國立虎尾科技大學 |

Face Representation Learning and Its Applications on Social Media.

紀錄類型:	書目-語言資料,印刷品 : Monograph/item
正題名/作者:	Face Representation Learning and Its Applications on Social Media./
作者:	Wang, Shuyang.
出版者:	Ann Arbor : ProQuest Dissertations & Theses, : 2018,
面頁冊數:	125 p.
附註:	Source: Dissertation Abstracts International, Volume: 79-09(E), Section: B.
Contained By:	Dissertation Abstracts International79-09B(E).
標題:	Computer engineering. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=10751578
ISBN:	9780355944792

Face Representation Learning and Its Applications on Social Media.
Wang, Shuyang.

Face Representation Learning and Its Applications on Social Media. - Ann Arbor : ProQuest Dissertations & Theses, 2018 - 125 p.

Source: Dissertation Abstracts International, Volume: 79-09(E), Section: B.

Thesis (Ph.D.)--Northeastern University, 2018.

Learning and extracting good feature representations for face images is always a hot topic in machine learning field, especially in this era of social media where tons of face datas are available and lots of novel real-world face related applications are waiting to be solved. The basic question is how to learn discriminant features or representations for varieties of real-world applications. In this dissertation, we focus and study on the face data in two lines: (1) developing efficient and effective machine learning tools by considering the locality and intrinsic structure of the data, (2) applying existing or developed machine learning tools to novel face related problems, e.g., beauty assessment, makeup analysis, and kinship verification.

ISBN: 9780355944792Subjects--Topical Terms:

569006
Computer engineering.

Face Representation Learning and Its Applications on Social Media.
LDR:02810nam a2200301 4500 001 931645
005 20190716101634.5
008 190815s2018 ||||||||||||||||| ||eng d
020 $a 9780355944792
035 $a (MiAaPQ)AAI10751578
035 $a (MiAaPQ)coe.neu:10961
035 $a AAI10751578
040 $a MiAaPQ $c MiAaPQ
100 1 $a Wang, Shuyang. $3 1213839
245 1 0 $a Face Representation Learning and Its Applications on Social Media.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2018
300 $a 125 p.
500 $a Source: Dissertation Abstracts International, Volume: 79-09(E), Section: B.
500 $a Adviser: Yun Fu.
502 $a Thesis (Ph.D.)--Northeastern University, 2018.
520 $a Learning and extracting good feature representations for face images is always a hot topic in machine learning field, especially in this era of social media where tons of face datas are available and lots of novel real-world face related applications are waiting to be solved. The basic question is how to learn discriminant features or representations for varieties of real-world applications. In this dissertation, we focus and study on the face data in two lines: (1) developing efficient and effective machine learning tools by considering the locality and intrinsic structure of the data, (2) applying existing or developed machine learning tools to novel face related problems, e.g., beauty assessment, makeup analysis, and kinship verification.
520 $a Three types of data representations are studied in this dissertation, including feature selection, auto-encoder, and dictionary learning. First, two novel auto-encoder based schemes are proposed. To capture the task-relevant ones from the task-irrelevant ones, we propose a Generalized Discriminative Encoder, which is a unified generative model that integrates feature selection and auto-encoder together. In cross-domain learning, there is a more challenging problem that the domain divergence involves more than one dominant factors, e.g., different viewpoints, various resolutions and changing illuminations. To this end, we propose a Coupled Marginalized Denoising Autoencoders framework to address the cross-domain problem. Moreover, motivated by existing sparse coding and low-rank representation techniques, we explored the enhancement of classification by adding locality constraint on discriminative low-rank dictionary learning. Above proposed dictionary learning and auto-encoder based approaches are appropriate for some novel face analytics problems such as beauty assessment, kinship verification, and makeup analysis.
590 $a School code: 0160.
650 4 $a Computer engineering. $3 569006
690 $a 0464
710 2 $a Northeastern University. $b Electrical and Computer Engineering. $3 1182193
773 0 $t Dissertation Abstracts International $g 79-09B(E).
790 $a 0160
791 $a Ph.D.
792 $a 2018
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=10751578