I
nd
o
ne
s
ia
n J
o
urna
l o
f
E
lect
rica
l En
g
ineering
a
nd
Co
m
pu
t
er
Science
Vo
l.
41
,
No
.
3
,
Ma
r
ch
20
2
6
,
p
p
.
9
7
7
~
9
9
9
I
SS
N:
2
502
-
4
7
52
,
DOI
: 1
0
.
1
1
5
9
1
/ijee
cs
.v
41
.
i
3
.
pp
977
-
9
9
9
977
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ee
cs
.
ia
esco
r
e.
co
m
Lev
erag
ing
CNN
to a
na
ly
ze f
a
cia
l e
x
press
io
ns for
aca
demic
eng
a
g
ement moni
toring
wit
h
insig
h
ts f
ro
m
the
m
ulti
-
so
urce
a
ca
demic a
ff
ec
tive eng
a
g
ement dat
a
set
No
o
ra
C.
T
.
,
P
.
T
a
m
il Selv
a
n
D
e
p
a
r
t
me
n
t
o
f
C
o
mp
u
t
e
r
S
c
i
e
n
c
e
,
K
a
r
p
a
g
a
m
A
c
a
d
e
m
y
o
f
H
i
g
h
e
r
Ed
u
c
a
t
i
o
n
,
C
o
i
m
b
a
t
o
r
e
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
No
v
1
9
,
2
0
2
3
R
ev
is
ed
No
v
3
,
2
0
2
5
Acc
ep
ted
Dec
1
3
,
2
0
2
5
Th
e
d
y
n
a
m
ics
o
f
stu
d
e
n
t
e
n
g
a
g
e
m
e
n
t
a
n
d
e
m
o
ti
o
n
a
l
sta
tes
sig
n
ifi
c
a
n
tl
y
in
flu
e
n
c
e
lea
rn
in
g
o
u
tco
m
e
s.
P
o
siti
v
e
e
m
o
ti
o
n
s,
ste
m
m
in
g
fr
o
m
su
c
c
e
ss
fu
l
tas
k
c
o
m
p
letio
n
,
c
o
n
tras
t
with
n
e
g
a
ti
v
e
e
m
o
ti
o
n
s
a
risin
g
fro
m
lea
rn
in
g
stru
g
g
les
o
r
fa
il
u
re
s.
E
ffe
c
ti
v
e
tran
siti
o
n
s
to
e
n
g
a
g
e
m
e
n
t
o
c
c
u
r
u
p
o
n
p
ro
b
lem
re
so
lu
ti
o
n
,
wh
i
le
u
n
r
e
so
lv
e
d
iss
u
e
s
lea
d
to
fru
stra
ti
o
n
a
n
d
su
b
se
q
u
e
n
t
b
o
re
d
o
m
.
F
a
c
ial
e
n
g
a
g
e
m
e
n
t
m
o
n
it
o
ri
n
g
is
c
r
u
c
ial
fo
r
a
ss
e
ss
in
g
stu
d
e
n
ts
’
a
tt
e
n
ti
o
n
,
in
tere
st,
a
n
d
e
m
o
ti
o
n
a
l
re
sp
o
n
se
s
d
u
rin
g
lea
rn
i
n
g
.
Re
c
e
n
t
a
d
v
a
n
c
e
m
e
n
ts
in
c
o
n
v
o
lu
ti
o
n
a
l
n
e
u
ra
l
n
e
two
r
k
s
(CNN
s)
sh
o
w
p
ro
m
ise
in
a
u
to
m
a
ti
c
a
ll
y
a
n
a
ly
z
in
g
fa
c
ial
e
x
p
re
ss
io
n
s
to
in
fe
r
e
n
g
a
g
e
m
e
n
t
le
v
e
ls.
