暂无图片
暂无图片
暂无图片
暂无图片
暂无图片
自动图像标注技术综述- 马艳春, 刘永坚, 解庆, 熊盛武, 唐伶俐.pdf
509
27页
4次
2021-12-07
免费下载
DOI
:
issn
JournalofCom
p
uterResearchandDevelo
p
ment
(
):
,
 
稿
:
;
:
 
:
(
);
(
WUT
:
YB
)
Thisworkwassu
pp
ortedb
y
theNationalNaturalScienceFoundationofChina
(
)
andtheFundamentalResearchFunds
fortheCentralUniversities
(
WUT
:
YB
)
 
:
(
felixx
q
@whuteducn
)
 
 
 
 
 
(
 
 
)
(
ma
y
anchun@whuteducn
)
ReviewofAutomaticIma
g
eAnnotationTechnolo
gy
MaYanchun
,
LiuYon
gj
ian
,
XieQin
g
,
Xion
g
Shen
g
wu
,
andTan
g
Lin
g
li
(
Schoolo
f
Com
p
uterScienceandTechnolo
gy
,
WuhanUniversit
y
o
f
Technolo
gy
,
Wuhan
)
Abstract Asoneofthemosteffectivewa
y
storeducethe
semantic
g
a
p
betweenima
g
edataandits
content
,
automaticima
g
eannotation
(
AIA
)
technolo
gy
hasshownits
g
reatsi
g
nificancetohel
pp
eo
p
le
understandima
g
econtentsandretrievethetar
g
etinformationfrom the massiveima
g
edataThis
p
a
p
ersummarizesthe
g
eneralframeworkofAIA modelsb
y
investi
g
atin
g
theliteraturesaboutima
g
e
annotationinrecent
y
ears
,
andanal
y
zesthe
g
eneral
p
roblemstosolvein AIA
p
roblemsb
y
combinin
g
theframeworkwithvariouss
p
ecificworksInthis
p
a
p
er
,
themainmethodsusedinvarious
AIA modelsareclassifiedintot
yp
es
:
correlationmodel
,
hiddenMarkovmodel
,
to
p
icmodel
,
matrix
factorizationmodel
,
nei
g
hborbased model
,
SVMbased model
,
g
ra
p
hbased model
,
CCA
(
KCCA
)
modelanddee
p
learnin
g
modelForeacht
yp
eofima
g
eannotation model
,
this
p
a
p
er
p
rovidesa
detailedstud
y
andanal
y
sisintermsof
basic
p
rinci
p
leintroductions
p
ecificmodeldifferencesmodel
summar
y
Inaddition
,
this
p
a
p
ersummarizessomecommonl
y
useddatasetsandevaluationindexes
,
andcom
p
aresthe
p
erformanceofsomeim
p
ortantima
g
eannotationmodelswithrelatedanal
y
sisonthe
advanta
g
esanddisadvanta
g
esofvarioust
yp
esof AIA modelsFinall
y
,
someo
p
en
p
roblemsand
researchdirectionsinthefieldofima
g
eannotationare
p
ro
p
osedandsu
gg
ested
Ke
y
words automaticima
g
eannotation
;
ima
g
esemanticanal
y
sis
;
ima
g
eta
gg
in
g
;
ima
g
econtentlabel
;
ima
g
econtentannotation
 
 
鸿
,
,
,
;
,
;
,
Markov
,
,
,
,
,
,
 
;
;
;
;
 TP
  
,
(
ChinaInternetNetworkInformation
Center
,
CNNIC
)
«
»
,
,
,
,
,
,
鸿
,
鸿
,
,
,
,
使
,
,
,
,
,
,
,
广
,
,
鸿
,
:
)
广
;
)
;
)
,
使
;
)
,
,
,
,
[
]
,
,
,
,
:
)
,
;
,
)
使
;
,
,
,
)
,
)
)
 
,
(
)
,
(
),
,
,
,
,
(
)
(
)
(
IW
)
,
;
,
(
II
)
(
WW
)
,
使
(
)
(
SIFT
(
scaleinvariantfeaturetrans
form
)
);
one
hot
wordvec
,
使
,
:
of 27
免费下载
【版权声明】本文为墨天轮用户原创内容,转载时必须标注文档的来源(墨天轮),文档链接,文档作者等基本信息,否则作者和墨天轮有权追究责任。如果您发现墨天轮中有涉嫌抄袭或者侵权的内容,欢迎发送邮件至:contact@modb.pro进行举报,并提供相关证据,一经查实,墨天轮将立刻删除相关内容。

评论

关注
最新上传
暂无内容,敬请期待...
下载排行榜
Top250 周榜 月榜