Diabetic retinopathy (DR) diagnosis delays pose a critical challenge for public healthcare systems such as the Brazilian Unified Health System (SUS), where long referral queues often prevent timely treatment and increase the risk of vision loss. Deep learning (DL) models offer an effective solution by automating retinal image analysis, but choosing an appropriate model requires balancing diagnostic accuracy with computational and energy efficiency. This study evaluates 38 convolutional neural networks (CNN) across four key dimensions: Area under the curve (AUC), energy consumption, model size, and training time. Our analysis identifies MobileNet as the superior architecture, demonstrating 77% lower energy use, 83% faster training, and 85% smaller model size than the InceptionV3 baseline, while achieving 3% higher AUC. We further optimize MobileNet through systematic hyperparameter tuning and evaluate its scalability on the ARM-based NVIDIA Grace Superchip, revealing peak efficiency at 36-thread configurations where energy use, CPU utilization, and memory access patterns reach optimal balance. All implementation scripts are publicly available to foster reproducible, sustainable AI development for clinical applications.

This paper reviews the literature on the visualization of individual mobility data, with a focus on thematic integration. It emphasizes the importance of visualization in understanding mobility patterns within a population and how it helps mobility experts to address domain-specific questions. We analyze 38 papers published between 2010 and 2024 in GIS and VIS venues that describe visualizations of multidimensional data related to individual movements in urban environments, concentrating on individual mobility rather than traffic data. Our primary aim is to report advances in interactive visualization for individual mobility analysis, particularly regarding the representation of thematic information about people’s motivations for mobility. Our findings indicate that the thematic dimension is only partially represented in the literature, despite its critical significance in transportation. This gap often stems from the challenge of identifying data sources that inherently provide this information, necessitating visualization designers and developers to navigate multiple, heterogeneous data sources. We identify the strengths and limitations of existing visualizations and suggest potential research directions for the field. • Analysis of 38 visual analytics solutions designed to aid the exploration of individual mobility data across spatial, temporal, and thematic dimensions. • An analysis of how visualization tools incorporate the theme dimension and which thematic properties are addressed. • Analysis of how the subject is approached within GIS and VIS communities.

Recent advancements in Natural Language Processing (NLP) have highlighted the success of transformer-based models in various tasks, including Named Entity Recognition (NER). This study evaluates and compares these models using GeoCorpus-3, Portuguese's most extensive annotated geological corpus. We investigate the performance of BERTimbau and XLM-RoBERTa models across 30 entity classes, incorporating both linear and CRF layers, and assess their effectiveness using micro, macro, and weighted F1-scores. Our results reveal that transformer models, particularly XLM-RoBERTa, achieve superior performance compared to previous methods. Statistical analysis indicates significant performance differences, with BERTimbau-Linear showing lower performance than the other models. This study provides a comprehensive evaluation of transformer models on a geological corpus and makes the updated corpus and models publicly available to advance research and development in this domain.

The Web of Linked Open Data (LOD) provides access to a great number of dynamic datasets containing valuable information to support decision-making processes in diverse application domains while being publicly accessible and up-to-date. While information visualization techniques are useful to explore, analyze, and explain relationships within LOD data, the existing tools are limited to visualizing a single dataset at a time and, often, use static and preprocessed data. In this article, we leverage the linked aspect of LOD to support the dynamic integration of data into a visualization system by connecting views and distributed LOD datasets using the so-called follow-up queries. We demonstrate how our approach uses dynamic SPARQL queries to integrate external data into the exploration flow through visualization techniques and enrich the ongoing analysis. We ran a semi-structured interview to assess the usefulness of our approach, which results were encouraging while showing its relevance to explore big linked data.

Trends are changes in variables or attributes over time, often represented by line plots or scatterplot variants, with time being one of the axes. Interpreting tendencies and estimating trends require observing the lines or points behavior regarding increments, decrements, or both (reversals) in the value of the observed variable. Previous work assessed variants of scatterplots like Animation, Small Multiples, and Overlaid Trails for comparing the effectiveness of trends representation using large and small displays and found differences between them. In this work, we study how best to enable the analyst to explore and perform temporal trend tasks with these same techniques in immersive virtual environments. We designed and conducted a user study based on the approaches followed by previous works regarding visualization and interaction techniques, as well as tasks for comparisons in three-dimensional settings. Results show that Overlaid Trails are the fastest overall, followed by Animation and Small Multiples, while accuracy is task-dependent. We also report results from interaction measures and questionnaires.

