close

Jürgen Bernard, Computer Science at UBC, Vancouver

Enhancing Interactive Machine Learning

December 10th, 2019, 4:00 pm CET

Room: Science Park Building 3, Room 063

Abstract:
Machine Learning (ML) is a key technology in the era of data-centered research and applications.
Given the availability of data, a series of data-driven learning tasks have successfully been formalized in the past and made ML-capable. As a result, many automated ML-based solutions already ease our everyday life or gain other types of promising benefits.

Interactive Machine Learning is a principle that may be applicable for the plethora of data-driven learning tasks for which formalizations and automations could not be achieved so far. Interactive ML explicitly includes the involvement of humans in the iterative and incremental ML process, following the idea to combine the strengths of both humans and machines to create more effective and efficient solutions. There is considerable evidence that many remaining data-driven learning challenges can be tackled with interactive ML principles. Example include data-oriented challenges like heterogeneous data, dirty data, uncertain data, or unlabeled data, as well as process-oriented challenges such as data preprocessing, model building, model parameterization, model quality assessment, model uncertainty assessment, model refinement, model explanation and interpretation, or model comparison. These performance-driven challenges are accompanied by at least two human-centered goals worth to be pursued: 1) creating personalized ML solutions tailored to the information need of individual users or application tasks, as well as 2) making ML applicable for larger user groups far beyond the group of data science experts.

This talk motivates the human-centered and interactive perspective of ML. Using two of my research branches as examples, I will show how users can be enabled to play an active role in the ML process and be an integral part of highly interactive and iterative training and validation cycles. The first example is the Segmentation and Labeling of Multivariate Time Series (VISSECT), a visual analytics approach that investigates the interplay between ML models, model parameters, and uncertainty in a systematic way. The second example includes Visual-Interactive Labeling (VIAL), an approach to solve the data labeling problem. The data labeling problem refers to the lack of labels in the substantial majority of datasets existing today. Finding solutions to label data effectively will make today’s unlabeled data feasible for supervised ML such as classification and regression.

About the Speaker:
Jürgen Bernard is a postdoctoral research fellow at the University of British Columba, Vancouver. He has joined the InfoVis group, led by Professor Tamara Munzner in 2019. Jürgen Bernard studied Computer Sciences with focus on Computer Graphics and Bio Technology at the University of Technology of Darmstadt. He received his PhD Degree in 2015, when he was with Fraunhofer IGD. His thesis “Exploratory Search in Time-Oriented Primary Data” was awarded with the Hugo-Geiger Prize for excellent PhD theses, granted by the Fraunhofer Foundation. In 2016, Jürgen Bernard started as a Post-doc researcher at TU Darmstadt at Interactive Graphics Systems Group (Computer Science), leading his own research group (Visual-Interactive Machine Learning). In 2017, he received the Dirk Bartz Prize for Visual Computing in Medicine, granted by the Eurographics Association (EG).

Johannes Behr, Fraunhofer IGD, Darmstadt

Challenges for industrial VR/AR Application

November 26th, 2019, 3:00 pm CET

Room: Science Park Building 3, Room 063

Abstract:
Within the past years, Virtual Reality (VR) and Augmented Reality (AR) have gained much attention. In addition, the rising interest in these topics is not limited to entertainment applications any more: Affordable, high-performance hardware and powerful 3D visualization software are currently making VR/AR technology available to everyone.As a consequence, industrial VR/AR scenarios are becoming more and more popular. However, several crucial challenges remained largely unsolved so far, limiting the broad use of VR/AR in industry. Within my talk, I will highlight the key factors that determine the applicability of VR/AR in industry. I will mention several related challenges, and I will show how far we currently are on the way of solving them. In this context, 3D Web technology and standards are of crucial importance. Finally, I will give examples of promising industrial use cases, where we have succeeded to create industry-scale pipelines and workflows for VR/AR experiences on all kinds of client devices.

About the Speaker:
Dr. Johannes Behr received his M. Sc. diploma in Advanced Software Engineering form the University of Wolverhampton in 1996. From 1997-2004 he was working as a research assistant within the Visual Computing department of the Zentrum für Graphische Datenverarbeitung (ZGDV). Areas of interests where Virtual Reality, Computer Vision, Stereo Vision and 3D Interaction Techniques. Since 2005 Johannes Behr is working at the Fraunhofer Institut für Graphische Datenverarbeitung (IGD). In 2005 he has received his doctoral degree from the Darmstadt University of Technology. The title of his thesis is “Avalon – A scalable Framework for dynamic Mixed Reality Applications“. From 2008 to 2010 was he leading the VR-group of the "Virtual and Augmented Reality" department. He is now the head of the "Visual Computing System Technologies" (igd.fraunhofer.de/vcst) department which coordinates the current and future efforts to develop the InstantReality (www.instantreality.org), X3DOM (www.x3dom.org) and instant3Dhub (https://instant3dhub.org) solutions

Tom Torsney-Weir, Swansea University

Visualizing multi-dimensional spaces

October 30th, 2019, 3:30 p.m. CET

Room: Science Park Building 3, Room 063

Abstract:
I will discuss work on visualization techniques for understanding machine learning models, numerical simulations, multi-objective optimization, and function spaces. This diverse set of application areas can be abstracted under the name "multi-dimensional spaces." These are spaces typically between 4 and 20 dimensions where the dimensions have specific semantic meaning. In this talk I will discuss the available challenges and recent solutions to visually understanding these multi-dimensional spaces on a 2D screen.

