Cover Page
The handle http://hdl.handle.net/1887/66480 holds various files of this Leiden University dissertation.
Author: Liu, Y.
Title: Exploring images with deep learning for classification, retrieval and synthesis Issue Date: 2018-10-24
INVITATION
You are cordially invited to attend the defence ofmy PhD dissertation:
Expl ori ng Images wi t h Deep Learni ng f or
Cl assi f i cati on, Retri eval and Synthesi s
Wednesday 24th October2018 At11:15 AM
In the Academiegebouw Rapenburg 73,Leiden
Afterwards,a lunch reception will be held in Yuniku Restaurant, Stationsweg 19,2312 AS Leiden
Yu Liu
y.liu@liacs.leidenuniv.nl
Paranymphs:
WeiChen Theodoros Georgiou
Expl ori ng Images wi t h Deep Learni ng
Cl assi f i cat i on, Ret ri eval & Synt hesi s
Ex plo rin g I m ag es w ith D ee p L ea rn ing fo r C las sif ica tio n, R etr iev al a nd S yn the sis Yu L iu
Yu Li u
刘宇“Intelligence is the ability to adapt to change”.
-- Stephen Hawking
In 2018,the numberofmobile phone users willreach about4.9 billion.Assuming an average of5 photos taken perday using the built-in cameras would resultin about9 trillion photos annually.Thus,itbecomes challenging to mine semantic information from such a huge amountofvisualdata.To solve this challenge, deep learning, an importantsub-field in machine learning,has achieved impressive developments in recent years.Inspired by its success,this thesis aims to develop new approaches in deep learning to explore and analyze image data from three research themes:classification,retrievaland synthesis.
(1)Classification aims to correctly predictsemantic labels in an image,and acts as a fundamentaltask in (1)Classification aims to correctly predictsemantic labels in an image,and acts as a fundamentaltask in computervision.Typically,itcan be divided to image-leveland pixel-levelclassification.
(2)Retrievalis to search forimage candidates from the database thatare similarto the query sample.This work studies notonly image retrieval,butalso cross-modalretrievalbetween images and texts.
(3)Synthesis is able to generate new image samples thatneverexisted in the image database.We mainly focus on two synthesis applications:image-to-image translation and fashion style transfer.