apidis is an ambitious and challenging project that builds on recent scientific and technological breakthroughs in the fields of user-centric design, signal processing, computer vision, machine learning, and planning and optimized decision process. The APIDIS Consortium is made up of leading experts from academia and industry.

Partner technical and scientific expertise:

To implement the APIDIS system, significant technological and scientific advances were made in fields related to:

• The extraction of meaningful content from the video sequences captured by a set of complementary cameras (low and high resolution, omnidirectional and conventional), distributed within/around an area of interest. It implies distributed scene analysis and event recognition mechanisms so as to focus on the content-of-interest (or identify the segments of content that should not be displayed). It also requires generating TV-quality video content by cropping the high-resolution image captured by a fixed camera or by processing the information collected by an array of omnidirectional cameras. We deliberately focus on still (omnidirectional) cameras because it prevents the cost associated to the maintenance and real-time control of a robotic moving PTZ. Regarding the audio part, we only consider standard ambiance sound capture and recognition (e.g. recognize the referee whistles), and do not envision specific advance or development regarding speech or audio processing.
  
  
• The automated production of video content. In particular, our project targets autonomous production of a coherent story from the captured video content. Given a raw content, and a set of hints about the informative value of its segments, the objective is to produce a nice-looking (because it respects general production rules. Here, nice-looking does not refer to the video resolution.) and informative (because it favors informative segments) video summary that meets a pre-defined duration constraint. Inclusion of advertisements within the produced content could obviously also be considered during the summarization process.
  
  
• The personalized summarization of the content. Our project aims at enabling flexible and personalized creation of the content. The content stored on a server is likely to be accessed for different purposes, through heterogeneous channels and devices. Both a short low-resolution overview and a long detailed report of the event are likely to be requested. In some cases, on-line user feedback may also control the production of the content. Hence, the format adopted to store and manipulate the initial raw content has to be flexible enough to enable efficient and personalized creation. Existing standards (JPEG2000, SVC) already offer flexible access to the raw video data, but have to be augmented to support the semantically-driven summarization of the content.

The apidis partners are:

* Université catholique de Louvain, UCL, Belgium, http://www.uclouvain.be/
* Queen Mary & Westfield College, University of London, QMUL, UK, http://www.qmul.ac.uk/
* Ecole Polytechnique Fédérale de Lausanne, EPFL, Switzerland, http://www.epfl.ch/
* Barcelona Media – Centre d’Innovació, BMCI, Spain, http://www.barcelonamedia.org/
* ACIC sa, ACIC, Belgium, http://www.acic-tech.be/
* Mediapro, MP, Spain, http://www.mediapro.es/