About the Speaker:
Tom is currently a Lecturer at Swansea University. His research interests are in visualizing multi-dimensional continuous spaces. He completed his PhD at the University of Vienna, MS in computer science from Simon Fraser University in 2012, and his BS in computer science from Georgetown University in 2002. His other projects include interactive visualization systems for understanding complex computer simulations and programming language design.

Noeska Smit, University of Bergen

Medical Visualization: Visualizing the Invisible

June 19th, 2019, 16:00 CET,

Room: Science Park Building 3, Room 063

Abstract:
In the field of medicine, there has been a tremendous growth in the amount of medical imaging data that is acquired, such as computed tomography (CT) or magnetic resonance imaging (MRI) scans. In medical visualization, computer-based interactive visual representations of such data are made, often aiming at improved diagnosis, treatment planning, treatment guidance, and/or doctor-patient communication. At times, there are structures that are not visible in the original medical imaging scans. When these invisible structures are nerves, and they are at risk for damage during surgery, visualizing these invisible structures can be crucial to avoid such damage. In this talk, I will present an overview of my research in this area, aiming at enhancing medical imaging data by integrating information from various sources into a combined interactive visual representation for educational, clinical, and research purposes.

About the Speaker:
Noeska Smit is an Associate Professor in Medical Visualization in the Visualization Group at the Department of Informatics, University of Bergen, Norway. She is also affiliated to the Mohn Medical Imaging and Visualization Centre at the Department of Radiology, Haukeland University Hospital, where she holds a position as a Senior Researcher. Her research interests include model-based visualization for surgical planning and education, as well as interactive visualization of multi-modal medical imaging data. She received the Dirk Bartz Prize for Visual Computing in Medicine (Eurographics Medical Prize) in 2019. Dr. Smit received her PhD from Delft University of Technology in the Netherlands.

Daisuke Iwai, Osaka University

Computational Projection Mapping

May 22nd, 2019, 10:00am CET,

Room: Science Park Building 3, Room 063

Abstract:
Projection mapping dynamically augments the appearance of a real surface by digital image projection. It has been applied in many application fields such as medicine, entertainment, and product design. However, the projected results are always suffered from the reflectance properties of the surface such as spatially varying textures, subsurface scattering, and inter-reflection, and also from projector's technical limitations such as low dynamic range, narrow depth-of-field, and latency. To resolve these issues and enhance the image quality of projection mapping, several technologies based on computational display approaches have been developed. Computational display is the joint design of hardware with computational algorithms. A unique and interesting property of computational display for projection mapping is that the hardware we try to optimize includes not only optics but also surfaces, while normally only optics has been considered in computational display researches for other displays. This talk introduces a series of computational projection mapping researches, and also discuss its new application field---human augmentation.

About the Speaker:
Daisuke Iwai received his B.S., M.S., and Ph.D. degrees from Osaka University, Japan, in 2003, 2005, and 2007, respectively. He was a visiting scientist at Bauhaus-University Weimar, Germany, from 2007 to 2008, and a visiting Associate Professor at ETH, Switzerland, in 2011. He is currently an Associate Professor at the Graduate School of Engineering Science, Osaka University. His research interests include spatial augmented reality (a.k.a. projection mapping) and projector-camera systems. He published 100+ research papers in the related areas and received several academic awards including Best Paper Award in IEEE VR 2015, Best Paper Award in IEEE 3DUI 2015, Best Paper Runner-up Award in IEEE ISMAR 2016, and Best Research Demonstration Runner-up Award in IEEE VR 2019. He is a member of the IEEE.

