Open standards for representing and exchanging interactive 3D scenes and data is a critical capability needed to publish visualization and graphics to the web. X3D, the successor to VRML, is a leading provider of such capabilities, enabling integration with web services, supporting cross-platform execution, and facilitating inter-application data exchange. Building on the capabilities of X3D and other open data standards, the addition of cross-platform, shareable execution modules implemented via generalized programming languages provides additional powerful web deployment capabilities. To this end, we have developed execution models based on the emerging open standards WebAssembly (WASM) and WebGPU. WASM transforms complex code, in our case C++, into portable, secure, and performant modules supported by web browsers; while WebGPU enables cross-platform graphics and processing capabilities. In this paper we describe how we use WASM and WebGPU to build cross-platform web applications using the C++ VTK toolkit as the underlying graphics and visualization engine.
Recent advances in 3D Gaussian Splatting (3DGS) have enabled high-resolution real-time rendering; however, the unstructured nature of 3DGS complicates dynamic deformation. Without inherent topology, 3DGS often suffers from structural tearing and visual artifacts during manipulation. We present APG-GS, a framework enabling semantically selective deformation while preventing non-physical interference between adjacent objects. To establish reliable topology, we construct a Gaussian Similarity Graph using a multi-attribute metric that integrates spatial distance, orientation, scale, and Spherical Harmonics. A Point-consistent ChainMail algorithm then propagates user-driven displacements through the graph to ensure interactive responsiveness. Furthermore, we maintain geometric consistency by estimating local deformation gradients, refined via polar decomposition, to adapt Gaussian covariance matrices in real time. Experimental results demonstrate that APG-GS effectively reduces visual artifacts and preserves structural coherence without sacrificing rendering efficiency. This approach extends 3DGS from static reconstruction to interactive, user-driven dynamic applications.
We introduce Disentangle-and-Diffuse, a framework for structure-aware 3D shape generation that combines automatic part decomposition, invariant/equivariant feature encoding, and diffusion-based part synthesis. The pipeline first segments a 3D shape into semantic parts, then encodes each part with geometry features that are invariant to global rotations and pose features that transform equivariantly with the object orientation. A dual-stream transformer fuses local part context and global shape context to condition a part diffusion model, and the generated parts are assembled with predicted pose features. Experiments on challenging benchmarks show that our method improves structural consistency and geometric quality over the evaluated baselines, while enabling flexible shape manipulation and interpolation. These results support the value of combining structural and feature-level disentanglement for controllable 3D generative modeling.
Interactive adjustment of agent-based cooperative tasks is essential for supporting flexible and complex task execution. However, existing approaches generally lack intuitive visual and effective interaction, which limits real-time task adjustment and human intervention in complex scenarios. To address these challenges, we propose an interactive adjustment framework for agent-based cooperative tasks driven by multi-view collaboration. First, we propose a task representation visualization method based on a Dual-Structured Graph to provide a unified representation of task structures and agent execution processes. Second, we propose a task interaction enhancement method driven by a large language model to accurately interpret and execute diverse user interaction commands. Building on these methods, we implement an integrated platform with three views to support intuitive and flexible task adjustment through multi-view collaboration. The effectiveness of the proposed framework and platform is validated through three usage scenarios and a user study.
We propose Rooms, an open-source XR content creation platform designed to democratize 3D modeling and animation by leveraging the intuitiveness of VR motion controls and the high-performance capabilities of the modern WebGPU API. Built on top of wgpuEngine, Rooms provides real-time, cross-platform immersive sculpting and animation tools directly accessible through XR-enabled web browsers and desktop systems. The system implements a Signed Distance Field (SDF) sculpting pipeline that enables smooth Constructive Solid Geometry (CSG) operations for clay-like modeling. This paper details the architecture, the XR-first user interaction workflow, and the technical implementation, which includes a GPU-driven, sparse SDF baking pipeline. User studies confirm that the immersive, motion-based interaction model successfully lowers the entry barrier for non-expert users. Rooms effectively paves the way for future web-based XR creation environments by guiding the development of accessible and intuitive tools that harness the full power of modern GPU APIs on the web.
