British Machine Vision Association and Society for Pattern Recognition

Understanding Visual Behaviour

One Day BMVA Workshop in association with the IEE/E4 at the British Institute of Radiology, London, UK on 24 January 2001

Chairpersons: Shaogang Gong, Jamie Sherrah (Queen Mary College ), Hilary Buxton (University of Sussex)


Computational understanding of human behaviour from visual data is crucial for computer vision systems that aim (1) to provide a more natural human-computer interface experience, (2) to create realistic visually augmented virtual sets in interactive 3D TV productions, (3) to facilitate visually mediated telecommunication, and (4) to build systems that can process surveillance data unsupervised.

However, segmenting and modelling humans in visual scenes is far from trivial due to the deformable articulated structure and complex dynamics of the human body, resulting in ambiguities when trying to measure human behaviour from visual data. The problem of interpreting behaviours of humans and their activities is necessarily made even more challenging because behaviour cannot be visually observed or measured from images. It is commonly hard even for human subjects to describe their own behaviour adequately.

This meeting will present recent progress and also highlight some of the challenges in modelling and understanding visual behaviour of human bodies and activities. In particular, it will cover the following topics of interest.


10:00 Registration and coffee

10:25-12:30 Morning Session (chair Shaogang Gong)

10:25 Introduction and welcome, Shaogang Gong

10:30 A Model of Human Interaction, David Hogg (Leeds) Download Real Movie of Presentation

11:00 A Connectionist System for Visually Mediated Interaction, Hilary Buxton (Sussex) Download Real Movie of Presentation

11:30 A Probabilistic Sensor for the Perception and the Recognition of Activities, Jim Crowley (INRIA Rhones Alpes) Download Real Movie of Presentation

12:00 A New Model for Visual Attention, Fred Stentiford (BT) Download Real Movie of Presentation

12:30-13:30  Poster Session (chair Jamie Sherrah)

Automatic Gait Recognition via the Generalised Symmetry Operator, Hayfron-Acquah, Nixon and Carter (University of Southampton)

New Area Measures for Automatic Gait Recognition, Foster, Prugel-Bennett and Nixon (University of Southampton)

A Hierarchical Model of Dynamics for Tracking People with a Single Video Camera, Karaulova, Hall and Marshall (Universities of Cardiff and Bath)

Modelling Human Gait using Motion Sensors, Gill Hayes (Edinburgh)

Learning some geometric properties of a 3D scene, Makris and Ellis (City University)

Tracking Drivers' Hand Movements, McAllister, McKenna and Ricketts (University of Dundee)

13:30 Lunch

14:30-16:30 Afternoon Session (chair Hilary Buxton)

14:30 Simultaneous Perception and Classification of Visual Motion, Andrew Blake (Microsoft) Download Real Movie of Presentation

15:00 Visual Interaction using Gesture and Behaviour, Shaogang Gong (Queen Mary College) Download Real Movie of Presentation

15:30 Tea

16:00 Action Reaction Learning for Predicting Interactive Behaviour, Tony Jebara, Alex Pentland (MIT) Download Real Movie of Presentation

16:30 Marker-less Motion Capture for Studio Production, Adrian Hilton (Surrey) Download Real Movie of Presentation

17:00 Summary and discussions

17:15 Closing remarks