Roman Pflugfelder, Austrian Institute of Techhnology (AIT)

Introducing Visual Object Tracking From Classical Views to Machine Learning

April 10th, 2019, 12:30pm CET,

Room: Science Park Building 2, Room 120

Abstract:
Visual object tracking is a fundamental and important task in computer vision. Tracking is an essential prerequisite of motion analysis which is important to many problems such as motion capturing, object recognition and scene understanding. Tracking research started in the early 1950s with Claude Shannon’s exceptional work on information theory and Philip Woodward’s contributions to radar research. Since then, tracking has become a research field in various scientific disciplines, considering beside visual data, state estimation in dynamical systems or the analysis of time series. Despite the efforts, visual tracking is an open problem, lacking in a sufficient theoretical understanding and in practical algorithms for a large number of applications. This lecture will give an introduction to tracking and motion analysis, its challenges and applications supported by practical examples. We will learn the theoretical views on the problem, which are currently prevalent in literature, especially by focusing on tracking single objects. Finally, the talk will present a rather popular view on tracking by seeing the problem from a machine learning perspective. A compact overview of different learning situations such as adaptation, semi-supervised learning, unsupervised deep learning with different representational models under different algorithmic design concepts is given. The talk concludes with a summary of the current state-of-the-art concerning performance, open problems and potential future work.

About the Speaker:
Roman Pflugfelder is Scientist at the AIT Austrian Institute of Technology and lecturer at TU Wien. He received in 2002 a MSc degree in informatics at TU Wien and in 2008 a PhD in telematics at the TU Graz, Austria. In 2001, he was academic visitor at the Queensland University of Technology, Australia. His research focuses on visual motion analysis, tracking and recognition applied to automated video surveillance. He aims to combine sciences and theories in novel ways to gain theoretical insights into learning and inference in complex dynamical systems and to develop practical algorithms and computing systems. Roman contributed with more than 55 papers and patents to research fields such as camera calibration, object detection, object tracking, event recognition where he received awareness of media as well as several awards and grants for his scientific achievements. Roman is senior project leader at AIT where he has been managing cooperations among universities, companies and governmental institutions. Roman co-organised the Visual Object Tracking Challenges VOT'13-14 and VOT'16-18 and was program co-chair of AVSS'15. Currently he is steering committee member of AVSS. He is regular reviewer for major computer vision conferences and journals.

David Gotz, University of North Carolina

Visual Analytics for Cohort Selection and Analysis

March 13th, 2019, 16:00pm CET,

Room: Science Park Building 3, Room 063

Abstract:
Given the widespread adoption of electronic health systems, clinical institutions are amassing ever-larger collections of patient-centered data. These repositories hold detailed longitudinal records capturing a vast variety of patient histories and outcomes.  Visual analysis techniques--designed to mine, analyze, and visualize data-driven insights--are enabling new opportunities to leverage these vast longitudinal resources for population health applications.  This talk will provide an overview of my research in this area, with a focus on visual analytics methods for large-scale temporal event data such as longitudinal electronic health records.  It will include demonstrations of new methods being developed to provide practitioners with exploratory cohort selection and analysis tools that are faster, more intuitive, and more reliable.

About the Speaker:
David Gotz is an Associate Professor of Information Science at the University of North Carolina at Chapel Hill (UNC), where he leads the Visual Analysis and Communications Lab (VACLab).  He is also Assistant Director for the Carolina Health Informatics Program (CHIP) and Associate Director for UNC's NIH-funded Big Data to Knowledge Pre-Doctoral Training Program.  Prior to joining UNC in 2014, Dr. Gotz was a Research Scientist at the IBM T.J. Watson Research Center in New York where he served as technical lead for visual analytics and data analysis systems in the Department of Health Informatics Research, and member of the Intelligent Information Interaction research group.  Dr. Gotz received his Ph.D. in Computer Science from UNC in 2005.

Miriah Meyer, University of Utah

Research Through Visualization Design Study

January 8th, 2019, 14:00pm CET,

Room: Science Park Building 2, Room 054

Abstract:
Designing effective visualizations requires a careful consideration of factors beyond aesthetics and functionality — it requires deeply understanding the needs, intuitions, and goals of target users. Visualization design studies are a methodical approach for acquiring this understanding. In this talk I’ll discuss the way we conduct design studies in my group, and how we use what we learn to contribute new visualization knowledge. I’ll also layout some open challenges in design study research along with a several proposed solutions that we’ve developed.

About the Speaker:
Miriah is an associate professor in the School of Computing at the University of Utah and a faculty member in the Scientific Computing and Imaging Institute. She co-directs the Visualization Design Lab, which focuses on the design of visualization systems for helping analysts make sense of complex data, as well on the development of design methods for helping visualization designers make sense of real-world problems. She obtained her bachelors degree in astronomy and astrophysics at Penn State University, and earned a PhD in computer science from the University of Utah. Prior to joining the faculty at Utah Miriah was a postdoctoral research fellow at Harvard University and a visiting scientist at the Broad Institute of MIT and Harvard.

Miriah is the recipient of a NSF CAREER grant, a Microsoft Research Faculty Fellowship, and a NSF/CRA Computing Innovation Fellow award. She was named a University of Utah Distinguished Alumni, both a TED Fellow and a PopTech Science Fellow, and included on MIT Technology Review's TR35 list of the top young innovators. She was also awarded an AAAS Mass Media Fellowship that landed her a stint as a science writer for the Chicago Tribune.