Web3D technologies enable global access to cultural heritage, yet online digital museums struggle to present vast macroscopic contexts alongside high-fidelity artifacts without disorienting users. We propose a lightweight Web-based visualization framework that bridges 2D spatiotemporal maps and 3D inspections through a dual-camera finite state machine and a "render-to-texture" visual anchor. This anchor projects the macro-context into the 3D space, preserving spatial cognition during seamless transitions. A prototype, "The Digital Silk Road," is implemented using HTML5 and WebGL. A within-subjects user study ($N=24$) shows the framework reduces navigation errors by 55% and task time by 32% (Cohen's $d \gt 1.3$) versus traditional click-and-load designs. Technical benchmarks confirm 78-88 FPS on consumer hardware with negligible overhead. The system aligns with IRVE theory, maps to X3D standards, and offers a scalable solution for presenting heterogeneous heritage data on the Web.
Broadening the reach of uncertainty visualization requires overcoming the technical barriers of delivering complex, multi-view interfaces across diverse devices. We present WebUVisBox, a framework that addresses these challenges through a modular, scenario-based architecture. By employing a registry pattern and declarative configuration, the framework decouples domain-specific visualization logic from core infrastructure, facilitating rapid extensibility. Simultaneously, its reactive state management and responsive layout engine ensure accessibility, automatically adapting complex analytical tools for both desktop workspaces and mobile field environments. We demonstrate the framework's capabilities through two case studies-wildfire ensemble forecasts and neural network flow uncertainty. The results validate that WebUVisBox effectively balances developer flexibility with the rigorous demands of reproducible, cross-platform scientific analysis
Avatar-mediated interaction increasingly spans virtual live entertainment and real-time telepresence, making interoperability and end-to-end latency central to user experience. This Perspective and Overview article surveys interoperable and latency-aware avatar pipelines with an emphasis on deployment contexts across the Asia-Pacific (AP)region. After introducing core animation primitives (blendshape-based facial deformation and skeletal animation with linear blend skinning), interoperability is formalized as a three-layer stack: asset containers, semantic control interfaces (highlighting VRM semantics and reusable motion via VRM Animation, .vrma), and streaming/transport rules. System bottlenecks are then outlined, including latency, robustness, etc., together with practical mitigation strategies based on prediction, scalable parameter streams, and compression of time-varying animation signals. Evaluation is organized around pipeline failure modes, temporal stability, signal-domain distortion, and multi-criteria subjective assessment. Finally, emerging directions in Asia-Pacific research and industry are outlined, including 3D Gaussian splatting (3DGS) avatars for mobile and augmented-reality telepresence, as well as text-to-motion models for intent-driven motion synthesis.
This article investigates the dissemination of geometric figures for education and academic research in architecture. When extracting 3D geometry and metadata from digital models created using computer-aided design (CAD) software, traditional interoperability workflows based on file exchange often introduce lossy conversions to standardized data schemas. This paper proposes an alternative approach based on a direct data stream between CAD software and a web-based visualization environment, using the open-source collaborative platform Speckle. After outlining the objectives and methods, a state-of-the-art review highlights key technical challenges. A case study based on geometric figures modeled in TopSolid is then presented, involving two development layers: (1) a Speckle connector implemented as a plug-in within TopSolid, and (2) a customized web viewer built using speckle-viewer and Three.js. The results show that a CAD environment can be effectively replicated on the web, enabling flexible and open dissemination of academic content in architectural geometry.