T
h
is
stu
d
y
p
r
o
p
o
se
s
a
CNN
-
b
a
se
d
a
p
p
ro
a
c
h
u
ti
li
z
i
n
g
th
e
m
u
lt
i
-
so
u
rc
e
a
c
a
d
e
m
ic
a
ffe
c
ti
v
e
e
n
g
a
g
e
m
e
n
t
d
a
tas
e
t
(M
AA
ED)
to
c
a
teg
o
rize
fa
c
ial
e
x
p
re
ss
io
n
s
in
t
o
b
o
re
d
o
m
,
c
o
n
fu
si
o
n
,
fru
stra
ti
o
n
,
a
n
d
y
a
w
n
in
g
.
B
y
e
x
trac
ti
n
g
fe
a
t
u
re
s
fro
m
fa
c
ial
ima
g
e
s,
th
is
m
e
th
o
d
o
ffe
rs
a
n
e
fficie
n
t
a
n
d
o
b
jec
ti
v
e
m
e
a
n
s
to
g
a
u
g
e
stu
d
e
n
t
e
n
g
a
g
e
m
e
n
t.
Re
c
o
g
n
izin
g
a
n
d
a
d
d
re
ss
in
g
n
e
g
a
t
iv
e
a
ffe
c
ti
v
e
sta
tes
,
su
c
h
a
s
c
o
n
fu
si
o
n
a
n
d
b
o
re
d
o
m
,
is
fu
n
d
a
m
e
n
tal
in
c
re
a
ti
n
g
su
p
p
o
rti
v
e
lea
rn
in
g
e
n
v
iro
n
m
e
n
ts.
Th
r
o
u
g
h
a
u
t
o
m
a
ted
fra
m
e
e
x
trac
ti
o
n
a
n
d
m
o
d
e
l
c
o
m
p
a
riso
n
,
t
h
is
stu
d
y
d
e
m
o
n
st
ra
tes
re
d
u
c
e
d
lo
ss
v
a
lu
e
s
with
imp
ro
v
i
n
g
a
c
c
u
ra
c
y
,
sh
o
wc
a
sin
g
t
h
e
e
ffe
c
ti
v
e
n
e
ss
o
f
th
is
m
e
th
o
d
in
o
b
jec
ti
v
e
l
y
e
v
a
lu
a
ti
n
g
stu
d
e
n
t
e
n
g
a
g
e
m
e
n
t.
F
a
c
ial
e
n
g
a
g
e
m
e
n
t
m
o
n
i
to
ri
n
g
w
it
h
CNN
s,
u
sin
g
t
h
e
M
AA
ED
d
a
tas
e
t,
is
p
i
v
o
tal
in
u
n
d
e
rsta
n
d
in
g
h
u
m
a
n
b
e
h
a
v
io
r
a
n
d
e
n
h
a
n
c
in
g
e
d
u
c
a
ti
o
n
a
l
e
x
p
e
rien
c
e
s.
Th
e
CNN
m
o
d
e
l
,
trai
n
e
d
o
n
M
AA
ED
a
n
n
o
tate
d
fa
c
ial
e
x
p
re
ss
io
n
s,
a
c
c
u
ra
tely
c
las
sifies
e
n
g
a
g
e
m
e
n
t
c
a
teg
o
ries
.
Ex
p
e
rime
n
tal
re
su
l
ts
u
n
d
e
rsc
o
re
th
e
CNN
-
ba
se
d
a
p
p
ro
a
c
h
’
s
e
f
fica
c
y
in
m
o
n
it
o
r
in
g
fa
c
ial
e
n
g
a
g
e
m
e
n
t,
h
ig
h
li
g
h
ti
n
g
i
ts
p
o
ten
ti
a
l
t
o
e
n
rich
e
d
u
c
a
ti
o
n
a
l
e
n
v
ir
o
n
m
e
n
ts
a
n
d
p
e
r
so
n
a
li
z
e
d
lea
rn
i
n
g
e
x
p
e
rien
c
e
s.
