Search

Appropriate evaluation and experimental design are fundamental for empirical sciences, particularly in data-driven fields. Due to the successes in computational modeling of languages, for instance, research outcomes are having an increasingly immediate impact on end users. As the gap in adoption by end users decreases, the need increases to ensure that tools and models developed by the research communities and practitioners are reliable, trustworthy, and supportive of the users in their goals. In this position paper, we focus on the issues of evaluating visual text analytics approaches. We take an interdisciplinary perspective from the visualization and natural language processing communities, as we argue that the design and validation of visual text analytics include concerns beyond computational or visual/interactive methods on their own. We identify four key groups of challenges for evaluating visual text analytics approaches (data ambiguity, experimental design, user trust, and "big picture" concerns) and provide suggestions for research opportunities from an interdisciplinary perspective.

Recently, an article by Seneff et al. entitled "Innate immunosuppression by SARS-CoV-2 mRNA vaccinations: The role of G-quadruplexes, exosomes, and MicroRNAs" was published in Food and Chemical Toxicology (FCT). Here, we describe why this article, which contains unsubstantiated claims and misunderstandings such as "billions of lives are potentially at risk" with COVID-19 mRNA vaccines, is problematic and should be retracted. We report here our request to the editor of FCT to have our rebuttal published, unfortunately rejected after three rounds of reviewing. Fighting the spread of false information requires enormous effort while receiving little or no credit for this necessary work, which often even ends up being threatened. This need for more scientific integrity is at the heart of our advocacy, and we call for large support, especially from editors and publishers, to fight more effectively against deadly disinformation.

Thermal science is a perennial obstacle for learners. Infrared camera technology provides an opportunity for pupils to confront challenging thermal ideas. From an embodied cognition perspective, sensory experiences form metaphoric relations that underpin conceptualisation and reasoning about abstract scientific phenomena. This study investigated eight groups of fourth grade pupils’ use of metaphoric relations when engaging thermal cameras to explore “heat” at a science centre. Pupils were videorecorded while collaboratively exploring thermal properties of the surroundings and during a thermos modelling exercise. Qualitative metaphor analyses of pupils’ dialogue and behaviour revealed various metaphoric and metonymic relations around spatial properties, colour, movement and change in their conceptualisation of thermal phenomena. “Heat” was almost exclusively conceptualised as a noun, manifested in utterances such as “harder for the heat to escape” and “it wants to get yellow”. In addition, pupils used colour as both a metonym and metaphor for heat and temperature. Expressions of heat as an entity were closely related to experiences of movement, indicating that spatial cognition is central to children’s conceptualisation of heat. Engagement with the cameras provided access to thermodynamic phenomena through unique sensory and nonsensory experiences. Future research will explore how these metaphoric relations can be exploited as a meaning-making resource in the classroom.

Application-oriented papers provide an important way to invigorate and cross-pollinate the visualization field, but the exact criteria for judging an application paper's merit remain an open question. This article builds on a panel at the 2016 IEEE Visualization Conference entitled "Application Papers: What Are They, and How Should They Be Evaluated?" that sought to gain a better understanding of prevalent views in the visualization community. This article surveys current trends that favor application papers, reviews the benefits and contributions of this paper type, and discusses their assessment in the review process. It concludes with recommendations to ensure that the visualization community is more inclusive to application papers., Funding Agencies|Office of Science, Office of Advanced Scientific Computing Research, of the US DOE [DE-AC02-05CH11231]; Canadas Natural Science and Engineering Research Council; Alberta Innovates Technology Futures; SMART Technology; US Department of Homeland Security Center of Excellence program [2009-ST-061-CI0003]; Boeing; Oracle; Adobe; Knut and Alice Wallenberg Foundation; Swedish eScience Research Center; Swedish Research Council

In industrial applications where the physical parameters are highly interconnected, keeping the process flow steady is a major concern for the operators. This is caused by the sensitivity of system to the process dynamics. As a result, a slight adjustment to a control parameter can significantly affect the efficiency of the system and thus impact the financial gain. Paper pulp production is an example of such a process, where operators continuously investigate the potential of changes in the process and predict the consequences of an adjustment before making a decision. Process parameter adjustments prescribed by simulated control models cannot be fully trusted as the external disturbances and the process inherent variabilities cannot be fully incorporated into the simulations. Therefore, to assess the viability of a strategy, operators often compare the situation with the historical records and trends during which the processes in the plant ran steadily. While previous research has mostly focused on developing advanced control models to simulate complex pulp production process, this work aims to support operators analytical reasoning by provision of effective data visualization. The contributions of our design study include a domain problem characterization and a linked-view visual encoding design, which aims to enhance operator's mental models independent of particular users or scenarios. Finally, by reflecting on the advantages of our choice of task abstraction technique, inherited from the ecological interface design framework [5], we reason for the generalizability of our approach to similar industrial applications., Funding agencies: The Sweden’s Innovation Agency (Vinnova) under the project EXPLAIN (2021-04336) and the Knutand Alice Wallenberg Foundation (grant KAW 2019.0024).

Training a deep neural network is computationally expensive, but achieving the same network performance with less computation is possible if the training data is carefully chosen. However, selecting input samples during training is challenging as their true importance for the optimization is unknown. Furthermore, evaluation of the importance of individual samples must be computationally efficient and unbiased. In this paper, we present a new input data importance sampling strategy for reducing the training time of deep neural networks. We investigate different importance metrics that can be efficiently retrieved as they are available during training, i.e., the training loss and gradient norm. We found that choosing only samples with large loss or gradient norm, which are hard for the network to learn, is not optimal for the network performance. Instead, we introduce an importance sampling strategy that selects samples based on the cumulative distribution function of the loss and gradient norm, thereby making it more likely to choose hard samples while still including easy ones. The behavior of the proposed strategy is first analyzed on a synthetic dataset, and then evaluated in the application of classification of malignant cancer in digital pathology image patches. As pathology images contain many repetitive patterns, there could be significant gains in focusing on features that contribute stronger to the optimization. Finally, we show how the importance sampling process can be used to gain insights about the input data through visualization of samples that are found most or least useful for the training., The fulltext is published under Creative Common license Attribution 4.0https://creativecommons.org/licenses/by/4.0/No changes have been made to the publication.

