Search

Proceedings of Graphics Interface 2019, Kingston, Ontario, Canada, 28 - 31 May 2019, Optimizing the use of a small display while presenting graphic data such as line charts is challenging. To tackle this, we propose GSparks, a compact visual representation of glanceable line graphs for smartwatches. Our exploration primarily considered the suitable compression axes for time-series charts. In a first study we examine the optimal line-graph compression approach without compromising perceptual metrics, such as slope or height detections. We evaluated compressions of line segments, the elementary unit of a line graph, along the x-axis, y-axis, and xyaxes. Contrary to intuition, we find that condensing graphs yield more accurate reading of height estimations than non-compressed graphs, but only when these are compressed along the x-axis. Building from this result, we study the effect of an x-axis compression on users' ability to perform "glanceable" analytic tasks with actual data. Glanceable tasks include quick perceptual judgements of graph properties. Using bio-metric data (heart rate), we find that shrinking a line graph to the point of representing one data sample per pixel does not compromise legibility. As expected, such type of compression also has the effect of minimizing the needed amount of flicking to interact with such graphs. From our results, we offer guidelines to application designers needing to integrate line charts into smartwatch apps.

Proceedings of Graphics Interface 2019, Kingston, Ontario, Canada, 28 - 31 May 2019, Free-motion quilting patterns are functional and decorative patterns sewn on pieced quilts using a single-line continuous stitch path for each region of the quilt. Seven families of quilting patterns are commonly used by quilters [3]. We present an approach for computationally generating three of these families. The user can control the design for each family based on a set of parameters, including the density and general layout of the pattern as well as the decorative elements. Our algorithm starts by sampling a point set in a designated region, generates a skeleton path over that set, then inserts decorative elements along the skeleton. We provide digital and quilted examples for each type of pattern.

Proceedings of Graphics Interface 2019, Kingston, Ontario, Canada, 28 - 31 May 2019, With millions of recipe videos increasingly available online, viewers often face the challenge of browsing through these videos and deciding among different styles of recipe demonstrations and instructions. Although state-of-the-art video summarization techniques using linear presentation formats have been shown to be effective in domains such as surveillance, sports or lecture videos, recipe videos are often more complex and may require a different summarization approach. We first investigated how viewers navigate recipe videos and what information they look for when seeking quick overviews of such videos. Based on our findings, we designed VideoWhiz, a novel interactive video summarization tool that provides a non-linear overview design allowing easy access to the key stages or milestones within the recipe and inter-milestone relationships. VideoWhiz uses a combination of computer vision techniques and an annotation workflow to generate these interactive overviews. Our evaluation showed that viewers found VideoWhiz to be effective and useful in providing quick overviews of recipe videos. We discuss the potential for future work to investigate non-linear overviews for other types of instructional videos and to explore more powerful representations for video summarization.

Proceedings of Graphics Interface 2019, Kingston, Ontario, Canada, 28 - 31 May 2019, Practitioners use low-pass filters to improve the quality of noisy input device signals whose optimal parameters depend on applicationspecific precision and latency requirements, as well as situational human and environmental factors. Two common calibration approaches are to learn optimal filter parameters from training data, or interactively tune via trial and error until satisfaction ensues, both having major drawbacks. We propose a novel, automatic custom calibration technique for pointing tasks called Pitch Pipe that in three straightforward steps is able to determine appropriate parameters for a given filter, and is therefore suitable for deployment into unknown environments. Specifically, we estimate noise and user speed, and then select those parameters that best meet system requirements. In a widely deployed Fitts’ task user study, we show that Pitch Pipe-tuned filters perform on par with their manually calibrated counterparts, demonstrating that one may use our automatic approach for custom calibration.

Proceedings of Graphics Interface 2019, Kingston, Ontario, Canada, 28 - 31 May 2019, In computer games, designers frequently leverage biologicallyinspired movement algorithms such as flocking, particle swarm optimization, and firefly algorithms to give players the perception of intelligent behaviour of groups of enemy non-player characters (NPCs). While extensive effort has been expended designing these algorithms, a comparison between biologically inspired algorithms and naive directional algorithms (travel towards the opponent) has yet to be completed. In this paper, we compare the biological algorithms listed above against a naive control algorithm to assess the effect that these algorithms have on various measures of player experience. The results reveal that the Swarming algorithm, followed closely by Flocking, provide the best gaming experience. However, players noted that the firefly algorithm was most salient. An understanding of the strengths of different behavioural algorithms for NPCs will contribute to the design of algorithms that depict more intelligent crowd behaviour in gaming and computer simulations.

Proceedings of Graphics Interface 2019, Kingston, Ontario, Canada, 28 - 31 May 2019, Generative design techniques use algorithmic encodings of domain knowledge to automate parts of the design process. This approach has worked well when success can be written as an optimization problem, but useful evaluation criteria are often discovered during the design process. To study how these criteria are developed we built DreamRooms, a room layout prototyping tool with a tight interaction loop between the designer and a generative process that does not encode a priori objective measures of quality. DreamRooms consists of a VR environment wherein the user can set constraints and gradually lower the entropy of a generative process that produces alternative layouts for the user to consider and iterate on. In addition to the DreamRooms system, we present the results of an observational study which revealed benefits to rapid collaboration between a designer and the generative process in an embodied environment and points towards mechanisms for communicating design intent to the generative process.

