Keynote Lectures

Abstract
The aim of crowd modelling is to propose mathematical representations and numerical methods for calculating the movement of simulated crowds in order to understand, reproduce or predict the behaviour of real crowds. This subject is at the interface of many disciplines, including mathematics, physics, biology, cognitive sciences and, of course, computer sciences. Because crowds generate visually fascinating patterns, computer graphics has proposed important contributions to the field from the outset, especially through the seminal work of Craig Reynolds who proposed the well-known Boids model in the late 80s to enable the visual effects of films such as Burton's Batman Returns.
Almost 40 years have passed.
In this talk, I will put into perspective some recent results and advances in the field from a computer graphics perspective. I will present a variety of specific and recent contributions from our discipline to the field of crowd modelling, and why I think our community has major assets to shape the future of this research, well beyond the limits of the field of visual effects for cinema, touching on the safety of mass events, for example.

Abstract
Machine learning and data science are receiving significant attention and rightly so. The results that can be produced by distilling large amounts data are amazing.  Yet, what society expects is human oversight at an appropriate level and trust of system results.  Oversight and trust are not for machines; oversight and trust are for humans.  Effective solutions thus require careful human-machine collaboration and in turn careful visualisation design.  In this talk, I motivate why visualisation design forms a necessary part of a data driven society.  I provide motivation for carefully designed visualisations that must take into account human perceptual factors, target audience, and the automated processes applied to the information before visualisation.  Throughout the talk, I provide practical examples where all three must be carefully thought about in order to deliver effective data science.

Abstract
Touch is far less understood than vision or hearing, since what you feel greatly depends on how you move, and since engineered haptic sensors, actuators, and algorithms typically struggle to match human capabilities. My team and I work to sharpen our understanding of haptic interaction by investigating the dual approaches of haptic interfaces and autonomous robots. Haptic interfaces are mechatronic systems that modulate the physical interaction between a human and their tangible surroundings. Typically taking the form of grounded kinesthetic devices, ungrounded wearable devices, or surface devices, such systems enable the user to act on and feel a remote or virtual environment. I will elucidate key approaches to creating outstanding haptic interfaces by showcasing examples of my research on both kinesthetic and wearable devices. Then I will draw parallels to autonomous robots, where the engineered system acts as an agent rather than a tool and needs to detect and intelligently react to (rather than generate) haptic signals. In addition to inventing tactile sensors and touch-processing approaches, we have created several physically interactive robots, including HuggieBot, a custom robot that uses real-time haptic sensing to give good hugs.

Abstract
Computer vision has found its way towards an increasingly large number of applications. Yet, most standard learning approaches lead to fragile models prone to drift when incremental updates are performed. Many lifelong learning methods have been proposed to mitigate this ‘catastrophic forgetting’ issue. Lifelong learning, which had been envisioned since the early days of computer science, has recently gained more traction in computer vision. It is now being revisited in light of the large pre-trained visual and multimodal models that have become available. This presentation will discuss recent methods that leverage such large models for continual learning applications.