BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Multi-class image segmentation in EM images
DTSTART:20180522T110000
DTEND:20180522T130000
DTSTAMP:20260406T194625Z
UID:0cb41b86b0675ca5f04e65090346a1b03d5ce2b8627dbf1856d73949
CATEGORIES:Conferences - Seminars
DESCRIPTION:Pamuditha Udaranga Wickramasinghe\nEDIC candidacy exam\nExam p
 resident: Prof. Wulfram Gerstner\nThesis advisor: Prof. Pascal Fua\nCo-exa
 miner: Dr. Graham Knott\n\nAbstract\nWe study the problem of multi-class i
 mage segmentation in Electron Microscopy(EM) images with the focus on extr
 acting intricate structures made up of multiple parts. The problem is form
 ulated as a structured prediction task and we review two approaches to sol
 ve it\, namely (1) Probabilistic Graphical Models and (2) Deformable Model
 s. Furthermore\, we propose a framework for learning structural relationsh
 ips from data. We present the results based on this preliminary work and d
 iscuss future directions for our research.\n\nBackground papers\nDeformabl
 e models in medical image analysis: a survey\,  Medical Image Analysis 19
 96\, by  T. McInerney\, D. Terzopoulos\nEfficient Inference in Fully Conn
 ected CRFs with Gaussian Edge Potentials\, NIPS 2011\, by  P. Krähenbüh
 l\, V. Koltun.\nSegmentation-Aware Convolutional Networks Using Local Atte
 ntion Masks\, ICCV 2017\, by A. W. Harley\, K. G. Derpanis\, I. Kokkinos
LOCATION:BC 329 https://plan.epfl.ch/?room==BC%20329
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR