Prof. Winston Hsu,
Professor, National Taiwan University & Director, NVIDIA AI Lab (NTU)
Prof. Winston Hsu is an active researcher dedicated to large-scale image/video retrieval/mining, visual recognition, and machine intelligence. He is a Professor in the Department of Computer Science and Information Engineering, National Taiwan University and co-leads Communication and Multimedia Lab (CMLab). He and his team have been recognized with technical awards in multimedia and computer vision research communities including IBM Research Pat Goldberg Memorial Best Paper Award (2018), Best Brave New Idea Paper Award in ACM Multimedia 2017, First Place for IARPA Disguised Faces in the Wild Competition (CVPR 2018), Third Place (mini-track) for Moments in Time Challenge (video action recognition) in CVPR 2018,Third Place for 2018 IEEE Signal Processing Society Video and Image Processing (VIP) Cup, First Prize in ACM Multimedia Grand Challenge 2011, First Place in MSR-Bing Image Retrieval Challenge 2013, ACM Multimedia 2013/2014 Grand Challenge Multimodal Award, ACM Multimedia 2006 Best Paper Runner-Up, etc.
He is keen to realizing advanced researches towards business deliverables via academia-industry collaborations and co-founding startups. Working closely with the industry, he was a Visiting Scientist at Microsoft Research Redmond (2014) and had his 1-year sabbatical leave (2016-2017) at IBM TJ Watson Research Center, New York, to enhance Watson's visual cognition, where he contributed the first AI produced movie trailer. He is the Founding Director for NVIDIA AI Lab (NTU), the 1st in Asia. He received Ph.D. (2007) from Columbia University, New York. Before that, he was a founding engineer and research manager in CyberLink Corp. He serves as the Associate Editor for IEEE Transactions on Circuits and Systems for Video Technology (TCSVT) and IEEE Transactions on Multimedia, two premier journals, and was in the Editorial Board for IEEE Multimedia Magazine (2010 - 2017). He also co-organized several premier conferences such as ACM Multimedia, ACCV, ICME, ICMR, ICIP, etc.
Keynote Title: Approximating Human Perception via Multimodal Data Streams
Abstract: Images, videos, audio, 3D point clouds (from LiDAR or RGB-D), etc., are major data types nowadays essential for disruptive opportunities in domains such as entertainment, security, healthcare, manufacturing, etc. Such multimodal data streams bring rich and diverse aspects for sensing the the real world and being able to approximate human cognitive capabilities for further manipulations. However, the current techniques are far behind the dire needs. In this talk, aiming to leverage the multimodal data streams, we will focus on very challenging computer vision problems ranging from large-scale object localization, disguised face recognition, low-resolution human recognition, fine-grained action recognition, 3D semantic labeling, etc. We will review the lessons we learned as designing the advanced neural networks which accommodate the multimodal signals in an end-to-end manner. Meanwhile, we identify few key criteria for realizing them into deployed applications as collaborating with industry partners. We will evaluate and compare the effective neural networks components with the state-of-the-art over the public benchmarks.
Prof. Mohan S Kanakanhalli,
National University of Singapore
Prof. Mohan S Kanakanhalli is currently the Professor and Dean, School of Computing and Provost's Chair Professor of Computer Science National University of Singapore. He was born in Ahmedabad, India and did his schooling in Ahmedabad and Vadodra. He then obtained his B-Tech (Electrical Engineering) from the Indian Institute of Technology, Kharagpur, in 1986 and his MS and PhD (Computer and Systems Engineering) from the Rensselaer Polytechnic Institute in 1988 and 1990, respectively.
He subsequently joined the Institute of Systems Science (ISS - then a part of NUS but now it is known as Institute for Infocomm Research and belongs to A*STAR) in October 1990. During 1997-1998, he was a faculty member at the Department of Electrical Engineering of the Indian Institute of Science, Bangalore.
He has been with the NUS School of Computing since May 1998 where he is Provost's Chair Professor of Computer Science. He is also the Dean of NUS School of Computing. Before that, he was the Vice Provost (Graduate Education) for NUS during 2014-2016 and Associate Provost (Graduate Education) during 2011-2013. He was earlier the School of Computing Vice-Dean for Research during 2001-2007 and Vice-Dean for Graduate Studies during 2007-2010. He visited the Garage Cinema Group at the University of California at Berkeley during Jan-Jun 2004 and the Media Futures Group at University College London during Jan-Apr 2013. Mohan is a Fellow of IEEE.
Keynote Title: Privacy-aware Computing for Smart Cities
Abstract: The vision of building smart cities relies on data-driven and data-hungry algorithms, services, and applications. However, the underlying data often contains very sensitive information about people such as their identity, demographic data (age, gender, ethnicity, etc.), location, health conditions, financial records, etc. A breach of privacy can have negative consequences for an individual, a company or an organization. Therefore, the trade-off between privacy and utility needs systematic exploration. Current efforts towards privacy preservation have focused more on structured data, i.e., data comprising of well-defined and individual attributes. Increasingly, much of the information collected and shared is unstructured and multimodal data such as text, images, video, and audio. This data potentially contains sensitive information such as faces, location information, (quasi-)identifiers (e.g., car registration number, passport numbers, etc.) shown in an image. However, naïve mechanisms to preserve privacy, e.g., simply blurring images in an uncontrolled manner, will significantly reduce data utility like impeding the predictive power of machine learning algorithms applied on such data. In this talk, we will present examples of work done by researchers at N-CRiPT. We will discuss some of the critical issues related to privacy in the context of smart cities and potential solutions towards addressing them. In particular, we will touch upon specific privacy problems related to video surveillance, biometrics, smart grids, Internet-of-Things as well as describe novel privacy techniques for data publishing and machine learning.
Prof. Josef Bigun,
Halmstad University, Sweden
Josef Bigun obtained his M.Sc. and Ph.D. degrees from Linkoeping University, in 1983 and 1988 respectively. In 1988, he joined the Swiss Federal Institute of Technology in Lausanne where he worked as Adjoint Scientifique until 1998 with the exception that in 1997 he was Visiting Professor at the Royal Institute of Technology, (KTH) Stockholm.
He has been elected Professor to the Signal Analysis Chair, his current position, at Halmstad University and Chalmers Institute of Technology in 1998. He has been in technical and organizational committees of several national and international conferences. In particular he co-chaired the first international conference on Audio and Video Based Person Authentication in 1997. He has been contributing as a referee or as an editorial board member of international journals including Image and Vision Computing, Pattern Recognition Letters and IEEE Image Processing. He served in the executive committees of several scientific associations, including the international association for pattern recognition, IAPR.
He has contributed to the initiation and progress of several national, e.g. VR and SSF projects, and international projects consortia, e.g. the EU projects BBfor2, BIOSECURE, IT-VIRSBS and ACTS-M2VTS. His scientific interests include a broad field in Computer Vision and pattern recognition including biometric signal analysis, texture analysis, motion analysis, 3-D modelling and understanding of the biological recognition mechanisms of audio-visual signals. He has been awarded the grades of fellow of IAPR, and fellow of IEEE.
Keynote Title: Dense Maps of Local Fourier Features Applied to Biometrics
Abstract: A summary of main Fourier features along with their importance to human visual intelligence will be given. These features comprise local direction, local (absolute) frequency, and local phase which in turn drive more complex models of shape, also they being dense.
Wheras fingerprint wave structures can be modelled by frequency and direction maps alone, minutiae in the form of ridgeends and bifurcations call for modelling these as first order angular variation of local phase. Likewise, complex global patterns
in the form of cores and deltas demand a modelling of dense direction maps using angular variations of order 1 and -1. Similar example applications from other biometrics will be visited, including identity recognition using iris, periocular regions, lip movements, and full faces.
However intricate, the underlying shape models demand a principled and robust estimation of local direction and frequency, independent of each other. With such estimations at hand one can recognize more complex patterns, including those using local phase, as delivered by Gabor filters associated with estimated frequencies and directions. Tools using complex valued filters on complex valued dense maps, easing the practice of more intricate shape recognition will be presented. Examples of shape models built on top of basic dense fields, which in themselves yield dense responses, will be given along their use.
Dr. Sharathchandra U. (Sharath) Pankanti,
Manager, Exploratory Computer Vision Group; Thomas J. Watson Research Center, Yorktown Heights, NY USA
Sharath Pankanti (http://researcher.ibm.com/person/us-sharat) is Principal Research Staff Member in Cognitive Computing Department at the Thomas J. Watson Research Center. He received Ph.D. degree in Computer Science from the Michigan State University. Sharath has led a number of safety, productivity, education, healthcare, and security focused projects involving biometrics, multi-sensor surveillance, rail-safety, driver assistance technologies that entail object/event modeling, detection and recognition from information provided by static and moving sensors/cameras. Results of many of these efforts have demonstrated competitive results in scientific evaluations (NIST TRECVID-2012, 2013 and 2014, ImageClef 2013) and been integrated into real world applications. His work contributed to world's first large scale biometric civilian fingerprint identification system in Peru and to award1 winning IBM surveillance offering2that have been featured in news media (ABC3/Fox/CBS/NBC), mentioned in popular TV media (CSI:Miami) and covered in social (good.is) media. He is a co-author of over 150 peer-reviewed publications (over 16,000 citations with h-index 46 Google Scholar) published in many reputed venues, including Scientific American, IEEE Computer, IEEE Spectrum, Comm. ACM, and Proc. IEEE. He is also co-inventor of more than 100 inventions which has generated significant IP revenue for IBM. His efforts have been recognized by IBM as significant accomplishments including Master Inventor (~1% of IBM employees are selected as master inventors) and Outstanding Accomplishment Awards. Dr. Pankanti co-edited the first comprehensive book on biometrics, "Biometrics: Personal Identification" Kluwer, 1999 and co-authored, "A Guide to Biometrics", Springer 2004 which has been used in many biometrics curricula. He is Fellow of IEEE, IAPR, and SPIE. He has served the computer vision and pattern recognition community in various capacities over last two decades and most recently volunteered as part of IEEE Distinguished Visitor and ACM Distinguished Speaker programs.
Keynote Title: Securing Senses/Operations to Free New World: Trades and tools for the emerging Future
Abstract: For centuries, we humans have prided in our uncanny ability to discriminate fakes from reality using our own senses. We were comforted that this our uncanny ability will serve us well in detecting manipulated reality/content and let us objectively find our bearings in increasingly deceptive adversarial world. If experts are to be believed, all this is about to change: thanks to increasingly accessible array of sophisticated technological tools supplied by confluence of advances in machine learning, availability of annotated data, and computing power, in not very distant future it would be relatively easy to project manipulated reality that will increasingly suspend, challenge, and mock at our very sense of securely grounded reality. Not surprisingly, there are vast implications of such technologies ranging from narrowly causing personal sense of disorientation to widespread dysfunctions of deeply rooted institutions that are critical to the very existence of fair and free societies. In this two part talk, we will first present burgeoning technologies that enable effective projection of fake reality and misrepresentations supported by real recent incidents in the mainstream news. The second part of the talk will explore various tools of the trade (such as blockchain and encrypted processing) that could potentially help us not only disambiguate the treacherous landscape of disbelief and distrust but also guide us to safely and securely operate therein.