In industrial applications where the physical parameters are highly interconnected, keeping the process flow steady is a major concern for the operators. This is caused by the sensitivity of system to the process dynamics. As a result, a slight adjustment to a control parameter can significantly affect the efficiency of the system and thus impact the financial gain. Paper pulp production is an example of such a process, where operators continuously investigate the potential of changes in the process and predict the consequences of an adjustment before making a decision. Process parameter adjustments prescribed by simulated control models cannot be fully trusted as the external disturbances and the process inherent variabilities cannot be fully incorporated into the simulations. Therefore, to assess the viability of a strategy, operators often compare the situation with the historical records and trends during which the processes in the plant ran steadily. While previous research has mostly focused on developing advanced control models to simulate complex pulp production process, this work aims to support operators analytical reasoning by provision of effective data visualization. The contributions of our design study include a domain problem characterization and a linked-view visual encoding design, which aims to enhance operator's mental models independent of particular users or scenarios. Finally, by reflecting on the advantages of our choice of task abstraction technique, inherited from the ecological interface design framework [5], we reason for the generalizability of our approach to similar industrial applications., Funding agencies: The Sweden’s Innovation Agency (Vinnova) under the project EXPLAIN (2021-04336) and the Knutand Alice Wallenberg Foundation (grant KAW 2019.0024).

Training a deep neural network is computationally expensive, but achieving the same network performance with less computation is possible if the training data is carefully chosen. However, selecting input samples during training is challenging as their true importance for the optimization is unknown. Furthermore, evaluation of the importance of individual samples must be computationally efficient and unbiased. In this paper, we present a new input data importance sampling strategy for reducing the training time of deep neural networks. We investigate different importance metrics that can be efficiently retrieved as they are available during training, i.e., the training loss and gradient norm. We found that choosing only samples with large loss or gradient norm, which are hard for the network to learn, is not optimal for the network performance. Instead, we introduce an importance sampling strategy that selects samples based on the cumulative distribution function of the loss and gradient norm, thereby making it more likely to choose hard samples while still including easy ones. The behavior of the proposed strategy is first analyzed on a synthetic dataset, and then evaluated in the application of classification of malignant cancer in digital pathology image patches. As pathology images contain many repetitive patterns, there could be significant gains in focusing on features that contribute stronger to the optimization. Finally, we show how the importance sampling process can be used to gain insights about the input data through visualization of samples that are found most or least useful for the training., The fulltext is published under Creative Common license Attribution 4.0https://creativecommons.org/licenses/by/4.0/No changes have been made to the publication.

With advances in artificial intelligence, machine learning, and cognitive modelling, unmanned aircraft are expected to actas human-like wingmen in the near future. For fluent and effective manned–unmanned teaming, synthetic wingmen must beable to account for and adapt to their partners’ intent with little or no communication. To enable such abilities, it becomescrucial to identify the requirements that makes intent explainable to synthetic wingmen, necessitating approaches to describeand analyse intent from a human-centric perspective. To address this issue, this paper reports on findings from using WorkDomain Analysis to design and analyse models of situated intent with six levels of cognitive control (frames, effects, values,generic, implementations, and physical). Through a literature review and seven subject matter expert interviews, a synthesizedmodel was designed to represent fighter pilots’ intent in a manned–unmanned teaming scenario. Using the synthesized modelas the context, a transfer of control and a link loss situation were further described and analysed. Experiences show that WorkDomain Analysis can provide a practical and applicable means to model situated intent, particularly since designed modelscan be re-utilised to model intent in similar situations. Furthermore, the model analyses show the importance of accountingfor fighter pilots’ adopted frames since small variations of the framing of the situations can propagate throughout the modelresulting in conflicting or inconsistent intent. The paper concludes that synthetic wingmen must be able to reason about all six levels of cognitive control, requiring a more holistic approach to make intent explainable., Funding: VINNOVA; Swedish Defence Material Administration [2017-04884,2017-04884,2017-04884,2017-04884]; NFFP (National Aviation Research Programme) - VINNOVA (Swedish Governmental Agency for Innovation Systems); Swedish Armed Forces, The Swedish Defence Material Administration and NFFP (National Aviation Research Pro- gramme), which is funded by VINNOVA (Swedish Governmental Agency for Innovation Systems, 2017-04884,2017-04884,2017- 04884,2017-04884), the Swedish Armed Forces, and the Swedish Defence Material Administration.

In industrial applications where the physical parameters are highly interconnected, keeping the process flow steady is a major concern for the operators. This is caused by the sensitivity of system to the process dynamics. As a result, a slight adjustment to a control parameter can significantly affect the efficiency of the system and thus impact the financial gain. Paper pulp production is an example of such a process, where operators continuously investigate the potential of changes in the process and predict the consequences of an adjustment before making a decision. Process parameter adjustments prescribed by simulated control models cannot be fully trusted as the external disturbances and the process inherent variabilities cannot be fully incorporated into the simulations. Therefore, to assess the viability of a strategy, operators often compare the situation with the historical records and trends during which the processes in the plant ran steadily. While previous research has mostly focused on developing advanced control models to simulate complex pulp production process, this work aims to support operators analytical reasoning by provision of effective data visualization. The contributions of our design study include a domain problem characterization and a linked-view visual encoding design, which aims to enhance operator's mental models independent of particular users or scenarios. Finally, by reflecting on the advantages of our choice of task abstraction technique, inherited from the ecological interface design framework [5], we reason for the generalizability of our approach to similar industrial applications., Funding agencies: The Sweden’s Innovation Agency (Vinnova) under the project EXPLAIN (2021-04336) and the Knutand Alice Wallenberg Foundation (grant KAW 2019.0024).

Training a deep neural network is computationally expensive, but achieving the same network performance with less computation is possible if the training data is carefully chosen. However, selecting input samples during training is challenging as their true importance for the optimization is unknown. Furthermore, evaluation of the importance of individual samples must be computationally efficient and unbiased. In this paper, we present a new input data importance sampling strategy for reducing the training time of deep neural networks. We investigate different importance metrics that can be efficiently retrieved as they are available during training, i.e., the training loss and gradient norm. We found that choosing only samples with large loss or gradient norm, which are hard for the network to learn, is not optimal for the network performance. Instead, we introduce an importance sampling strategy that selects samples based on the cumulative distribution function of the loss and gradient norm, thereby making it more likely to choose hard samples while still including easy ones. The behavior of the proposed strategy is first analyzed on a synthetic dataset, and then evaluated in the application of classification of malignant cancer in digital pathology image patches. As pathology images contain many repetitive patterns, there could be significant gains in focusing on features that contribute stronger to the optimization. Finally, we show how the importance sampling process can be used to gain insights about the input data through visualization of samples that are found most or least useful for the training., The fulltext is published under Creative Common license Attribution 4.0https://creativecommons.org/licenses/by/4.0/No changes have been made to the publication.

