What information can be extracted? Very high value in multimedia content Security Identify risks or provide information Quasi real-time news Get information from people witnessing an event in real time Marketing What are the new emerging trends? Entertainment People spend a lot of time with on-line media. What are they looking for? Human behaviour analysis What do people do in this moment.
Boston bombing Rome – Feyenoord FC supporter riots Security / quasi real-time news
Challenges Effectiveness Extract information relevant pertinent useful Scalability Keep track and handle a continuous flow of media Billion images and videos
Our Skill Multimedia information retrieval on very large scale Prototypes using 100 millions images Similarity search on very large scale Searching by similarity on faces, images, objects, audios, etc. Real time image matching on very limited resources Prototypes recognizing in real-time around 1000 images on mobile phones Image content classification/recognition
Research group Permanent Giuseppe Amato Fausto Rabitti Pasquale Savino Claudio Gennaro Fabrizio Falchi Franca Debole Temporary Costantino Thanos (AS) Fraco A. Cardillo (PD) Claudio Vairo (PD) Paolo Bolettieri (TS)
Fundings and Projects National 210k€ (120k€ soBigData related) /year Inter. 1380k€ (190€ soBigData related) /year On going soBigData related projects: FP7 STREP - RUBICON (2011-2014) CIP-ICT-PSP - EAGLE (2013-2016) PON – DiCET (2013-2014) POR CREO - SECURE (2013-2014),
Research aim Study and develop techniques to Analyze Aggregate Annotate Make searchable user generated multimedia content available on social networks