1. 제 목 : Iris and ear segmentation, recognition, covariates analysis

2. 발 표 자 : Professor Peter peer

3. 일 시 :

Lecture#1) 2018. 06. 11. (15:00~16:30)
Lecture#2) 2018. 06. 12. (10:30~12:00)
Lecture#3) 2018. 06. 12. (14:00~15:30)

4. 장 소 : IT-1 313호

5. 초청교수 : 전자공학부 김민영

6. 강사약력 :
Associate Professor / University of Ljubljana, Faculty of Computer and Information Science, Slovenia – December 2014 ~ Present
Assistant Professor / University of Ljubljana, Faculty of Computer and Information Science, Slovenia – December 2006 ~ 2014
Invited Researcher / CEIT their spine-off Asiris Vision Technologies SA, San Sebastian, Spain – 2004 ~ 2005
Assistant / University of Primorska, Faculty of Education, Slovenia – November 2003 ~ September 2004

7. 내용요약 :

Lecture#1) Iris recognition pipeline
The presentation will start by introducing historical timeline, move to general description of the classical pipeline and then to details of each step in the pipeline. Thus, we will talk about acquisition, segmentation, masking, normalization, iris code, matching, performance evaluation, open research questions

Lecture#2) Why are the challenges at the conferences good?
The presentation will cover the outcomes of three challenges we participated (we organised one, we won at the other two) in 2017/18. Eventhough certain high-level details about the methods will be given, for instance deep learning approaches always won, the focus will be more on the lessons learned through participation.

Lecture#3) Ear segmentation, recognition, covariates analysis
First, novel ear detection technique based on convolutional encoder-decoder networks (CEDs) with shortcut connections will be presented. Then, deep-learning-based techniques for recognition will be discussed. Finally, a comprehensive analysis of several descriptor- and deep-learning-based techniques for ear recognition will be given.