The advent of single-cell RNA sequencing (scRNA-seq) technologies has increased the demand for accessible analysis tools. Although a standard multi-stage scRNA-seq workflow is well established, existing visualization systems typically support only isolated stages or data types, leaving full-pipeline exploration under-addressed. Supporting the complete workflow introduces a broad problem and design space, requiring diverse analytical tasks and interactions to be systematically organized into a cohesive analytical process. To address this challenge, we conducted a design study with visualization and bioinformatics experts. From this process, we derived a ‘Task-to-Interface’ strategy for translating workflow-oriented tasks into coordinated interface structures. scFlowVis realizes this strategy as a unified visual analytics environment for exploration, comparison, and workflow tracking across the scRNA-seq pipeline. By detailing our design process and insights, we aim to inform the design of future visual analytics systems involving complex analytical workflows and integrated interfaces. The source code is available at https:// github.com/ EavanXing0416/ scFlowVis.
Reassembling fragmented 3D objects is challenging, particularly for archaeological artifacts affected by surface degradation and material heterogeneity. We propose a graph-based method integrating geometric and topological features to match fragments by analyzing surface patch arrangements instead of individual shapes, improving robustness to erosion and noise. Fragments are first segmented using Reeb graph-based partitioning. From this, an adjacency graph encoding the spatial organization of surface regions is constructed. Local geometry is described using the shape index to detect complementary areas across fragments. Pairwise matches are obtained by comparing structural patterns within adjacency graphs and filtered using geometric consistency constraints. Global reassembly is achieved through confidence-guided clustering with collision checks. Experiments on two public datasets achieve 80% and 95% pairwise matching recall and enable correct reconstruction of objects with up to 62 fragments, outperforming classical and recent AI-based methods.
Design students increasingly work in data-rich contexts but may struggle to interpret information, translate insights into form, and connect data to lived experiences. This article presents a pedagogical model that uses data physicalization both as an experiential learning medium and as an instrument for making learning visible. The model combines a three-project studio scaffold with an instructor-crafted meta-physicalization: a participatory artifact that transforms public audience engagement into interpretable indicators of reflection depth, emotional resonance, and perceived impact. Used across three public showcases in an undergraduate studio, the instrument revealed progression in audience responses, including sustained engagement, deeper reflections, and broader emotional expression. These patterns provide students with means to evaluate how their artifacts provoke thought, invite interpretation, and resonate with diverse audiences. The article contributes a replicable pedagogical model, an assessment tool grounded in experiential learning, and a strategy for using audience feedback as evidence to support reflection-in-action in data physicalization education
Extended Reality (XR) and Large Language Models (LLMs) can support context-aware, hands-free information access in industrial environments. This paper presents LlymX, a prototype combining a Unity-based 3D environment with an LLM-driven backend leveraging retrieval-augmented generation (RAG) and vector databases. LlymX introduces an interaction-oriented XR-LLM pipeline in which spatial user actions, such as gestures and object selection, are grounded in semantic object representations and linked to retrieval from a structured knowledge base. Through speech, gestures, and text, users can identify objects and access contextually relevant information grounded in domain documentation. A user study comparing LlymX against menu-based and infographic-based XR interfaces shows that the LLM-based modality is significantly more usable (SUS) than static infographics and comparable to traditional menus. Task completion time and cognitive workload showed no statistically significant differences, though participants appreciated the flexibility of conversational interaction.
Accurate digital twin reconstruction of articulated objects is essential for graphics-driven simulation, visualization, and embodied applications. However, existing 3D Gaussian Splatting (3DGS) approaches primarily target static scenes and fail to capture the structural dynamics of multi-joint objects, leading to geometric inconsistencies and limited realism. We present a structure-aware reconstruction framework that attaches articulated Gaussian primitives to the kinematic hierarchy, enabling dynamic and editable digital twin reconstruction from multi-view observations via differentiable rendering. By explicitly modeling link-level geometry and motion coupling, the proposed method produces temporally consistent reconstructions that generalize to novel viewpoints and unseen joint configurations. Extensive experiments demonstrate improved visual fidelity, geometric accuracy, and cross-view consistency on articulated robotic systems. Furthermore, the reconstructed digital twins facilitate scalable simulation and enhance sim-to-real policy transfer, highlighting the importance of dynamic Gaussian-based modeling for high-fidelity digital twin applications.