Forensic practitioners analyse intrinsic 3D data daily on 2D screens. We explore novel immersive visualisation techniques that enable digital autopsy through analysis of 3D imagery. We employ a user-centred design process involving four rounds of user feedback: (1) formative interviews eliciting opportunities and requirements for mixed-reality digital autopsies; (2) a larger workshop identifying our prototypes limitations and further use-cases and interaction ideas; (3+4) two rounds of qualitative user validation of successive prototypes of novel interaction techniques for pathologist sense-making. Overall, we find MR holds great potential to enable digital autopsy, initially to supplement physical autopsy, but ultimately to replace it. We found that experts were able to use our tool to perform basic virtual autopsy tasks, MR setup promotes exploration and sense making of cause of death, and subject to limitations of current MR technology, the proposed system is a valid option for digital autopsies, according to experts feedback. - Warning: This paper contains sensitive images which are 3D visualisation of deceased people., Funding Agencies|Victorian Institute of Forensic Medicine (VIFM); Knut and Alice Wallenberg Foundation [KAW 2019.0024]

The SESAR-funded Modern ATM via Human / Automation Learning Optimisation (MAHALO) project recently completed two years of technical work exploring the human performance impacts of AI and Machine Learning (ML), as applied to enroute ATC conflict detection and resolution (CD&R). It first developed a hybrid ML CD&R capability, along with a realtime simulation platform and experimental User Interface. After a series of development trials, the project culminated in a pair of field studies (i.e., human-in-the-loop trials) across two EU countries, with a total of 35 operational air traffic controllers. In each of these two field studies, controller behaviour was first captured in a pre-test phase, and used to train the ML system. Subsequent main experiment trials then experimentally manipulated within controllers both Conformance (as either a personalised-, group average-, or optimized model) and Transparency (as ether a baseline vector depiction, an enhanced graphical diagram, or a diagramplus-text presentation). The proposed paper presents guidelines on the design and implementation of ML systems in Air Traffic Control, derived from the results and lesson learned from the Simulations, as well as the qualitative feedback received from the controllers themselves.

Dimensionality reduction (DR) plays a vital role in the visual analysis of high-dimensional data. One main aim of DR is to reveal hidden patterns that lie on intrinsic low-dimensional manifolds. However, DR often overlooks important patterns when the manifolds are distorted or masked by certain influential data attributes. This paper presents a feature learning framework, FEALM, designed to generate a set of optimized data projections for nonlinear DR in order to capture important patterns in the hidden manifolds. These projections produce maximally different nearest-neighbor graphs so that resultant DR outcomes are significantly different. To achieve such a capability, we design an optimization algorithm as well as introduce a new graph dissimilarity measure, named neighbor-shape dissimilarity. Additionally, we develop interactive visualizations to assist comparison of obtained DR results and interpretation of each DR result. We demonstrate FEALMs effectiveness through experiments and case studies using synthetic and real-world datasets., Funding Agencies|Knut and Alice Wallenberg Foundation [KAW 2019.0024]; U.S. National Science Foundation [ITE-2134901]; National Institute of Health [1R01CA270454-01]

Exploration and rapid prototyping of design ideas for VR experiences is challenging for both novice and experienced designers as well as people that are not designers, especially in co-design projects. In this speculative paper, I propose Glowing Bodies, a bodystorming technique that leverages on the intrinsic immersive quality of a dark environment where participants use a simple prop, glow sticks, to create an embodied ideation and rapid prototyping space for VR.

In industrial applications where the physical parameters are highly interconnected, keeping the process flow steady is a major concern for the operators. This is caused by the sensitivity of system to the process dynamics. As a result, a slight adjustment to a control parameter can significantly affect the efficiency of the system and thus impact the financial gain. Paper pulp production is an example of such a process, where operators continuously investigate the potential of changes in the process and predict the consequences of an adjustment before making a decision. Process parameter adjustments prescribed by simulated control models cannot be fully trusted as the external disturbances and the process inherent variabilities cannot be fully incorporated into the simulations. Therefore, to assess the viability of a strategy, operators often compare the situation with the historical records and trends during which the processes in the plant ran steadily. While previous research has mostly focused on developing advanced control models to simulate complex pulp production process, this work aims to support operators analytical reasoning by provision of effective data visualization. The contributions of our design study include a domain problem characterization and a linked-view visual encoding design, which aims to enhance operator's mental models independent of particular users or scenarios. Finally, by reflecting on the advantages of our choice of task abstraction technique, inherited from the ecological interface design framework [5], we reason for the generalizability of our approach to similar industrial applications., Funding agencies: The Sweden’s Innovation Agency (Vinnova) under the project EXPLAIN (2021-04336) and the Knutand Alice Wallenberg Foundation (grant KAW 2019.0024).

Training a deep neural network is computationally expensive, but achieving the same network performance with less computation is possible if the training data is carefully chosen. However, selecting input samples during training is challenging as their true importance for the optimization is unknown. Furthermore, evaluation of the importance of individual samples must be computationally efficient and unbiased. In this paper, we present a new input data importance sampling strategy for reducing the training time of deep neural networks. We investigate different importance metrics that can be efficiently retrieved as they are available during training, i.e., the training loss and gradient norm. We found that choosing only samples with large loss or gradient norm, which are hard for the network to learn, is not optimal for the network performance. Instead, we introduce an importance sampling strategy that selects samples based on the cumulative distribution function of the loss and gradient norm, thereby making it more likely to choose hard samples while still including easy ones. The behavior of the proposed strategy is first analyzed on a synthetic dataset, and then evaluated in the application of classification of malignant cancer in digital pathology image patches. As pathology images contain many repetitive patterns, there could be significant gains in focusing on features that contribute stronger to the optimization. Finally, we show how the importance sampling process can be used to gain insights about the input data through visualization of samples that are found most or least useful for the training., The fulltext is published under Creative Common license Attribution 4.0https://creativecommons.org/licenses/by/4.0/No changes have been made to the publication.

A common research process in visualization is for visualization researchers to collaborate with domain experts to solve particular applied data problems. While there is existing guidance and expertise around how to structure collaborations to strengthen research contributions, there is comparatively little guidance on how to navigate the implications of, and power produced through the socio-technical entanglements of collaborations. In this paper, we qualitatively analyze reflective interviews of past participants of collaborations from multiple perspectives: visualization graduate students, visualization professors, and domain collaborators. We juxtapose the perspectives of these individuals, revealing tensions about the tools that are built and the relationships that are formed — a complex web of competing motivations. Through the lens of matters of care, we interpret this web, concluding with considerations that both trouble and necessitate reformation of current patterns around collaborative work in visualization design studies to promote more equitable, useful, and care-ful outcomes.

