Keynote Speakers

Professor Touradj Ebrahimi

Professor Touradj Ebrahimi

Professor, École Polytechnique Fédérale de Lausanne (EPFL), CH

Convenor of JPEG Standardization Committee

Title: From Seamless to Trustworthy Visual Communication: The Case of JPEG Trust

Date 14 December 2026 Time 9.30am - 10.30am Location Auditorium, 2.101

Abstract

Historically, communication systems were engineered with a singular, overarching goal: to provide universal access to information, anytime, anywhere, and for anyone. This mission is particularly challenging for visual information. As a universal vector of exchange, images and video transcend linguistic barriers, while they require the transmission and storage of massive volumes of data. The development of highly efficient coding technologies, most notably the JPEG and MPEG image and video coding formats, successfully removed these technical barriers, facilitating the seamless access to visual information we rely on today.

However, as content creation becomes increasingly democratized, accelerated by the rapid rise of Generative AI, two critical challenges have emerged. First, in a seamlessly connected ecosystem, protecting the intellectual property of content creators and ensuring proper copyright attribution is paramount. Second, this frictionless accessibility can be weaponized by ill-intentioned actors to spread propaganda, deepfakes, and fraud.

To address these vulnerabilities, efficient visual formats must evolve beyond compression to natively support media authenticity and trust. In this talk, we will focus on this critical transition and introduce JPEG Trust (ISO/IEC 21617), a new international standard recently published by the JPEG standardization committee. JPEG Trust provides a robust, international framework for establishing trust in digital media assets, offering secure tools for provenance tracking and authenticity verification, while safeguarding authorship and ownership in an era where seeing is no longer believing.


Biography

Touradj Ebrahimi is a Professor of Image Processing at the École Polytechnique Fédérale de Lausanne (EPFL), where he directs the Multimedia Signal Processing Group (MMSPG). A globally recognized leader in media technologies and INternational standardization, he serves as the Convenor of the JPEG standardization committee.


Professor Lin Weisi

Professor Lin Weisi

Professor, College of Computing & Data Science, Nanyang Technological University, SG

Associate Dean (Research), President's Chair in Computer Science

Title: Beyond Scores: Toward Evidence-Grounded, Generalizable Visual Quality Assessment with LMMs

Date 15 December 2026 Time 9.00am - 10.00am Location Auditorium, 2.101

Abstract

Visual quality assessment (VQA) aims to tell every imaging system, from cameras and codecs to AIGC models, whether it is doing its job, for human users and increasingly for machines, and has evolved from handcrafted signal features to deep predictors. Yet most systems merely return a scalar score with limited evidence of what degraded a signal, where problems lie, if the judgment could transfer beyond the data on which the model was trained, and what is the actionable improvement.

This talk argues that large multimodal models (LMMs) can help move VQA from scalar prediction toward evidence-grounded quality reasoning, in diversified scenarios. It first examines what LMMs can and cannot perceive about low-level visual quality, including subtle artifacts, spatial localization, and hallucination. It then discusses how to teach LMMs to use text-defined quality levels, provide localized quality diagnosis, and make pairwise comparisons—a setting that can be more stable than absolute scoring.

We then extend the discussion to AI-generated imagery, where perceptual quality, plausibility, and prompt consistency interact, and to machine-oriented quality, where the relevant observer is a downstream model and the criterion is task reliability rather than human preference.

It concludes with a research agenda: combining foundation models with perceptual priors such as masking, attention, and just-noticeable difference; evaluating explanation faithfulness and uncertainty; and addressing computational cost and temporal quality of long video. The goal is a quality assessor to assess, explain, compare, and more importantly, guide actions.


Biography

Weisi Lin is an active researcher in image and video processing, perception-based signal modelling and assessment, and multimedia communication. He had been the Lab Head, Visual Processing, Institute for Infocomm Research (I2R), Singapore. He is currently a President’s Chair Professor in College of Computing and Data Science, Nanyang Technological University (NTU), Singapore, where he also serves as the Associate Dean (Research). He is a Fellow of IEEE and IET.


Dr. Liu Shan

Dr. Liu Shan

Distinguished Scientist and General Manager, Tencent, US

Editor-in-Chief, IEEE Transactions on Circuits and Systems for Video Technology

Title: Perceptual AI – Video Coding for Machines

Date 14 December 2026 Time 11.00am - 12.00pm Location Auditorium, 2.101

Abstract

Video is occupying about 80% of today’s Internet traffic. Out of this, more and more video contents are consumed by machines, in broad applications such as video surveillance, healthcare monitoring, transportation, smart cities, etc.

Machine vision is different from human vision in many aspects, therefore video codecs optimized for human vision may be sub-optimal for machines. Great interests and demands have risen in recent years to develop video coding technologies and solutions for machine vision requirements and use cases.

This talk will introduce and discuss technology and standard development, applications, challenges, as well as envision future potentials that are related to video coding for machines (VCM).


Biography

Shan Liu (Fellow, IEEE) received her B.E. degree from Tsinghua University and her M.S. and Ph.D. degrees from University of Southern California. She held senior technical and management positions at Sony, MERL, MediaTek, etc. and is now a Distinguished Scientist and General Manager at Tencent.


Professor Jiwen Lu

Professor Jiwen Lu

Department of Automation, Tsinghua University, CN

Professor and Associate Chair of the Department of Automation

Title: Embodied Intelligent Perception and Manipulation

Date 15 December 2026 Time 10:30 AM – 11:30 AM Location Auditorium, 2.101

Abstract

Embodied intelligence is a hot and important research topic in the fields of artificial intelligence and robotics, with significant application prospective in manufacturing, agriculture, and service industry. This talk will present recent major advancements in embodied intelligent perception and manipulation, covering methodologies and technologies such as online perception, autonomous navigation, mobile manipulation, and intelligent control.

It will also explore applications in areas like modern services, industrial manufacturing, deep-sea exploration, and low-altitude security, concluding with an outlook on future development trends.


Biography

Jiwen Lu (IEEE/IAPR Fellow) received the B.Eng. degree in Mechanical Engineering and the M.Eng. degree in Electrical Engineering from Xi'an University of Technology, Xi'an, China, in 2003 and 2006, respectively, and the Ph.D. degree in Electrical Engineering from Nanyang Technological University, Singapore, in 2011. From 2011 to 2015, he was a Research Scientist with the Advanced Digital Sciences Center, Singapore. In November 2015, he joined the Department of Automation at Tsinghua University, where he is currently a tenured Full Professor and Associate Chair of the department.