28.07.2013 Views

Project Proposal (PDF) - Oxford Brookes University

Project Proposal (PDF) - Oxford Brookes University

Project Proposal (PDF) - Oxford Brookes University

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

FP7-ICT-2011-9 STREP proposal<br />

18/01/12 v1 [Dynact]<br />

workshops normally are attended by some 50 specialists, and lead to the publication of books and collections<br />

with the potential of further disseminate the results of the project. These occasions will provide natural venue<br />

in which to establish new academic links and prepare the ground for spin-off or follow-up projects and<br />

national and European level.<br />

Relevant expertise is present in our consortium: Dr Antonucci has organised two special sessions on<br />

imprecise probabilistic graphical models at the IPMU conference (International Conference on Information<br />

Processing and Managing of Uncertainty in Knowledge-based Systems) in both 2010 and 2012.<br />

Organization of Grand Challenges on action recognition in the different scenarios. In recent years, the<br />

PASCAL VOC grand challenge (http://pascallin.ecs.soton.ac.uk/challenges/VOC/) on visual object classes<br />

recognition has greatly contributed to revolutionise computer vision. The VOC challenge was kick-started by<br />

a collaborative project such as the one proposed here. In recent times, an “Action Classification taster” has<br />

been introduced which copes with recognition from single images.<br />

A similar grand challenge on full action recognition would be a natural development along this line.<br />

For its characteristic, this project has a clear potential to given a similar contribution to action recognition in<br />

terms of both methodologies and test-beds: the launch of a grand challenge on this theme could potently<br />

magnify its impact.<br />

Organisation of special issues of international journals on the project's topic. Thanks to the role of many<br />

of the partners in the Editorial Boards of several international journals, we plan to organise special issues on<br />

the various topics covered by the project in the course of the years. In particular, an issue on impreciseprobabilistic<br />

graphical models could fit very well IJAR or AIJ; spatio-temporal discriminative models could<br />

be the topic of a PAMI or Machine Learning special issue; an action or activity recognition challenge could<br />

result in a special issue of IJCV or PAMI.<br />

Building on existing nextwork of links. We also plan to build on the consortium's already strong network<br />

of existing professional links in order to maximise the impact of this project. For instance, OBU is currently<br />

engaged with organisations interested in security matters, such as HMGCC (http://www.hmgcc.gov.uk/), the<br />

UK government centre of excellence whose aim is to design and develop secure communication systems,<br />

hardware and software specifically for HM Government use, both at home and overseas. HMGCC could be<br />

involved towards the final part of the project in order to test government interest in novel biometrics.<br />

Concerning action recognition, OBU can bring added value to the project given its extensive links with<br />

world class companies active in the area such as Microsoft Research Europe, Google, Yotta, Sony<br />

Entertainment. It will be quite straightforward to attract the interest of such industrial behemoths in case the<br />

performances of our prototype meet our expectations. Contacts will be made informally (at least at first) to<br />

probe potential interest. Sony and <strong>Oxford</strong> Metric Group have already manifested interest in writing support<br />

letters for UK level EPSRC project proposals, are since long time collaborating with OBU on the supervision<br />

of several Ph.D. students, and have a history of providing equipment in the form of traditional and range<br />

cameras.<br />

Building new industrial links. At the same time, new industrial links with companies active in recognition<br />

will have to be pursued in order to establish Knowledge Transfer Protocol agreements, proofs of concept or<br />

patents.<br />

An example is a tentative collaboration brokered by <strong>Brookes</strong>’ consultant Eugene Sweeney with DECTEL<br />

Inc. (http://www.dectel.co.uk/), a company active in video surveillance. Many other companies active in<br />

biometrics, vision applications and gait analysis are present in the European Union. They will be targeted via<br />

“customised” seminars in order to present the results obtained by the project in the most interesting way<br />

from the specific point of view and focus of each individual company. Microsoft Research in the person of<br />

Andrew Fitzgibbon, involved in the Kinect project, is being contacted to verify their interest in a formal<br />

collaboration. As commercial enterprises are mostly focused on simpler, indoor environment (as in the<br />

gaming scenario) setting new industrial partnerships could already be possible during the second part of the<br />

project. The Coordinator plans to exploit his personal contacts at Google Zurich (M. Marszalek) and Santa<br />

Monica (A. Bissacco) in this sense. In addition, he plans to attend the next ICT conference in Brussels as an<br />

opportunity for creating new links at EU level, to be later exploited for commercialisation or patenting.<br />

Engaging government bodies and security agencies. Given its repercussions in security applications such<br />

as semi-automatic event detection, or identity recognition from gait, we believe a range of government<br />

<strong>Proposal</strong> Part B: page [58] of [67]

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!