Embeddings are powerful tools for transforming complex and unstructured data into numeric formats suitable for computational analysis tasks. In this paper, we extend our previous work on using multiple embeddings for text similarity calculations to the field of networks. The embedding ensemble approach improves network reconstruction performance compared to single-embedding strategies. Our visual analytics methodology is successful in handling both text and network data, which demonstrates its generalizability beyond its originally presented scope.

Stacking (or stacked generalization) is an ensemble learning method with one main distinctiveness from the rest: even though several base models are trained on the original data set, their predictions are further used as input data for one or more metamodels arranged in at least one extra layer. Composing a stack of models can produce high-performance outcomes, but it usually involves a trial-and-error process. Therefore, our previously developed visual analytics system, StackGenVis, was mainly designed to assist users in choosing a set of top-performing and diverse models by measuring their predictive performance. However, it only employs a single logistic regression metamodel. In this paper, we investigate the impact of alternative metamodels on the performance of stacking ensembles using a novel visualization tool, called MetaStackVis. Our interactive tool helps users to visually explore different singular and pairs of metamodels according to their predictive probabilities and multiple validation metrics, as well as their ability to predict specific prob- lematic data instances. MetaStackVis was evaluated with a usage scenario based on a medical data set and via expert interviews.

Bibliographic data and bibliometric analyses play an important role in the professional life of academic researchers, and the quality of the respective publication records is essential for establishing the big picture of the relationships between particular publications, their authors and affiliations, or further data facets associated with publications. In this paper, we report on the design and outcomes of an interactive visual data exploration project conducted within the scope of a university with the goal of gaining overview of the university publication data. The project has been carried out by information visualization researchers in collaboration with several groups of stakeholders, including the university library and administration staff. We describe the design considerations, the resulting interactive visual interface, and the feedback received from the stakeholders with respect to the tool functionality and the insights discovered in the bibliographic data.

Clinical deployment of systems based on deep neural networks is hampered by sensitivity to domain shift, caused by e.g. new scanners or rare events, factors usually overcome by human supervision. We suggest a correct-then-predict approach, where the user labels a few samples of the new data for each slide, which is used to update the network. This few-shot meta-learning method is based on Model-Agnostic Meta-Learning (MAML), with the goal of training to adapt quickly to new tasks. Here we adapt and apply the method to the histopathological setting by identifying a task as a whole-slide image with its corresponding classification problem. We evaluated the method on three datasets, while purposefully leaving out-of-distribution data out from the training data, such as whole-slide images from other centers, scanners or with different tumor classes. Our results show that MAML outperforms conventionally trained baseline networks on all our datasets in average accuracy per slide. Furthermore, MAML is useful as a robustness mechanism to out-of-distribution data. The model becomes less sensitive to differences between whole-slide images and is viable for clinical implementation when used with the correct-then-predict workflow. This offers a reduced need for data annotation when training networks, and a reduced risk of performance loss when domain shift data occurs after deployment., Copyright 2021 Society of Photo-Optical Instrumentation Engineers. One print or electronic copy may be made for personal use only. Systematic reproduction and distribution, duplication of any material in this paper for a fee or for commercial purposes, or modification of the content of the paper are prohibited.

Single-image high dynamic range (SI-HDR) reconstruction has recently emerged as a problem well-suited for deep learning methods. Each successive technique demonstrates an improvement over existing methods by reporting higher image quality scores. This paper, however, highlights that such improvements in objective metrics do not necessarily translate to visually superior images. The first problem is the use of disparate evaluation conditions in terms of data and metric parameters, calling for a standardized protocol to make it possible to compare between papers. The second problem, which forms the main focus of this paper, is the inherent difficulty in evaluating SI-HDR reconstructions since certain aspects of the reconstruction problem dominate objective differences, thereby introducing a bias. Here, we reproduce a typical evaluation using existing as well as simulated SI-HDR methods to demonstrate how different aspects of the problem affect objective quality metrics. Surprisingly, we found that methods that do not even reconstruct HDR information can compete with state-of-the-art deep learning methods. We show how such results are not representative of the perceived quality and that SI-HDR reconstruction needs better evaluation protocols., Funding Agencies|Wallenberg Autonomous Systems and Software Program (WASP); strategic research environment ELLIIT; Knut and Alice Wallenberg Foundation (KAW)Knut & Alice Wallenberg Foundation; European Research Council (ERC) under the European Unions Horizon 2020 research and innovation programmeEuropean Research Council (ERC) [725253-EyeCode]

Communicating evolutionary theory continues to demonstrate significant challenges. One particularly demanding area is comprehending evolutionary time scales. Furthermore, visual representations such as the Tree of Life usually incorporate temporal aspects but require interpreting the underlying biological concepts as well as the representational features used to convey evolutionary meaning. In this paper, we investigate how users engage with the DeepTree exhibit, an interactive touch table visualization of a dynamic Tree of Life (Block et al., 2012). The main purpose of DeepTree is to visualize the relationship of all life on earth. Specific focus of this paper is to explore how a zooming interactive feature which enables users to move virtually from the origin of life to present day species, influences perception and understanding of evolutionary time associated with the tree metaphor. The aim of the study is to explore how the zooming feature affords new ways of understanding the information which the dynamic tree metaphor conveys, what properties of zooming interactions are associated with various temporal aspects and how moving in time is perceived. A video recorded clinical, semi-structured interview was conducted with each participant while s/he interacted with the DeepTree application. The analytical method used to treat the generated data was a combination of deductive and inductive coding of the transcripts and accompanying visual recordings. The findings suggest that DeepTree is a natural and intuitive touch-based interface for navigating the metaphor of biological relationships. Interacting with the visualization induced various positive affective reactions. However, in its present design, DeepTree may not completely support users' understanding of evolutionary time. The results indicate that zooming can be interpreted in different ways. This suggests that the embodied and immersive experience offered by such interactive tree applications might strongly i

Clinical deployment of systems based on deep neural networks is hampered by sensitivity to domain shift, caused by e.g. new scanners or rare events, factors usually overcome by human supervision. We suggest a correct-then-predict approach, where the user labels a few samples of the new data for each slide, which is used to update the network. This few-shot meta-learning method is based on Model-Agnostic Meta-Learning (MAML), with the goal of training to adapt quickly to new tasks. Here we adapt and apply the method to the histopathological setting by identifying a task as a whole-slide image with its corresponding classification problem. We evaluated the method on three datasets, while purposefully leaving out-of-distribution data out from the training data, such as whole-slide images from other centers, scanners or with different tumor classes. Our results show that MAML outperforms conventionally trained baseline networks on all our datasets in average accuracy per slide. Furthermore, MAML is useful as a robustness mechanism to out-of-distribution data. The model becomes less sensitive to differences between whole-slide images and is viable for clinical implementation when used with the correct-then-predict workflow. This offers a reduced need for data annotation when training networks, and a reduced risk of performance loss when domain shift data occurs after deployment., Copyright 2021 Society of Photo-Optical Instrumentation Engineers. One print or electronic copy may be made for personal use only. Systematic reproduction and distribution, duplication of any material in this paper for a fee or for commercial purposes, or modification of the content of the paper are prohibited.

