicitly support comparing interaction effects across alternative action sets. Accordingly, we investigate visualization approaches for analyzing interaction effects in combinatorial cost-benefit analysis. We propose a method integrating multi-attribute set rankings with small-scale visualizations to facilitate comparative analysis. Through a user study, we evaluate the effectiveness of three techniques for representing two- and threeway interactions. We present preliminary findings and discuss design implications to inform future visualization research.EuroVis 2025 - Short PapersTechniques and Tool
Accessible Text Descriptions for UpSet Plots
No full textData visualizations are typically not accessible to blind and low-vision (BLV) users. Automatically generating text descriptions offers an enticing mechanism for democratizing access to the information held in complex scientific charts, yet appropriate procedures for generating those texts remain elusive. Pursuing this issue, we study a single complex chart form: UpSet plots. UpSet Plots are a common way to analyze set data, an area largely unexplored by prior accessibility literature. By analyzing the patterns present in real-world examples, we develop a system for automatically captioning any UpSet plot. We evaluated the utility of our captions via semi-structured interviews with (N=11) BLV users and found that BLV users find them informative. In extensions, we find that sighted users can use our texts similarly to UpSet plots and that they are better than naive LLM usage.Computer Graphics ForumInclusive Visualizatio
Large Language Models for Museum Label Optimization: A Case Study on Ministerial Compliance and Cultural Accessibility in Galleria Sabauda's Flemish Collection
No full textThis paper examines how Large Language Models (LLMs) can be applied to optimize art museum labels to meet specific cultural accessibility needs and ministerial guidelines and best practices in matters of exhibit texts redaction. The paper illustrates the first results of an experiment on optimizing the current labels of the Flemish section of the Galleria Sabauda in Turin. The study tests three state-of-art-LLMs, GPT4.o mini, Claude 3.7 and DeepSeek V-3, according to three different few-shot prompting techniques. An evaluation grid is created to assess the performance of LLMs in this rewriting task considering both content optimization and readability and formatting aspects.Digital HeritageFrom DH Theory and Principles to Innovative Practice: Participation, Training, and Acces
A Natural Language Interface for the Visualization and Analysis of 3D Point Cloud Saliency Maps
No full textDeep learning models for analyzing 3D point clouds have been extensively researched. However, there is limited research on explainable artificial intelligence techniques for such models. Approaches relying on saliency maps seek to attribute model outputs to their input; nevertheless, their practical use for extracting meaningful insights on model behavior depends on user knowledge, available visualization tools, and ad hoc data analytics pipelines. In this work, we propose a system that combines a natural language interface, backed by a large language model, with a traditional 3D point cloud visualization interface as an interactive technique for visualizing and analyzing 3D point cloud saliency maps. Our system can 1) answer general questions in the context of 3D point cloud saliency maps, 2) answer data analytics-related questions about the 3D point cloud being analyzed and its associated saliency map, and 3) generate additional visualizations that aid in the extraction of meaningful insights from the 3D point cloud data. We demonstrate the proposed system's workflow through a use case study involving a sample input to a 3D point cloud semantic segmentation model containing color, ground truth, predicted class, curvature, and attribution values. Our results indicate that including a natural language interface facilitates the exploration of the data and enhances the efficiency of the data analytics process, enhancing the value provided by the 3D point cloud saliency maps. We release our code as open source at https://github.com/JorgeFCS/llm-3DpointCloud-viz.Computer Graphics and Visual Computing (CGVC)Visualisatio
Exploration of Delays and Tonality for LLM-powered Game Narrators
No full textAs Large Language Models (LLMs) are becoming more widely spread as part of games and interactive media, more research is needed in introducing them in areas where they can provide deeper immersion, interactivity, and customization. The role of LLMs as storytellers and narrators has the potential for just that however, not enough research has been done focusing on how users react to real-time storytelling by AI. In this research, we test several factors associated with integrating LLM narrators in an interactive experience. Namely, acceptable narration delay for intentional and unintentional user-triggered interactions, for exploration of an environment, and in connection with the importance of the objects in the environment, and finally, how these factors are connected to user perception of LLM narration tonality in a VR historical setting. As part of the results, we provide recommendations for acceptable delays in narration, connecting narration to objects, and the choice of tonality of the LLM.ICAT-EGVE 2025 - International Conference on Artificial Reality and Telexistence and Eurographics Symposium on Virtual EnvironmentsLarge Language Model (LLM
Exact and Efficient Mesh-Kernel Generation
No full textThe mesh kernel for a star-shaped mesh is a convex polyhedron given by the intersection of all half-spaces defined by the faces of the input mesh. For all non-star-shaped meshes, the kernel is empty. We present a method to robustly and efficiently compute the kernel of an input triangle mesh by using exact plane-based integer arithmetic to compute the mesh kernel. We make use of several ways to accelerate the computation time. Since many applications just require information if a non-empty mesh kernel exists, we also propose a method to efficiently determine whether a kernel exists by developing an exact plane-based linear program solver. We evaluate our method on a large dataset of triangle meshes and show that in contrast to previous methods, our approach is exact and robust while maintaining a high performance. It is on average two orders of magnitude faster than other exact state-of-the-art methods and often about one order of magnitude faster than non-exact methods.Computer Graphics ForumMesh Processing44
A Searchable Multimodal Dataset of Rococo-Era Ornamental Prints
No full textWe present a curated multimodal dataset and an accompanying multimodal retrieval system designed to promote reproducible research in art historical information access. The dataset consists of 1,605 digitized photographs of eighteenth-century original prints, with a specific focus on Rocaille ornamentation. Each image is paired with rich metadata as well as additional domain expert commentary. The multimodal retrieval system exposes this corpus through a search engine, implemented with a lightweight architecture. Semantic search is enabled by dense multimodal embeddings. Full-text and fuzzy queries are enabled by conventional database indices. Both types of queries can be easily made through a simple website, exposing the search engine. Our implementation also provides a simple, uniform, queryable REST API, which makes the collection easily and flexibly accessible to researchers with programming skills. Emphasizing scalability and extensibility, the platform can serve as a practical blueprint for deploying multimodal search across specialized image-text datasets. Note that this paper describes work-in-progress; in particular, the multimodal embedding model is currently being implemented.Digital HeritageData Analysis, Datasets and Multimodal Approache