Visualization techniques are useful tools to explore data by enabling the discovery of meaningful patterns and causal relationships. The discovery process is often exploratory and requires multiple views to support analyzing different or complementary perspectives to the data. In this context, analytic provenance shows great potential to understand users’ reasoning process through the study of their interactions on multiple view systems. In this paper, we present an approach based on the concept of chained views to support the incremental exploration of large, multidimensional datasets. Our goal is to provide visual representation of provenance information to enable users to retrace their analytical actions and to discover alternative exploratory paths without loosing information on previous analyses. We demonstrate that our implementation of the approach, MGExplorer (Multidimensional Graph Explorer), allows users to explore different perspectives to a dataset by modifying the input graph topology, choosing visualization techniques, arranging the visualization space in meaningful ways to the ongoing analysis and retracing their analytical actions. MGExplorer combines multiple visualization techniques and visual querying while representing provenance information as segments connecting views, which each supports selection operations that help define subsets of the current dataset to be explored by a different view. We demonstrate the usage of the tool through a study case where we explore co-authorship data. We assess the approach through performance metrics, temporal ordering of tasks, number of physical actions, and amount of information to be recalled inbetween actions applied to the chosen visual exploration scenarios using chained views.

International audience

Whenever accessing indoor spaces such as classrooms or auditoriums, people might attempt to analyze and choose an appropriate place to stay while attending an event. Several criteria may be accounted for, and most are not always self-evident or trivial. This work proposes the use of data visualization allied to an Augmented Reality (AR) user interface to help users defining the most convenient seats to take. We consider sets of arbitrary demands and project information directly atop the seats and all around the room. Users can also narrow down the search by switching and combining the attributes being displayed, e.g., temperature, wheelchair accessibility. The proposed approach was tested against a comparable 2D interactive visualization of the same data in usability assessments of seat-choosing tasks with a set of users (N = 16) to validate the solution. Qualitative and quantitative data indicated that the AR-based solution is promising, suggesting that AR may help users make more accurate decisions, even in an ordinary daily task. Regarding Augmented Situated Visualization, our results open new avenues for the exploration of context-aware data.

Three-dimensional datasets from biological tissues have increased with the evolution of confocal microscopy. Hepatology researchers have used confocal microscopy for investigating the microanatomy of bile ducts. Bile ducts are complex tubular tissues consisting of many juxtaposed microstructures with distinct characteristics. Since confocal images are difficult to segment because of the noise introduced during the specimen preparation, traditional quantitative analyses used in medical datasets are difficult to perform on confocal microscopy data and require extensive user intervention. Thus, the visual exploration and analysis of bile ducts pose a challenge in hepatology research, requiring different methods. This paper investigates the application of unsupervised machine learning to extract relevant structures from confocal microscopy datasets representing bile ducts. Our approach consists of pre-processing, clustering, and 3D visualization. For clustering, we explore the density-based spatial clustering for applications with noise (DBSCAN) algorithm, using gradient information for guiding the clustering. We obtained a better visualization of the most prominent vessels and internal structures.

The use of novel displays and interaction resources to support immersive data visualization and improve analytical reasoning is a research trend in the information visualization community. In this work, we evaluate the use of an HMD-based environment for the exploration of multidimensional data, represented in 3D scatterplots as a result of dimensionality reduction (DR). We present a new modeling for this problem, accounting for the two factors whose interplay determine the impact on the overall task performance: the difference in errors introduced by performing dimensionality reduction to 2D or 3D, and the difference in human perception errors under different visualization conditions. This two-step framework offers a simple approach to estimate the benefits of using an immersive 3D setup for a particular dataset. Here, the DR errors for a series of roll call voting datasets when using two or three dimensions are evaluated through an empirical task-based approach. The perception error and overall task performance, on the other hand, are assessed through a comparative user study with 30 participants. Results indicated that perception errors were low and similar in all approaches, resulting in overall performance benefits in both desktop and HMD-based 3D techniques. The immersive condition, however, was found to require less effort to find information and less navigation, besides providing much larger subjective perception of accuracy and engagement.