Clinical deployment of systems based on deep neural networks is hampered by sensitivity to domain shift, caused by e.g. new scanners or rare events, factors usually overcome by human supervision. We suggest a correct-then-predict approach, where the user labels a few samples of the new data for each slide, which is used to update the network. This few-shot meta-learning method is based on Model-Agnostic Meta-Learning (MAML), with the goal of training to adapt quickly to new tasks. Here we adapt and apply the method to the histopathological setting by identifying a task as a whole-slide image with its corresponding classification problem. We evaluated the method on three datasets, while purposefully leaving out-of-distribution data out from the training data, such as whole-slide images from other centers, scanners or with different tumor classes. Our results show that MAML outperforms conventionally trained baseline networks on all our datasets in average accuracy per slide. Furthermore, MAML is useful as a robustness mechanism to out-of-distribution data. The model becomes less sensitive to differences between whole-slide images and is viable for clinical implementation when used with the correct-then-predict workflow. This offers a reduced need for data annotation when training networks, and a reduced risk of performance loss when domain shift data occurs after deployment., Copyright 2021 Society of Photo-Optical Instrumentation Engineers. One print or electronic copy may be made for personal use only. Systematic reproduction and distribution, duplication of any material in this paper for a fee or for commercial purposes, or modification of the content of the paper are prohibited.

Clinical deployment of systems based on deep neural networks is hampered by sensitivity to domain shift, caused by e.g. new scanners or rare events, factors usually overcome by human supervision. We suggest a correct-then-predict approach, where the user labels a few samples of the new data for each slide, which is used to update the network. This few-shot meta-learning method is based on Model-Agnostic Meta-Learning (MAML), with the goal of training to adapt quickly to new tasks. Here we adapt and apply the method to the histopathological setting by identifying a task as a whole-slide image with its corresponding classification problem. We evaluated the method on three datasets, while purposefully leaving out-of-distribution data out from the training data, such as whole-slide images from other centers, scanners or with different tumor classes. Our results show that MAML outperforms conventionally trained baseline networks on all our datasets in average accuracy per slide. Furthermore, MAML is useful as a robustness mechanism to out-of-distribution data. The model becomes less sensitive to differences between whole-slide images and is viable for clinical implementation when used with the correct-then-predict workflow. This offers a reduced need for data annotation when training networks, and a reduced risk of performance loss when domain shift data occurs after deployment., Copyright 2021 Society of Photo-Optical Instrumentation Engineers. One print or electronic copy may be made for personal use only. Systematic reproduction and distribution, duplication of any material in this paper for a fee or for commercial purposes, or modification of the content of the paper are prohibited.

Clinical deployment of systems based on deep neural networks is hampered by sensitivity to domain shift, caused by e.g. new scanners or rare events, factors usually overcome by human supervision. We suggest a correct-then-predict approach, where the user labels a few samples of the new data for each slide, which is used to update the network. This few-shot meta-learning method is based on Model-Agnostic Meta-Learning (MAML), with the goal of training to adapt quickly to new tasks. Here we adapt and apply the method to the histopathological setting by identifying a task as a whole-slide image with its corresponding classification problem. We evaluated the method on three datasets, while purposefully leaving out-of-distribution data out from the training data, such as whole-slide images from other centers, scanners or with different tumor classes. Our results show that MAML outperforms conventionally trained baseline networks on all our datasets in average accuracy per slide. Furthermore, MAML is useful as a robustness mechanism to out-of-distribution data. The model becomes less sensitive to differences between whole-slide images and is viable for clinical implementation when used with the correct-then-predict workflow. This offers a reduced need for data annotation when training networks, and a reduced risk of performance loss when domain shift data occurs after deployment., Copyright 2021 Society of Photo-Optical Instrumentation Engineers. One print or electronic copy may be made for personal use only. Systematic reproduction and distribution, duplication of any material in this paper for a fee or for commercial purposes, or modification of the content of the paper are prohibited.

Clinical deployment of systems based on deep neural networks is hampered by sensitivity to domain shift, caused by e.g. new scanners or rare events, factors usually overcome by human supervision. We suggest a correct-then-predict approach, where the user labels a few samples of the new data for each slide, which is used to update the network. This few-shot meta-learning method is based on Model-Agnostic Meta-Learning (MAML), with the goal of training to adapt quickly to new tasks. Here we adapt and apply the method to the histopathological setting by identifying a task as a whole-slide image with its corresponding classification problem. We evaluated the method on three datasets, while purposefully leaving out-of-distribution data out from the training data, such as whole-slide images from other centers, scanners or with different tumor classes. Our results show that MAML outperforms conventionally trained baseline networks on all our datasets in average accuracy per slide. Furthermore, MAML is useful as a robustness mechanism to out-of-distribution data. The model becomes less sensitive to differences between whole-slide images and is viable for clinical implementation when used with the correct-then-predict workflow. This offers a reduced need for data annotation when training networks, and a reduced risk of performance loss when domain shift data occurs after deployment., Copyright 2021 Society of Photo-Optical Instrumentation Engineers. One print or electronic copy may be made for personal use only. Systematic reproduction and distribution, duplication of any material in this paper for a fee or for commercial purposes, or modification of the content of the paper are prohibited.

The challenge of how cities can be designed and developed in an inclusive and sustainable direction is monumental. Smart city technologies currently offer the most promising solution for long-term sustainability. However, smart city projects have been criticised for ignoring diverse needs of the local population and increasing social divides. A sustainable urban environment depends as much on creating an inclusive space that is safe, accessible and comfortable for a diverse group of citizens as it does on deploying "smart" technologies for energy efficiency or environmental protection. This is because citizens will be more likely to adopt technologies promoting sustainability if they are well-aligned with their lived needs and experiences. In this paper, we present the rationale behind an ongoing interdisciplinary research project that aims to address exactly the problem outlined above by using a participatory design approach. Focusing on a smart city test site in Sweden where sensors are currently being deployed to collect data on noise, particles, vehicle numbers and types (amongst other), the goal is to bring local residents and government representatives into dialogue with technical developers by adopting a "meet-in-the-middle" approach. This paper comprises a brief presentation of early findings and a reflection on this approach., Funding Agencies | FORMAS Swedish Research Council Formas; Swedish Research Council for Sustainable Development [2019-01281]

The challenge of how cities can be designed and developed in an inclusive and sustainable direction is monumental. Smart city technologies currently offer the most promising solution for long-term sustainability. However, smart city projects have been criticised for ignoring diverse needs of the local population and increasing social divides. A sustainable urban environment depends as much on creating an inclusive space that is safe, accessible and comfortable for a diverse group of citizens as it does on deploying "smart" technologies for energy efficiency or environmental protection. This is because citizens will be more likely to adopt technologies promoting sustainability if they are well-aligned with their lived needs and experiences. In this paper, we present the rationale behind an ongoing interdisciplinary research project that aims to address exactly the problem outlined above by using a participatory design approach. Focusing on a smart city test site in Sweden where sensors are currently being deployed to collect data on noise, particles, vehicle numbers and types (amongst other), the goal is to bring local residents and government representatives into dialogue with technical developers by adopting a "meet-in-the-middle" approach. This paper comprises a brief presentation of early findings and a reflection on this approach., Funding Agencies | FORMAS Swedish Research Council Formas; Swedish Research Council for Sustainable Development [2019-01281]

The challenge of how cities can be designed and developed in an inclusive and sustainable direction is monumental. Smart city technologies currently offer the most promising solution for long-term sustainability. However, smart city projects have been criticised for ignoring diverse needs of the local population and increasing social divides. A sustainable urban environment depends as much on creating an inclusive space that is safe, accessible and comfortable for a diverse group of citizens as it does on deploying "smart" technologies for energy efficiency or environmental protection. This is because citizens will be more likely to adopt technologies promoting sustainability if they are well-aligned with their lived needs and experiences. In this paper, we present the rationale behind an ongoing interdisciplinary research project that aims to address exactly the problem outlined above by using a participatory design approach. Focusing on a smart city test site in Sweden where sensors are currently being deployed to collect data on noise, particles, vehicle numbers and types (amongst other), the goal is to bring local residents and government representatives into dialogue with technical developers by adopting a "meet-in-the-middle" approach. This paper comprises a brief presentation of early findings and a reflection on this approach., Funding Agencies | FORMAS Swedish Research Council Formas; Swedish Research Council for Sustainable Development [2019-01281]

The challenge of how cities can be designed and developed in an inclusive and sustainable direction is monumental. Smart city technologies currently offer the most promising solution for long-term sustainability. However, smart city projects have been criticised for ignoring diverse needs of the local population and increasing social divides. A sustainable urban environment depends as much on creating an inclusive space that is safe, accessible and comfortable for a diverse group of citizens as it does on deploying "smart" technologies for energy efficiency or environmental protection. This is because citizens will be more likely to adopt technologies promoting sustainability if they are well-aligned with their lived needs and experiences. In this paper, we present the rationale behind an ongoing interdisciplinary research project that aims to address exactly the problem outlined above by using a participatory design approach. Focusing on a smart city test site in Sweden where sensors are currently being deployed to collect data on noise, particles, vehicle numbers and types (amongst other), the goal is to bring local residents and government representatives into dialogue with technical developers by adopting a "meet-in-the-middle" approach. This paper comprises a brief presentation of early findings and a reflection on this approach., Funding Agencies | FORMAS Swedish Research Council Formas; Swedish Research Council for Sustainable Development [2019-01281]

Monitoring conditions in the periglacial areas of Antarctica helps geographers and geologists to understand physical processes associated with mesoscale land systems. Analyzing these unique temporal datasets poses a significant challenge for domain experts, due to the complex and often incomplete data, for which corresponding exploratory tools are not available. In this paper, we present a novel visual analysis tool for extraction and interactive exploration of temporal measurements captured at the polar station at the James Ross Island in Antarctica. The tool allows domain experts to quickly extract information about the snow level, originating from a series of photos acquired by trail cameras. Using linked views, the domain experts can interactively explore and combine this information with other spatial and non-spatial measures, such as temperature or wind speed, to reveal the interplay of periglacial and aeolian processes. An abstracted interactive map of the area indicates the position of measurement spots to facilitate navigation. The design of the tool was made in tight collaboration with geographers, which resulted in an early prototype, tested in the pilot study. The following version of the tool and its usability has been evaluated in the user study with five domain experts and their feedback was incorporated into the final version, presented in this paper. This version was again discussed with two experts in an informal interview. Within these evaluations, they confirmed the significant benefit of the tool for their research tasks.

The challenge of how cities can be designed and developed in an inclusive and sustainable direction is monumental. Smart city technologies currently offer the most promising solution for long-term sustainability. However, smart city projects have been criticised for ignoring diverse needs of the local population and increasing social divides. A sustainable urban environment depends as much on creating an inclusive space that is safe, accessible and comfortable for a diverse group of citizens as it does on deploying "smart" technologies for energy efficiency or environmental protection. This is because citizens will be more likely to adopt technologies promoting sustainability if they are well-aligned with their lived needs and experiences. In this paper, we present the rationale behind an ongoing interdisciplinary research project that aims to address exactly the problem outlined above by using a participatory design approach. Focusing on a smart city test site in Sweden where sensors are currently being deployed to collect data on noise, particles, vehicle numbers and types (amongst other), the goal is to bring local residents and government representatives into dialogue with technical developers by adopting a "meet-in-the-middle" approach. This paper comprises a brief presentation of early findings and a reflection on this approach., Funding Agencies | FORMAS Swedish Research Council Formas; Swedish Research Council for Sustainable Development [2019-01281]

