IEEE Computer Graphics and Applications最新文献

The need for large, high-quality annotated datasets continues to represent a primary limitation in training Object Detection (OD) models. To mitigate this challenge, we present VILOD, a Visual Interactive Labeling tool that integrates Active Learning (AL) with a suite of interactive visualizations to create an effective Human-in-the-Loop (HITL) workflow for OD annotation and training. VILOD is designed to make the AL process more transparent and steerable, empowering expert users to implement diverse, strategically guided labeling strategies that extend beyond algorithmic query strategies. Through comparative case studies, we evaluate three visually guided labeling strategies against a conventional automated AL baseline. The results show that a balanced, human-guided strategy-leveraging VILOD's visual cues to synthesize information about data structure and model uncertainty-not only outperforms the automated baseline but also achieves the highest overall model performance. These findings emphasize the potential of visually guided, interactive annotation to enhance both the efficiency and effectiveness of dataset creation for OD.

This paper presents a case study focusing on the exploratory visual analysis of a unique historical dataset consisting of approximately 4000 visual sketches and associated captions from an encyclopedic book published in 1909-1910. The book, which offers insight into Vietnamese crafts and social practices, poses the challenge of extracting cultural meaning and narrative structure from thousands of drawings and multilingual captions. Our research aims to explore and evaluate the effectiveness of multiple visualization techniques in uncovering meaningful relationships within the dataset while working closely with professional historians. The main contributions of this study include refining historical research questions through task and data abstraction, combining and validating visualization techniques for historical data interpretation, and involving a focus group of historians for further evaluation. These contributions offer generalizable insights for the development of domain-specific visualization tools and support interdisciplinary engagement in historical data visualization and critical digital humanities research.

We present PLUTO (Public VaLUe Assessment TOol), a framework for assessing the public value of specific instances of data use. Grounded in the concept of data solidarity, PLUTO aims to empower diverse stakeholders-including regulatory bodies, private enterprises, NGOs, and individuals-to critically engage with data projects through a structured assessment of the risks and benefits of data use, and by encouraging critical reflection. This paper discusses the theoretical foundation, development process, and initial user experiences with PLUTO. Key challenges include translating qualitative assessments of benefits and risks into actionable quantitative metrics while maintaining inclusivity and transparency. Initial feedback highlights PLUTO's potential to foster responsible decision-making and shared accountability in data practices.

This article proposes a method to design protective foam for packaging 3-D objects. Users first load a 3-D object and define a block-based design space by setting the block resolution and the size of each block. The system then constructs a block map in the space using depth textures of the input object, separates the map into two regions, and outputs the regions as foams. The proposed method is fast and stable, allowing the user to interactively make protective foams. The generated foam is a height field in each direction, so the foams can easily be fabricated using various materials, such as LEGO blocks, sponge with slits, glass, and wood. This article shows some examples of fabrication results to demonstrate the robustness of our system. In addition, we conducted a user study and confirmed that our system is effective for manually designing protective foams envisioned by users.

The study investigates the role of human-AI interaction in animating illustration with a case study of John Gilbert's visual interpretation of Shakespeare's play As You Like It. Through a multilayered animation, the research highlighted the irreplaceable role of human direction, particularly as a creator, to achieve narrative and visual coherence in AI-assisted animation. Drawing on theories of creativity and collaborative spaces, this article argues that human guidance is inherent for successful AI-empowered animation. It proposes a structured HAI workflow where the human remains the creative agent and main lead, while AI augments the process. This case study showcased how cocreative workflow can ensure visual and narrative coherence rather than foster mutual extinction.

Personal affective physicalization is the process by which individuals express emotions through tangible forms to record, reflect on, and communicate. Yet such physical data representations can be challenging to design due to the abstract nature of emotions. Given the shown potential of AI in detecting emotion and assisting design, we explore opportunities in AI-assisted design of personal affective physicalization using a research-through-design method. We developed PhEmotion, a tool for embedding LLM-extracted emotion values from human-AI conversations into the parametric design of physical artifacts. A lab study was conducted with 14 participants creating these artifacts based on their personal emotions, with and without AI support. We observed nuances and variations in participants' creative strategies, meaning-making processes, and their perceptions of AI support in this context. We found key tensions in AI-human cocreation that provide a nuanced agenda for future research in AI-assisted personal affective physicalization.

Virtual surgical simulation offers promising training for complex procedures, such as robotic internal mammary artery harvesting. Building upon previous work on dynamic virtual simulation with haptic feedback, we present an adaptive human-AI interaction framework that dynamically adjusts cardiac pulsation parameters based on surgeon behavior analysis. Our system captures surgical tool movements and performance metrics to create personalized training through dynamic difficulty adjustment, context-aware parameter selection, personalized learning paths, and real-time feedback. In a study with three cardiac surgeons across 24 sessions, our adaptive approach showed significant improvements over static simulations: 18% reduction in spatial asymmetry, 22% faster completion, and 48% fewer tissue trauma events. The system demonstrated consistent benefits across different skill levels and sustained learning progression, preventing performance plateaus seen in fixed-difficulty conditions.

This article describes the design and implementation of a course that evaluated the applicability of an artistic studio model, to educating students on the subject of artificial intelligence (AI). Specifically, four sections of an asynchronous, studio-style understanding and exploring AI course ran once per season via the Rhode Island School of Design online learning facility during the 2024-2025 academic year. The artistic studio model engages in bottom-up learning methods that include student-directed exploration, engagement, and artifact creation. As generative AI tools can output artistic artifacts based on human prompting, the research aligned well with typical course objectives. The qualitative study describes students' experiences of integrating Large Learning Models in support of their creative process. The results from 36 students are considered as evidence that the dominant model is applicable, and case studies from individual students are provided to assist the reader in considering the model for their own needs and interests.

Virtual reality (VR) technologies have emerged as valuable tools for medical and emergency training, providing safe, immersive, and repeatable environments where complex procedures can be practiced effectively. This article presents an immersive VR system designed to train workplace first-aid responders, with a particular focus on cardiopulmonary resuscitation (CPR). The platform integrates a physical CPR manikin with virtual patient overlays through mixed-reality calibration, incorporates realistic emergency scenarios with environmental hazards, and enables synchronous multiuser interaction between trainees and instructors. To assess its potential, we provide a detailed description of the system's architecture and functionalities, introducing the results of an extensive user study employing validated questionnaires on usability, performance, and user experience. The proposed framework contributes to the advancement of VR-based medical education, highlighting its benefits, current limitations, and future research opportunities.

This Visualization Viewpoints article explores how visualization helps uncover and communicate the internal chain-of-thought trajectories and generative pathways of large language models (LLMs) in reasoning tasks. As LLMs become increasingly powerful and widespread, a key challenge is understanding how their reasoning dynamics unfold, particularly in natural language processing (NLP) applications. Their outputs may appear coherent, yet the multistep inference pathways behind them remain largely hidden. We argue that visualization offers an effective avenue to illuminate these internal mechanisms. Moving beyond attention weights or token saliency, we advocate for richer visual tools that expose model uncertainty, highlight alternative reasoning paths, and reveal what the model omits or overlooks. We discuss examples, such as prompt trajectory visualizations, counterfactual response maps, and semantic drift flows, to illustrate how these techniques foster trust, identify failure modes, and support deeper human interaction with these systems. In doing so, visualizing the chain of thought in LLMs lays critical groundwork for transparent, interpretable, and truly collaborative human-AI reasoning.