Abstract 3D representations are potentially useful under many circumstances, but suffer from long known perception and interaction challenges. Current immersive technologies, which combine stereoscopic displays and natural interaction, are being progressively seen as an opportunity to tackle this issue, but new guidelines and studies are still needed, especially regarding information visualization. Many proposed approaches are impractical for actual usage, resulting in user discomfort or requiring too much time or space. In this work, we implement and evaluate an alternative data exploration metaphor where the user remains seated and viewpoint change is only realisable through physical movements. All manipulation is done directly by natural mid‐air gestures, with the data being rendered at arm's reach. The virtual reproduction of the analyst's desk aims to increase immersion and enable tangible interaction with controls and two dimensional associated information. A comparative user study was carried out against a desktop‐based equivalent, exploring a set of 9 perception and interaction tasks based on previous literature and a multidimensional projection use case. We demonstrate that our prototype setup, named VirtualDesk, presents excellent results regarding user comfort and immersion, and performs equally or better in all analytical tasks, while adding minimal or no time overhead and amplifying user subjective perceptions of efficiency and engagement. Results are also contrasted to a previous experiment employing artificial flying navigation, with significant observed improvements.

Network Functions Virtualization (NFV) is driving a paradigm shift in telecommunications networks, fostering new business models and creating innovation opportunities. In NFV-enabled networks, Service Providers (SPs) have the opportunity to build a business model where tenants can purchase Virtual Network Functions (VNFs) that provide distinct network services and functions. However, the chance to negotiate VNFs requires a change in traditional network planning strategies to accommodate tenants demands. In this context, the planning tasks perform a critical role in the introducing of business strategies that encompass both profit and health of services, which requires operators to have a broad understanding of the environment. In this paper, we propose the usage of two interactive visualization techniques to help NFV network operators in planning and strategic decisions. We advocate that our visualizations can aid in NFV planning tasks, such as infrastructure investment, resources allocation, and service pricing. We present three case studies to provide evidence of the feasibility and effectiveness of our visualizations.

Information Visualization (InfoVis) is a powerful tool to assist network administrators in daily tasks, and several authors report investigations about visualization techniques for network management. However, we have observed a fundamental issue that has not been addressed yet: how to promote the adoption of visualizations by network administrators, focusing on improving productivity and reducing costs? We claim that, in general, administrators do not have expertise in the InfoVis domain. Thus, the adoption of visualizations tends to be expensive and decrease the productivity of administrators, in special, because they lose focus on their core tasks. To overcome the raised issue, in this paper, we introduce a reuse-based approach (named as VisNSM) that aims to promote and encourage the adoption of visualizations for network management. To analyze and evaluate the feasibility of VisNSM, we have conducted a case study, and the obtained results show that our approach can significantly reduce costs and improve productivity.

Visualization means the use of Information Visualization (InfoVis) techniques to augment human cognitive capacity to support users tasks more efficiently. According to the current literature, visualization is a relevant requirement to support the management of Software-Defined Networking (SDN). For example, visualization can help network administrators in everyday tasks such as in the identification of traffic patterns, or in the implementation and test of new networking services. However, in most cases, network administrators are unskilled on InfoVis, which makes them naturally reluctant in the use of visualizations. Also, building up visualizations from scratch can spend a significant amount of effort, which directly impacts on the increase of cost and the decrease of administrators' productivity. Thus, in this paper, we introduce a reuse-based approach to facilitating the employment of visualizations in the SDN context with the focus on improving productivity and reduce costs. We developed a prototype to evaluate and demonstrate the feasibility of our approach. The obtained results to one of the usage scenarios show, for example, that our solution can decrease in 264 hours the administrators' workload (i.e., around U$ 9,976.56) when employing visualizations in daily tasks.

Debugging is a laborious activity in which developers spend lot of time navigating through code, looking for starting points, and stepping through statements. In this paper, we present the Swarm Debug Infrastructure (SDI) with which researchers can collect and share data about developers' interactive debugging activities. SDI allows collecting and sharing debugging data that are useful to answer research questions about interactive debugging activities. We assess the effectiveness of the SDI through an experiment to understand how developers apply interactive debugging

Debugging is a laborious activity in which developers spend lot of time navigating through code, looking for starting points, and stepping through statements. Yet, although debuggers exist for 40 years now, there have been few research studies to understand this important and laborious activity. Indeed, to perform such a study, researchers need detailed information about the different steps of the interactive debugging process. In this paper, to help research studies on debugging and, thus, help improving our understanding of how developers debug systems using debuggers, we present the Swarm Debug Infrastructure (SDI), with which practitioners and researchers can collect and share data about developers' interactive debugging activities. We assess the effectiveness of the SDI through an experiment that aims to understand how developers apply interactive debugging on five true faults found in JabRef, toggling breakpoints and stepping code. Our study involved five freelancers and two student developers performing 19 bug location sessions. We collect videos recording and data about 6 hours of effective debugging activities. The data includes 110 breakpoints and near 7,000 invocations. We process the collected videos and data to answer five research questions showing that (1) there is no correlation between the number of invocations (respectively the number of breakpoints toggled) during a debugging session and the time spent on the debugging task, ρ = -0.039 (respectively 0.093). We also observed that (2) developers follow different debugging patterns and (3) there is no relation between numbers of breakpoints and expertise. However, (4) there is a strong negative correlation between time of the first breakpoint (ρ = -0.637), and the time spent on the task, suggesting that when developers toggle breakpoints carefully, they complete tasks faster than developers who toggle breakpoints too quickly. We conclude that the SDI allows collecting and sharing debugging data that can provide interesting insights about interactive debugging activities. We discuss some implications for tool developers and future debuggers.