A surface-filling curve is a non-self-intersecting single curve uniformly covering a 3D object’s surface. These curves are actively studied for their broad applicability. State-of-the-art generation algorithms rely on geometric flows. However, curve initialization for geometric flows has received little attention despite its significant impact on performance. Moreover, straightforward alternative initializations suffer from low robustness, slow speed, and inflexibility. To address this, we propose two complementary initialization algorithms: the ultra-fast Steiner Traversal Initialization (STI) and the highly robust Poisson Traversal Initialization (PTI). STI traverses around an approximate minimum Steiner tree on the dual graph, while PTI uses geodesic Poisson disk sampling to construct and traverse a tree structure. Extensive experiments show that while STI's performance degrades on meshes with sliver triangles, it accelerates curve generation by up to 37.14× on high-quality meshes. Meanwhile, PTI achieves up to a 32.63× speedup, operating robustly even on poorly tessellated meshes containing severe sliver triangles.
This Guest Editorial introduces the special issue Intelligent CAD in the Age of AI, which examines how artificial intelligence (AI) is reshaping the foundations of computer-aided design (CAD). As CAD evolves beyond geometry-centric modeling toward intelligent, context-aware design partners, AI enables new capabilities for understanding design intent, generating controllable alternatives, reconstructing parametric histories, and supporting richer modes of human–AI interaction. At the same time, CAD provides a uniquely structured and semantically grounded testbed that exposes open challenges for AI—including data efficiency, constraint handling, and multimodal geometric reasoning. The four articles featured in this issue reflect the field’s expanding scope: advances in generative modeling for engineering CAD, new insights into dataset adequacy and augmentation, methods for recovering design logic from boundary-representation geometry, and hybrid workflows that bridge scanned data and editable representations. Together, these contributions illustrate a rapidly advancing research frontier and point toward a future in which intelligent CAD systems integrate automation, semantic understanding, and human expertise more seamlessly than ever.
Machine learning (ML) is widely used in medicinal chemistry, but accurate predictions alone are insufficient. Researchers need insight into which molecular features determine compound properties. We present an application-oriented case study that analyzes a trained model for compound potency as a source of domain knowledge. The model is converted into decision rules, and topic-guided visual analytics is used to identify co-occurring feature conditions associated with high predicted potency. These patterns are then mapped back to molecular substructures, yielding chemically interpretable motifs and testable hypotheses about structure-activity relationships. The study demonstrates how combining rule-based representations, topic modeling, and visual exploration can turn potency predictions into mechanistic insight, and outlines a reusable workflow for interpreting ML models of molecular properties.
As a federally funded initiative, the Insight Toolkit (ITK) is intended to make a lasting software investment for biomedical image computing, enabling our communities with open-source algorithms for image segmentation and registration. Its 25-year success has been the result of a combination of carefully integrated factors: executive support, project management expertise, user involvement and skilled developers, the adoption of agile processes and tools, and maintaining clear business objectives in our execution. This history of ITK covers these factors, from the inception of the toolkit through its agile implementation and continuous testing and integration. The story includes motivations and fortunate enabling opportunities and ends with summary advice on how to succeed in creating a successful project of open-source software.
We are researchers from six different countries with diverse research backgrounds. We gathered at Schloss Dagstuhl to reflect on data visualizations for public health communication during the COVID-19 pandemic. There, we shared what we had learned and discussed our experiences visualizing COVID-19 data to gain useful insights into the role of data visualization in future pandemics. In this article, we highlight missed opportunities and considerations and provide recommendations. We learned that visualization practices should include cross-disciplinary collaborators with socio-technical expertise, consider personalization and audience more extensively, and use coordinated communication pathways to ensure effective, timely communication in future pandemics.