Proceedings of Graphics Interface 2019, Kingston, Ontario, Canada, 28 - 31 May 2019, To address the problem of improving text entry accuracy in mobile devices, we present a new tablet keyboard that offers both immediate and delayed feedback on language quality through autocorrection, prediction, and grammar checking. We combine different visual representations for grammar and spelling errors, accepted predictions, and auto-corrections. and also support interactive swiping/tapping features and improved interaction with previous errors, predictions, and auto-corrections. Additionally, we added smart error correction features to the system to decrease the overhead of correcting errors and to decrease the number of operations. We designed our new input method with an iterative user-centered approach through multiple pilots. We conducted a lab-based study with a refined experimental methodology and found that WiseType outperforms a standard keyboard in terms of text entry speed and error rate. The study shows that color-coded text background highlighting and underlining of potential mistakes in combination with fast correction methods can improve both writing speed and accuracy.

Proceedings of Graphics Interface 2018, Toronto, Ontario, Canada, 8-11 May 2018, 18 - 25, Users can benefit from using an auxiliary peripheral that could mitigate many concerns with direct text entry on wearable devices. We introduce ThumbText, a thumb-operated text entry approach for a ring-sized touch surface. Through a multi-part exploration, we first identify a suitable discretization of the miniature touch surface for thumb input. We then design a number of two-step selection techniques for supporting the input of at least 28 characters. On a miniature touch surface, we find that a continuous touch-slide-lift selection technique in a 2×3 grid discretization offers improved performance gains over other selection methods. Finally, we evaluate ThumbText against techniques also designed for wearable devices and find that ThumbText allows for higher text entry rates than SwipeBoard and H4-Writer.

Proceedings of Graphics Interface 2018, Toronto, Ontario, Canada, 8-11 May 2018, 146 - 153, Aesthetics, specifically visual appeal, is an important aspect of user experience. It is included as a principle in frameworks such as Fogg's Functional Triad and the Persuasive Systems Design. Yet, literature that directly investigates the influence of aesthetics on persuasion is limited, especially in the context of mobile applications. To understand how aesthetics influences persuasion if it includes the concept of operant conditioning, we designed a mobile app called Couch, which aims to reduce sedentary behaviour. We devised a 2x2 between-subject experiment, creating four versions of the app with two levels of aesthetics and two levels of persuasion (with and without). Measuring persuasion through self-reports, we found that higher levels of persuasion had a significant impact in reducing sedentary behaviour over aesthetics. However, visual appeal had no significant effect on persuasion. We comment on the level of visual appeal of the app and discuss the implications for future work.

Proceedings of Graphics Interface 2018, Toronto, Ontario, Canada, 8-11 May 2018, 154 - 161, Non-player characters (NPCs) serve important functions for game narratives and influence player immersion. However, the visual design of NPCs for specific narrative roles is often approached by relying on designers' previous experience or guesswork. We contribute to the understanding of player perception of narrative NPC roles in games, by proposing a methodological approach towards the visual design of NPCs to fit specific narrative roles. We demonstrate this approach through the visual design of characters for the three narrative roles of mentor, companion, and enemy. The results of an online survey (n=45) indicate trait expectations towards these narrative roles, and differences therein based on participant gender. Further, the characters were generally perceived as the targeted role based on visual design alone. This method of designing characters for narrative roles is beneficial to both game designers and researchers for further exploring effects of NPCs on player experience.

Proceedings of Graphics Interface 2018, Toronto, Ontario, Canada, 8-11 May 2018, 168 - 175, Cybersickness in virtual reality (VR) is an on-going problem, despite recent advances in technology. In this paper, we propose a method for reducing the likelihood of cybersickness onset when using stationary (e.g., seated) VR setups. Our approach relies on reducing optic flow via inconsistent displacement - the viewpoint is “snapped” during fast movement that would otherwise induce cybersickness. We compared our technique, which we call viewpoint snapping, to a control condition without viewpoint snapping, in a custom-developed VR first-person shooter game. We measured participant cybersickness levels via the Simulator Sickness Questionnaire (SSQ), and user reported levels of nausea, presence, and objective error rate. Overall, our results indicate that viewpoint snapping significantly reduced SSQ reported cybersickness levels by about 40% and resulted in a reduction in participant nausea levels, especially with longer VR exposure. Presence levels and error rate were not significantly different between the viewpoint snapping and the control condition.