Debugging sessions require a methodical process of finding causes and reducing the number of software problems. During such sessions, developers run a software project, traversing method invocations, setting breakpoints, stopping or restarting executions. In these sessions, developers explore different parts of the code and create knowledge about them. When debugging sessions are over, it is likely that such knowledge is lost, and developers cannot use it in other sessions or sharing it with collaborators. We have developed Swarm Debugging, a new approach for visualizing and sharing information obtained during debugging sessions, providing interactive and real-time visualization techniques, and several searching tools. Through usage scenarios, we demonstrate that it can aid developers to decrease the required time for deciding where to toggle a break-point and locate bug causes. We show how Swarm Debugging offers more useful support for many typical development tasks than a traditional debugger tool.

Abstract Node‐link infographics are visually very rich and can communicate messages effectively, but can be very difficult to create, often involving a painstaking and artisanal process. In this paper we present an investigation of node‐link visualizations for communication and how to better support their creation. We begin by breaking down these images into their basic elements and analyzing how they are created. We then present a set of techniques aimed at improving the creation workflow by bringing more flexibility and power to users, letting them manipulate all aspects of a node‐link diagram (layout, visual attributes, etc.) while taking into account the context in which it will appear. These techniques were implemented in a proof‐of‐concept prototype called GraphCoiffure, which was designed as an intermediary step between graph drawing/editing software and image authoring applications. We describe how GraphCoiffure improves the workflow and illustrate its benefits through practical examples.

MagnetViz was designed for the interactive manipulation of force-directed graph layouts, allowing the user to obtain visualizations based on the graph topology and/or the attributes of its nodes and edges. The user can introduce virtual magnets anywhere in the graph and these can be set to attract nodes and edges that fulfill user-defined criteria. When a magnet is placed, the force-directed nature of the layout forces it to reorganize itself in order to reflect the changes in the balance of forces, consequently changing the visualization into one that is more semantically relevant to the user. This paper describes MagnetViz’s concepts, illustrating them with examples and a case study based on a usage scenario. We also describe how the MagnetViz has evolved since its original version and present the evaluation of its latest version. This evaluation consists of two user studies aiming at assessing generated layout quality and how well the concepts can be apprehended and employed, and a task taxonomy assessment focusing on establishing which graph visualization tasks the technique is able to handle.

Usually, ontologies are represented as static 2D graphs, being the relationships exhibited as overlapping edges, which may cause a cognitive overload. On the other hand, 3D representations can also lead to confusion in terms of navigation due to occlusion. Moreover, as the ontologies grow, incorporating new concepts (and their relationships), the visualization complexity increases either in 2D or in 3D. In this paper, we discuss a visualization tool for exploring relationships between ontology concepts. We employ linked tree structures that capture the hierarchical feature of parts of the ontology while preserving the different categories of relationships between concepts. Ontologies are displayed as trees on a plane, representing only the hierarchical relationships between concepts, allowing the user to explore other connections by creating projections of nodes (concepts) in another plane and linking them according to the relationships to be analyzed.

This paper presents an interactive physics-based technique for the exploration and dynamic reorganization of graph layouts that takes into account semantic properties which the user might need to emphasize. Many techniques have been proposed that take a graph as input and produce a visualization solely based on its topology, seldom ever relying on the semantic attributes of nodes and edges. These automatic topology-based algorithms might generate aesthetically interesting layouts, but they neglect information that might be important for the user. Among these are the force-directed or energy minimization algorithms, which use physics analogies to produce satisfactory layouts. They consist of applying forces on the nodes, which move until the physical system enters a state of mechanical equilibrium. We propose an extension of this metaphor to include tools for the interactive manipulation of such layouts. These tools are comprised of magnets, which attract nodes with user-specified criteria to the regions surrounding the magnets. Magnets can be nested and also used to intuitively perform set operations such as union and intersection, becoming thus an intuitive visual tool for sorting through the datasets. To evaluate the technique we discuss how they can be used to perform common graph visualization tasks.

