Output details
11 - Computer Science and Informatics
Queen Mary University of London
Identifying Rare and Subtle Behaviors: A Weakly Supervised Joint Topic Model
<23>Provides the first framework for identifying behaviours from sparse and subtle (few pixels) examples with weak supervision. These key missing capabilities prevent existing systems' deployment for security/safety according to user partners (gary.butcher@met.pnn.police.uk) in the FP7 project (FP7-217899) for which this is an output. It is an extension of highly cited earlier work (ICCV'09). It resulted in: invitation as an expert panel member to FP7 HIDE project (irma.vanderploeg@zuyd.nl) on surveillance ethics, invited seminars at Oxford (eric@robots.ox.ac.uk) and Microsoft Research (pkohli@microsoft.com). It also won development contracts from US DOD through QMUL spin-out VSL (rpkoger@visionsemantics.com) and was part of a NewScientist feature http://bit.ly/63AtFW.