BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Semantic Image Segmentation and Web-Supervised Visual Learning - F
 lorian Schroff\, University of Oxford
DTSTART:20090608T103000Z
DTEND:20090608T110000Z
UID:TALK17613@talks.cam.ac.uk
CONTACT:Dr Fabien Petitcolas
DESCRIPTION:*Abstract*: In object recognition\, the goal is to recognise o
 bjects of certain categories\, usually known and trained in advance\, desp
 ite intra-class appearance variations and small inter-class differences. T
 he appearance of objects is influenced by lighting\, scale\, different pos
 es\, viewpoints\, articulation of objects\, clutter and occlusion. Two dif
 ferent aspects of object recognition are investigated in this thesis. The 
 first part develops models for semantic object segmentation of natural ima
 ges and relies on groundtruth labelling for training. The second part uses
  the implicit supervision that is available on the Internet to learn visua
 l object-class models automatically. It can then provide a groundtruth lab
 elling for object detection or segmentation algorithms.\n\nThe goal in the
  first part is to label connected regions in an image as belonging to spec
 ific object classes\, such as grass or cow. We introduce a compact model t
 o the bag of visual words approach\, where each class is modelled by one s
 ingle histogram of visual words\, this is in contrast to common nearest-ne
 ighbour approaches which model each class by many histograms. After introd
 ucing segmentation algorithms based on these histogram models we extend th
 e Random Forest classifier and evaluate its feature selection properties a
 s well as the suitability of certain low-level features for the semantic o
 bject segmentation task. \n\nMost object recognition methods rely on label
 led training images. For each object category to be recognised\, the syste
 m is trained on a set of images containing instances of these categories. 
 The last part of this thesis focuses on the automatic creation of sets of 
 images that contain a certain object class. The idea is to download an ini
 tial set of images from the Internet based on a search query ( penguin). G
 iven the images a text based ranking that exploits the information on the 
 web-pages is performed. This ranking is then used to automatically learn v
 isual models for 18 object categories. We compare the performance of our s
 ystem to previous work and show that it performs equally well without the 
 need of explicit manual supervision.\n\n*Biography*: Florian Schroff is cu
 rrently a fourth year DPhil student in the Departement of Engineering Scie
 nce at the University of Oxford funded by Microsoft Research through the E
 uropean PhD Scholarship Programme. He is jointly supervised by Professor A
 ndrew Zisserman and Antonio Criminisi at Microsoft Research Cambridge. Bef
 ore joining the Visual Geometry Group (VGG) in Oxford he was working as a 
 researcher at the German Research Center for Artificial Intelligence in Ka
 iserslautern. He received his degree (Diploma) in computer science at the 
 University of Karlsruhe end of 2004\, where he was working with Professor 
 H.-H. Nagel on camera calibration and focused on artificial intelligence\,
  cryptography and algebra. In 2003 he received the Master of Science in co
 mputer science from the University of Massachusetts - Amherst\, where he h
 ad started his studies under the Baden-Württemberg exchange scholarship i
 n 2002.
LOCATION:Small Lecture Room\, Microsoft Research\, Roger Needham Building\
 , 7 J J Thomson Avenue\, Cambridge CB3 0FB
END:VEVENT
END:VCALENDAR