This paper presents a framework that fully leverages the advantages of a deferred rendering approach for the interactive visualization of large-scale datasets. Geometry buffers (G-Buffers) are generated and stored in situ, and shading is performed post hoc in an interactive image-based rendering front end. This decoupled framework has two major advantages. First, the G-Buffers only need to be computed and stored once-which corresponds to the most expensive part of the rendering pipeline. Second, the stored G-Buffers can later be consumed in an image-based rendering front end that enables users to interactively adjust various visualization parameters-such as the applied color map or the strength of ambient occlusion-where suitable choices are often not known a priori. This paper demonstrates the use of Cinema Darkroom on several real-world datasets, highlighting CDs ability to effectively decouple the complexity and size of the dataset from its visualization., Funding Agencies|U.S. Department of Homeland SecurityUnited States Department of Homeland Security (DHS) [2017-ST-061-QA0001, 17STQAC00001-03-03]; National Science Foundation ProgramNational Science Foundation (NSF) [1350573]; German research foundation (DFG)German Research Foundation (DFG) [IRTG 2057]; Swedish Foundation for Strategic Research (SSF)Swedish Foundation for Strategic Research [BD15-0082]; SeRC (Swedish e-Science Research Center); ELLIIT environment for strategic research in Sweden

As the problem of reconstructing high dynamic range (HDR) imagesfrom a single exposure has attracted much research effort, it isessential to provide a robust protocol and clear guidelines on howto evaluate and compare new methods. In this work, we comparedsix recent single image HDR reconstruction (SI-HDR) methodsin a subjective image quality experiment on an HDR display. Wefound that only two methods produced results that are, on average,more preferred than the unprocessed single exposure images. Whenthe same methods are evaluated using image quality metrics, astypically done in papers, the metric predictions correlate poorlywith subjective quality scores. The main reason is a significant toneand color difference between the reference and reconstructed HDRimages. To improve the predictions of image quality metrics, we propose correcting for the inaccuracies of the estimated cameraresponse curve before computing quality values. We further analyzethe sources of prediction noise when evaluating SI-HDR methodsand demonstrate that existing metrics can reliably predict onlylarge quality differences., Funding: This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement N° 725253–EyeCode)

As the problem of reconstructing high dynamic range (HDR) imagesfrom a single exposure has attracted much research effort, it isessential to provide a robust protocol and clear guidelines on howto evaluate and compare new methods. In this work, we comparedsix recent single image HDR reconstruction (SI-HDR) methodsin a subjective image quality experiment on an HDR display. Wefound that only two methods produced results that are, on average,more preferred than the unprocessed single exposure images. Whenthe same methods are evaluated using image quality metrics, astypically done in papers, the metric predictions correlate poorlywith subjective quality scores. The main reason is a significant toneand color difference between the reference and reconstructed HDRimages. To improve the predictions of image quality metrics, we propose correcting for the inaccuracies of the estimated cameraresponse curve before computing quality values. We further analyzethe sources of prediction noise when evaluating SI-HDR methodsand demonstrate that existing metrics can reliably predict onlylarge quality differences., Funding: This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement N° 725253–EyeCode)

How can I tell what I think till I see what I make and do? — Christopher Frayling Just a couple of months before the start of the new decade and the pandemic began haunting the world, we had the privilege to attend the Halfway to the Future symposium (https://www.halfwaytothefuture.org/2019/) in person and present a paper. The symposium was organized on the occasion of the 20th anniversary of the Mixed Media Lab at the University of Nottingham. There were many exciting keynotes, presentations, and panels. At that time, we did not know that this would be one of the last physical presentations we would give for a while. In this blog post, we would like to talk about the stage performance we did to present our paper and how it motivated us to open a discussion on alternative ways of communicating scientific knowledge, in particular that which is produced through HCI and design research.

Haptic interaction with artificial intelligence (AI) entities that have a physical embodiment is a developing area of research which investigates how to design communicative tangible expressions for human-AI interactions. However, traditional sketching techniques such as storyboards or making fully functional prototypes do not seem to be convenient options to capture these agents’ potential for haptic interaction in context and rapidly explore possible ways of expression. In a recent study, we designed and built a Wizard of Oz (WOz) rig where users engaged in haptic interaction with a handheld non-humanoid robot as a form of embodied AI agent that was not working yet physically controlled with strong magnets by a human wizard hidden under the table. The paper reports on this WOz rig, shares reflections on the human wizard’s bodily engagement with users through his magnet-driven magic wand, and discusses its potential as a low-tech responsive experimentation environment for exploring movement-based AI expressions. Our contribution is the WOz mechanism and design-related insights derived from its operation, which could be useful for similar studies on haptic interaction with embodied AI agents.

As the problem of reconstructing high dynamic range (HDR) imagesfrom a single exposure has attracted much research effort, it isessential to provide a robust protocol and clear guidelines on howto evaluate and compare new methods. In this work, we comparedsix recent single image HDR reconstruction (SI-HDR) methodsin a subjective image quality experiment on an HDR display. Wefound that only two methods produced results that are, on average,more preferred than the unprocessed single exposure images. Whenthe same methods are evaluated using image quality metrics, astypically done in papers, the metric predictions correlate poorlywith subjective quality scores. The main reason is a significant toneand color difference between the reference and reconstructed HDRimages. To improve the predictions of image quality metrics, we propose correcting for the inaccuracies of the estimated cameraresponse curve before computing quality values. We further analyzethe sources of prediction noise when evaluating SI-HDR methodsand demonstrate that existing metrics can reliably predict onlylarge quality differences., Funding: This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement N° 725253–EyeCode)

Identifying sociolinguistic attributes of inter-community interactions is essential for understanding the polarization of social network communities. A wide range of computational text and network analysis methods may be applicable for this task, however, interpretation of the respective results and investigation of particularly interesting cases and subnetworks are difficult due to the scale and complexity of the data, e.g., for the Reddit platform. In this poster paper, we present an interactive visual analysis interface that facilitates network exploration and comparison at different topological and multivariate attribute scales. Users are able to investigate text- and network-based properties of social network community interactions, identify anomalies of conflict starters, or gain insight into multivariate anomalies behind groups of negative social media posts.