Proceedings of Graphics Interface 2018, Toronto, Ontario, Canada, 8-11 May 2018, 162 - 167, Teleportation is a popular locomotion technique that lets users navigate beyond the confines of limited available positional tracking space. Because it discontinuously translates the viewpoint, it is considered a safe locomotion method because it doesn't generate any optical flow, and thus reduces the risk of vection induced VR sickness. Though the lack of optical flow minimizes VR sickness, it also limits path integration, e.g., estimating the total distance traveled, and which can lead to spatial disorientation. This paper evaluates a teleportation technique called Dash that quickly but continuously displaces the user's viewpoint and which retains some optical flow cues. A user study with 16 participants compares Dash to regular teleportation and found that it significantly improves path integration while there was no difference in VR sickness.

Proceedings of Graphics Interface 2018, Toronto, Ontario, Canada, 8-11 May 2018, 93 - 99, Movement artificial intelligence of simulated humanoid characters has been advancing rapidly through joint efforts of the computer animation, robotics, and machine learning communitites. However, practical real-life applications are still rare. We propose applying the technology to mental practice in sports, which we denote as computer-aided imagery (CAI). Imagery, i.e., rehearsing the task in one's mind, is a difficult cognitive skill that requires accurate mental simulation; we present a novel interactive computational sport simulation for exploring and planning movements and strategies. We utilize a fully physically-based avatar with motion optimization that is not limited by a movement dataset, and customize the avatar with computer vision measurements of user's body. We evaluate the approach with 20 users in preparing for real-life wall climbing. Our results indicate that the approach is promising and can affect body awareness and feelings of competence. However, more research is needed to achieve accurate enough simulation for both gross-motor body movements and fine-motor control of the myriad ways in which climbers can grasp climbing holds or shapes.

Proceedings of Graphics Interface 2018, Toronto, Ontario, Canada, 8-11 May 2018, 76 - 83, Audio podcasts have been widely used for more than a decade where millions of people listen to digital content on mobile devices. Despite a large amount of research on podcasts, there have not been any studies that explore the detailed listening practices of frequent podcast users, in particular, with a focus on understanding how podcasts support alone time. We conducted an interview study to understand and learn from such practices. Our results point to the characteristics of podcast technology that made it suitable for supporting people's ability to be alone yet still feel like they were connected to others. This included being able to multitask while listening to a podcast, escape from times of boredom, and even have experiential moments of self-reflection. These behaviors were supported by the flexibility of podcasts as a storytelling medium, a feeling of intimacy and connection with the podcast host, and podcasts' ability to make people feel like they are part of a conversation even when alone. We explore how these features suggest direction for technologies that can support alone time.

Proceedings of Graphics Interface 2018, Toronto, Ontario, Canada, 8-11 May 2018, 131 - 137, Interactive 3D models have the potential to enhance accessibility and education, but can be complex and time-consuming to produce. We present Blowhole, a technique for embedding blowing-activated tags into 3D-printed models to add interactivity. Requiring no special printing techniques, components, or assembly and working on consumer-level 3D printers, Blowhole adds acoustically resonant cavities to the interior of a model with unobtrusive openings at the surface of the object. A gentle blow into a hole produces a unique sound that identifies the hole, allowing a computer to provide associated content. We describe the theory behind Blowhole, characterize the performance of different cavity parameters, and describe our implementation, including easy-to-use software to automatically embed blowholes into preexisting models. We illustrate Blowhole's potential with multiple working examples.

Proceedings of Graphics Interface 2018, Toronto, Ontario, Canada, 8-11 May 2018, 34 - 41, Word Clouds have gained an impressive momentum for summarizing text documents in the last years. They visually communicate in a clear and descriptive way the most frequent words of a text. However, there are only very few word cloud visualizations that support a contrastive analysis of multiple documents. The available approaches provide comparable overviews of the documents, but have shortcomings regarding the layout, readability, and use of white space. To tackle these challenges, we propose MultiCloud, an approach to visualize multiple documents within a single word cloud in a comprehensible and visually appealing way. MultiCloud comprises several parameters and visual representations that enable users to alter the word cloud visualization in different aspects. Users can set parameters to optimize the usage of available space to get a visual representation that provides an easy visual association of words with the different documents. We evaluated MultiCloud with visualization researchers and a group of domain experts comprising five humanities scholars.

Proceedings of Graphics Interface 2018, Toronto, Ontario, Canada, 8-11 May 2018, 123 - 130, We present an evaluation of a new selection technique for virtual reality (VR) systems presented on head-mounted displays. The technique, dubbed EZCursorVR, presents a 2D cursor that moves in a head-fixed plane, simulating 2D desktop-like cursor control for VR. The cursor can be controlled by any 2DOF input device, but also works with 3/6DOF devices using appropriate mappings. We conducted an experiment based on ISO 9241-9, comparing the effectiveness of EZCursorVR using a mouse, a joystick in both velocity-control and position-control mappings, a 2D-constrained ray-based technique, a standard 3D ray, and finally selection via head motion. Results indicate that the mouse offered the highest performance in terms of throughput, movement time, and error rate, while the position-control joystick was worst. The 2D-constrained ray-casting technique proved an effective alternative to the mouse when performing selections using EZCursorVR, offering better performance than standard ray-based selection.