MEETING-DATE:November 22, 2012 MEETING-LOCATION:DC 1331 MEETING-TIME:11:30 MEETING-CHAIR:Elodie Fourquet MEETING-CHAIRPIC:elodie.jpg COFFEE-HOUR-LAST-WEEK:Zainab COFFEE-HOUR-THIS-WEEK:Volunteers? COFFEE-HOUR-NEXT-WEEK:Volunteers? FORTH-DATE1:November 29, 2012 FORTH-DATE2:December 6, 2012 FORTH-DATE3:December 13, 2012 FORTH-DATE4:December 20, 2012 FORTH-LOCATION1:DC 1331 11:30 FORTH-LOCATION2:DC 1331 11:30 FORTH-LOCATION3:DC 1331 11:30 FORTH-LOCATION4:DC 1331 11:30 FORTH-CHAIR1:Bill Cowan FORTH-CHAIR2:Tiffany Inglis FORTH-CHAIR3:Craig Kaplan FORTH-CHAIR4:Marta Kryven FORTH-CHAIRPIC1:cowan_unflipped.gif FORTH-CHAIRPIC2:tiffany.jpg FORTH-CHAIRPIC3:cr.jpg FORTH-CHAIRPIC4:martaPicture.jpg FORTH-TP1:Zainab AlMeraj FORTH-TP2:Bill Cowan FORTH-TP3:Elodie Fourquet FORTH-TP4:Tiffany Inglis FORTH-TPPIC1:zainab.jpg FORTH-TPPIC2:cowan_unflipped.gif FORTH-TPPIC3:elodie.jpg FORTH-TPPIC4:tiffany.jpg TPNAME:Grace Yao TPTITLE:An algorithm which represents objects in a background using discriminative non-orthogonal binary space(DNBS) TPABSTRACT:In visual tracking problem, how to represent the object is a crucial problem. This DNBS is a succinct and discriminative representation of the object spanned by Harr-like feature. However, the problem of finding the DNBS bases from an over-complete dictionary is NP-hard. The previous D-OOMP can be done in a greedy way and make it faster. Due to the huge search space, the algorithm is still expensive. At the same time they only care about the foreground with only one template which is not that stable. The DNBS is an efficient, generative and discriminative method. I will explain the original one, the iterative one and the hierarchical one in my tech presentation. TPPIC:grace-small.jpg DIONE: DITWO: DITHREE: DIFOUR: AIONE:Papier Mache painting AITWO: AITHREE: AIFOUR: LEONE: LETWO: LETHREE: LEFOUR: DMONE: DMTWO: DMTHREE: DMFOUR: SEMINARS: