Workshop on

 Human Activity Recognition and Modelling

HAREM 2005

To be held at the
BMVC - British Machine Vision Conference

September 9th, 2005
Oxford, UK

Sponsored by EU Project IST 2001 37540 CAVIAR 

Local organisation by
Instituto Superior Técnico/ Institute for Systems and Robotics




The HAREM registration is free of charge but registration is required.
Registration is however mandatory for at least one author for each paper that will attend the workshop and present the work. The proceedings will only be distributed to registered participants, with priority to (registered) authors or PC-members. All other (registered) participants will be served on a first-come-first-served basis. A maximum number of participants may be set at any time before the workshop.
Registration form: [ doc ] [ pdf ]

The HAREM organizers are not providing the accomodation logistics. Please book your accomodation through the BMVC web site, even if you are not attending BMVC. To do that, you have to fill in the registration form where you can book for the desired number of extra nights. You do not have to attend BMVC if you do not want to. Then the form must be faxed (or emailed), with payement instructions, to the number given in the form.

Getting there
Those attendees flying to Heathrow or Gatwick Airports can use regular busses to Oxford (about 30 minutes frequency) that stop at Brookes University. Harem will be held at the Gipsy Lane Campus of Oxford Brookes University about one mile from Oxford city centre. Check the BMVC travel page for additional information.

The HAREM workshop will be held at the Darcy building, see map for details.


Friday, 9 September 2005

09h30 - 10h45: DETECTION/DYNAMIC MODELING (3x25)
10h45 - 11h20: COFFEE BREAK
13h00 - 14h00: LUNCH
14h00 - 15h15: ACTIVITY RECOGNITION I (3x25)
15h15 - 15h45: COFFEE BREAK
15h45 - 17h00: ACTIVITY RECOGNITION II (3x25)
17h00 - 17h30: DISCUSSION


Detailed Program:


A Vision System for Automated Customer Tracking for Marketing Analysis: Low Level Feature Extraction, Alex Leykin and Mihran Tuceryan, Indiana university, USA.

Modeling Behavior Trends and Detecting Abnormal Events using Seasonal Kalman Filters,
James W. Davis and Mark A. Keck, Ohio State university, USA.

Classification and Prediction of Motion Trajectories using Spatiotemporal Approximations,
Andrew Naftel and Shehzad Khalid, University of Manchester, UK.

10h45 - 11h20: COFFEE BREAK

Human Motion Recognition based on Dynamic Shape Analysis, Ning Jin and Farzin Mokhtarian, University of surrey, UK.

Continuous Time-Varying Gesture Segmentation by Dynamic TimeWarping of Compound Gesture Models, Hong Li and Michael Greenspan, Queen's University, Kingston, Canada.

What are you looking at? Gaze estimation in medium-scale images, Neil Robertson, Ian Reid and Michael Brady, QinetiQ and Oxford University, UK.

Human Gait Recognition with 3DWavelets and Kernel based Subspace Projections, Nasir Rajpoot and Khalid Masood, Univ Warwick, UK; Inform.Inst. Technology, Pakistan.

13h00 - 14h00: LUNCH

Human Activity Recognition from Video: modeling, feature selection and classification architecture. Pedro Canotilho Ribeiro and José Santos-Victor, Instituto Superior Técnico, ISR, Portugal.

Activity Recognition via Autoregressive Prediction of Velocity Distribution, Miha Peternel and Ales Leonardis, University of Ljubljana, Slovenia

Segmentation and Classification of Human Activities, J. Nascimento, M. Figueiredo and J. S. Marques, Instituto Superior Técnico, Portugal.

15h15 - 15h45: COFFEE BREAK

Human Activity Learning and Segmentation using Partially Hidden Discriminative Models, Tran The Truyen, Hung H. Bui and Svetha Venkatesh, Curtin Univ, Australia; SRI Intl, USA.

Exploring Techniques for Behaviour Recognition via the CAVIAR Modular Vision Framework, David Tweed, Wanli Feng, Robert Fisher, José Bins and Thor List, Univ. of Edimburgh, UK.

Recognition of Action, Activity and Behaviour in the ActIPret Project, Kingsley Sage, Jonathan Howell and Hilary Buxton, University of Sussex, UK.

17h00 - 17h30: DISCUSSION



Recent advances in computer vision and learning methodologies, together with the massive increase of computational power of standard computers enabled the deployment of a new generation of computer vision systems that go beyond more traditional approaches designed mainly for modelling the scene geometry from video data. Instead, the challenge is to develop systems able to detect and signal interesting events and understand, interpret and describe the observed video sequence. 
One potential application of such methodologies lies in the context of video surveillance. Due to the massive number of cameras deployed in public spaces, it is no longer possible (or efficient) to have human operators continuously monitoring a multitude of video channels. Instead we need to provide computer vision systems able to process these video streams, use and learn contextual information, characterize the behaviour in a given situation and trigger an alert, only when an interesting “event” is detected. Needless to say, these systems must be designed to function and learn in an open-ended way and must have built-in self-regulatory, and reconfigurable capabilities. 
The goal of the workshop is to bring together researchers in cognitive computer vision and particular in the domains of surveillance and human activity recognition, with an emphasis on the following topics:
- Low-level feature extraction and selection
- Modelling of attention and control
- Human activity recognition
- Cognitive and self-adaptive architectures
- Video interpretation


Submission Deadline:           June 1st, 2005  
Notification of acceptance:  July 1st, 2005 ------- [ ACCEPTED PAPERS ]
Camera-ready copy due:     July 20th 2005
Workshop:                        September 9, 2005
Please submit your paper as a PDF email attachment to
Papers should not be longer than 10 pages (use BMVC format). The paper should be anonymous during the review process. Any indications of name and affiliation should be removed from the paper. 
In the email body, please include the following information:

* Title
* Authors (first and last names)
* Affiliation of authors (institute, address)
* Corresponding author (including his/her address, email, phone, fax)

Papers that have also been submitted to the main BMVC conference will be considered for review. Double submission must be indicated by authors. Papers must be withdrawn if accepted for BMVC.


José Santos-Victor  Instituto Superior Técnico, ISR, Portugal (Chair)
Bob Fisher              Univiversity of Edinburgh, UK.
James Crowley        INRIA- Rhone Alpes, France.



Alexandre Bernardino, Instituto Superior Técnico / ISR (Portugal)
Aaron Bobick,           Giorgia Tech, (USA )
James Ferryman,       University of Reading (UK)
Daniela Hall,             INRIA Rhône-Alpes (France)
David Hogg,             University of Leeds (UK)
Graham Jones,          Kingston University (UK)
Seong-Whan Lee,      Korea University (Korea)
Jorge Marques,         Instituto Superior Técnico/ ISR (Portugal)
Filiberto Pla,             University Jaume I, (Spain)
Hiroshi Sako,            Hitachi, (Japan)
Monique Thonnat,     INRIA Sophia Antipolis, (France)


Prof. José Santos-Victor
Institute for Systems and Robotics
Instituto Superior Técnico
Av. Rovisco Pais,
1049-001 Lisboa
Phone: +351 (218 418 294) (direct line)
Fax: +351 218 418 291