Direct volume rendering techniques allow visualization of volume data without extracting intermediate geometry. The mapping from voxel attributes to optical properties is performed by transfer functions which, consequently, play a crucial role in building informative images from the data. One-dimensional transfer functions, which are based only on a scalar value per voxel, often do not provide proper visualizations. On the other hand, multidimensional transfer functions can perform more sophisticated data classification, based on vectorial voxel signatures. The transfer function design is a non-trivial and unintuitive task, especially in the multi-dimensional case. In this paper we propose a multi-dimensional transfer function design technique that uses self-organizing maps to perform dimensional reduction. Our approach gives uniform treatment to volume data containing voxel signatures of arbitrary dimension, and allows the use of any type of voxel attribute as part of the voxel signatures.

Multidimensional data sets can be visualized in a variety of forms provided by many techniques described in the literature. Depending on specific tasks, users might need to analyze the same data set using different representations. Moreover, they might need to interact with a view and have the results shown also in the others. This paper presents a system to provide multiple coordinated views of multidimensional data. It activates the coordination of techniques provided by InfoVis as well as visualizations we implemented using its basic resources. We allow users to set which visualizations they want to coordinate through a diagram representing the different visualizations. We present the user-driven coordination scheme and extensions we made in InfoVis to allow coordinated views through different interaction tools.

A stock market investor relies on two schools of analysis of market behavior to determine a trading strategy: Technical Analysis and Fundamental Analysis. Fundamental Analysis is based on the study of the fundamental data of a company and more directed to long-term investments, not taking into account small short-term price oscillations. On the other hand, Technical Analysis is mostly used for mid and short-term investments and is based on the study of past price behavior through the use of statistical tools and price history charts, taking into consideration the hypothesis that prices form patterns and reflect all the relevant information about a company and about the psychology of other investors. The application of Technical Analysis requires a computational system capable of displaying price history charts and providing tools to be used with them. This paper presents a prototype for a portable, extensible and easy-to-use tool for desktop/laptop and handheld computers that provides the investor with techniques for the visualization of stock market data. Classical visualization techniques and tools, such as Line, Bar, Candlestick and Point and Figure Charts, as well as extra tools, such as candlestick pattern recognition, are available as built-in functions, but new tools and visualizations can be easily added. The software was built with the .NET and .NET Compact frameworks and utilizes XML to store the data set.

The broad-phase step of collision detection in scenes composed of n moving objects is a challenging problem because enumerating collision pairs has an inherent O(n2) complexity. Spatial data structures are designed to accelerate this process, but often their static nature makes it difficult to handle dynamic scenes. In this work we propose a new structure called Semi-Adjusting BSP-tree for representing scenes composed of thousands of moving objects. An scheduling algorithm evaluates locations where the BSP-tree becomes unbalanced, uses several strategies to alter cutting planes, and defer updates based on their re-structuring cost. We show that the tree does not require a complete re-structuring even in highly dynamic scenes, but adjusts itself while maintaining desirable balancing and height properties.

Evaluating user interfaces is usually accomplished to detect design problems in layout and interaction. One possible way to evaluate image quality in computer graphics is visual inspection by experts. Information visualization techniques are usually presented showing their use in experimental situations, employing some kind of analysis. Nevertheless, few works have specifically addressed the evaluation of such techniques. This work reports our results towards the definition of criteria for evaluating information visualization techniques, addressing the evaluation of visual representation and interaction mechanisms as a first step.

This paper describes a new joint model for articulated human skeletons. The model is based in human anatomy and was conceived for use in medical applications. In this model, we try to overcome some limitations of other models used in computer graphics. For validation, we compared a knee modeled by our approach with a simulator, a plastic model of a knee, and images taken during a video-arthroscopy.

Cooperative manipulation refers to the simultaneous manipulation of a virtual object by multiple users in an immersive virtual environment. This paper describes a framework supporting the development of collaborative manipulation techniques, and example techniques we have tested within this framework. We describe the modeling of cooperative interaction techniques, methods of combining simultaneous user actions, and the awareness tools used to provide the necessary knowledge of partner activities during the cooperative interaction process. Our framework is based on a Collaborative Metaphor concept that defines rules to combine user interaction techniques. The combination is based on the separation of degrees of freedom between two users. Finally, we present novel combinations of two interaction techniques (Simple Virtual Hand and Ray-casting).