In this work, we explore the performance of supervised stance classification methods for social media texts in under-resourced languages and using limited amounts of labeled data. In particular, we focus specifically on the possibilities and limitations of the application of classic machine learning versus deep learning in social sciences. To achieve this goal, we use a training dataset of 5.7K messages posted on Flashback Forum, a Swedish discussion platform, further supplemented with the previously published ABSAbank-Imm annotated dataset, and evaluate the performance of various model parameters and configurations to achieve the best training results given the character of the data. Our experiments indicate that classic machine learning models achieve results that are on par or even outperform those of neural networks and, thus, could be given priority when considering machine learning approaches for similar knowledge domains, tasks, and data. At the same time, the modern pre-trained language models provide useful and convenient pipelines for obtaining vectorized data representations that can be combined with classic machine learning algorithms. We discuss the implications of their use in such scenarios and outline the directions for further research., This paper is an extended version of our paper published in the Proceedings of the Swedish Workshop on Data Science (SweDS ’21), Växjö, Sweden, December 2–3, 2021.Funding: EU CEF [2394203]; Linnaeus University

This paper started by asking the core question of why trust-building is hard in virtual meetings. Theories about "what is trust?" are drawn from the literature on interpersonal trust and trust in the business and marketing academic field. The sales meeting is the main observation object of this paper because it is a type of meeting that is more purpose-driven and certainly expected for economic relative output, compared with the other less-emergent daily meeting kinds. When many offline meetups are forced to be held online due to pandemic restrictions, the core of how to generate sales and manage customer relationships does not change much for a salesperson. It is commonly agreed by business meeting attendees that online trust-building is harder than it is in offline meetings. But what is missing from trust-building? This study uses user experience researching methods to identify the missed cues of essential trust elements both online and offline. Compare the similarities and differences of key cues of trust-building between attendees in their virtual sales meeting experience and face-to-face sales meeting experience. Finally, a few subtle aiding tools are designed for compensating for what is missed in the online meeting. This paper finds that it is not easy for attendees in virtual meeting being conscious and precise about which moment increases or decreases their trusting level towards counterparts. To trust a salesperson through online meeting requires a long and step by step process of building the trust, this process must expose a salesperson’s vulnerability and authenticity so can gradually trigger the happening in terms of trust. Design of applying visual cues can bring about the result of delivering eye gaze and listening status feedback in online meeting. The method is using qualitative study process includes organizing observations online and on-site for sales meetings. Analyze the user experience through studying on the recordings of the meetings, Examensarbetet är utfört vid Institutionen för teknik och naturvetenskap (ITN) vid Tekniska fakulteten, Linköpings universitet

To meet the increasing demands within the air traffic control (ATC) industry, automation is seen as one solution to increase efficiency, safety, and capacity. In ATC, teamwork between air traffic controllers (ATCO) is a key component to safely direct the aircraft. With increasing automation, the human-automation teamwork will become even more important to keep the human operator in the loop and contribute to the operator’s involvement in decision-making. However, operators can be cautious to new automation due to experience has taught them to be cautious and the willingness to accept automation can relate to the purpose of the automation. Therefore, this study presents how ATCOs want to work with an automated tool in the future, what qualities the automation should have, and what the human should be able to do when working with an automated system. Through an online questionnaire, 113 operational ATCOs answered questions about ATCOs’ experiences and expectations of current and future ATM systems, as well as teamwork between the ATCO and automated systems. For this paper, the teamwork aspects of the questionnaire are presented and discussed. The ATCOs believed that they should be able to trust the automation, know the reasoning behind the automation’s suggestions and decisions, and be able to always take control of the situation from the automation. The results will be used in early development processes of automation, in both the ATM domain but also in other control domains, that faces similar challenges.

Operational demands in safety-critical systems impose a risk of failure to the operators especially during urgent situations. Operators of safety-critical systems learn to make decisions effectively throughout extensive training programs and many years of experience. In the domain of air traffic control, expensive training with high dropout rates calls for research to enhance novices' ability to detect and resolve conflicts in the airspace. While previous researchers have mostly focused on redesigning training instructions and programs, the current paper explores possible benefits of novel visual representations to improve novices' understanding of the situations as well as their decision-making process. We conduct an experimental evaluation study testing two ecological visual analytics interfaces, developed in a previous study, as support systems to facilitate novice decision-making. The main contribution of this paper is threefold. First, we describe the application of an ecological interface design approach to the development of two visual analytics interfaces. Second, we perform a human-in-the-loop experiment with forty-five novices within a simplified air traffic control simulation environment. Third, by performing an expert-novice comparison we investigate the extent to which effects of the proposed interfaces can be attributed to the subjects' expertise. The results show that the proposed ecological visual analytics interfaces improved novices' understanding of the information about conflicts as well as their problem-solving performance. Further, the results show that the beneficial effects of the proposed interfaces were more attributable to the visual representations than the users' expertise., Funding: KAW Scholar Grant

Coffee cup reading is a popular social activity in some cultures and seen as a playful way of speculating on a person's past, present and future life. If we think of coffee cup reading as a way to envision futures, can we use this practice as a method for creating speculative design fictions? This paper proposes a workshop in which 10 people living in Venice will enact future inhabitants of the city and read each other's coffee cup in pairs to create speculative design fictions that envision Venice in 2030. The workshop will result in a catalogue of design fictions comprising text and drawings produced by each pair that will present different future visions for Venice. After the exhibition, this catalogue will be exhibited in a dedicated online space to present the design fictions and photos and/or videos created during the workshop as well as visitors' feedback. The workshop aims to provide insights about how coffee cup reading as a traditional way of looking into possible futures could be used as a low-threshold design activity to reveal urban inhabitants’ visions and concerns towards uncertain futures as well as today's challenges.

As Virtual Reality (VR) is becoming a readily available tool in an increasing number of fields, we see also an increasing number of non-technical users entering virtual environments (VEs). One pose manipulation technique that has the potential to be both fast and intuitive to use, is the Hand-centered Object Manipulation Extending Ray-casting (HOMER), presented first by Bowman and Hodges [1]. In this paper we take a look at how a naive implementation of HO-MER can lead to unintuitive interaction and present mathematics for avoiding this problem, by using a rotational offset of the grab vector instead of a static offset vector. Subsequently we compare the naive and rotation-based versions with respect to task completion time and number of re-grabs, and measure their respective System Usability Scale (SUS) score and Raw NASA Task Load Index (RTLX)

As the problem of reconstructing high dynamic range (HDR) imagesfrom a single exposure has attracted much research effort, it isessential to provide a robust protocol and clear guidelines on howto evaluate and compare new methods. In this work, we comparedsix recent single image HDR reconstruction (SI-HDR) methodsin a subjective image quality experiment on an HDR display. Wefound that only two methods produced results that are, on average,more preferred than the unprocessed single exposure images. Whenthe same methods are evaluated using image quality metrics, astypically done in papers, the metric predictions correlate poorlywith subjective quality scores. The main reason is a significant toneand color difference between the reference and reconstructed HDRimages. To improve the predictions of image quality metrics, we propose correcting for the inaccuracies of the estimated cameraresponse curve before computing quality values. We further analyzethe sources of prediction noise when evaluating SI-HDR methodsand demonstrate that existing metrics can reliably predict onlylarge quality differences., Funding: This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement N° 725253–EyeCode)