K
ey
w
o
r
d
s
:
Aca
d
em
ic
af
f
ec
tiv
e
en
g
ag
em
e
n
t
E
m
o
tio
n
r
ec
o
g
n
itio
n
Facial
ex
p
r
ess
io
n
s
Mu
lti
-
s
o
u
r
ce
ac
ad
em
ic
af
f
ec
tiv
e
en
g
a
g
em
en
t
Stu
d
en
t e
n
g
a
g
em
en
t
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
No
o
r
a
C
.
T
.
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
,
Kar
p
ag
a
m
Aca
d
em
y
o
f
H
ig
h
er
E
d
u
ca
tio
n
C
o
im
b
ato
r
e,
I
n
d
ia
E
m
ail:
n
o
o
r
ac
t@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
s
tu
d
en
t
in
a
class
r
o
o
m
m
ay
ex
p
er
ie
n
ce
d
if
f
e
r
en
t
m
i
x
tu
r
es
o
f
m
en
tal
s
tates,
wh
ich
ar
e
v
er
y
im
p
o
r
tan
t
f
ac
to
r
s
to
r
ev
ea
l
t
h
e
co
g
n
itiv
e
lear
n
in
g
an
d
en
g
ag
em
en
t
o
f
s
tu
d
en
ts
.
W
h
en
t
h
e
s
tu
d
en
ts
ten
d
to
m
ak
e
m
is
tak
es o
r
f
ac
e
f
ailu
r
es o
r
s
tr
u
g
g
le
at
lear
n
in
g
m
a
y
ar
o
u
s
e
a
n
eg
ativ
e
em
o
tio
n
al
s
tate
s
u
ch
as ir
r
itatio
n
,
f
r
u
s
tr
atio
n
,
an
d
an
g
er
o
n
th
e
o
th
er
h
an
d
,
if
th
ey
ca
n
co
m
p
lete
an
y
tas
k
o
r
co
n
q
u
er
ch
al
len
g
es/d
if
f
icu
lties
,
p
o
s
itiv
e
m
en
tal
s
tates
s
u
ch
as
d
elig
h
t,
ex
citem
en
t,
an
d
s
atis
f
ac
tio
n
will
b
e
th
e
r
esu
lt
[
1
]
.
Acc
o
r
d
i
n
g
to
D
’
Me
llo
an
d
Gr
ae
s
s
er
[
2
]
,
in
n
o
r
m
al
ca
s
es,
th
e
s
tu
d
en
t
en
ter
s
in
to
th
e
lear
n
in
g
ac
tiv
ity
in
an
en
g
a
g
ed
an
d
co
n
ce
n
tr
ated
s
tate
it
will
co
n
tin
u
e
u
n
til
th
e
y
r
ea
c
h
o
u
t
in
a
n
y
k
in
d
o
f
d
if
f
icu
lt
s
itu
atio
n
g
r
ad
u
ally
lead
in
g
to
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
41
,
No
.
3
,
Ma
r
ch
20
2
6
:
9
7
7
-
999
978
co
n
f
u
s
io
n
an
d
b
o
r
ed
o
m
.
At
th
is
p
o
in
t,
eith
e
r
o
n
e
o
f
t
h
e
tr
a
n
s
itio
n
s
will
o
cc
u
r
.
Po
s
itiv
e
wa
y
s
o
f
tr
an
s
itio
n
will
o
cc
u
r
w
h
en
th
e
s
tu
d
e
n
t
r
et
u
r
n
s
to
t
h
e
en
g
a
g
ed
s
tate
b
y
r
eso
l
v
in
g
t
h
e
p
r
o
b
lem
s
h
e/sh
e
f
ac
e
d
.
Neg
ativ
e
way
s
o
f
tr
an
s
itio
n
o
cc
u
r
wh
en
th
e
p
r
o
b
lem
h
e/sh
e
f
ac
ed
at
th
e
tim
e
o
f
lis
ten
in
g
/d
is
cu
s
s
io
n
m
ay
n
o
t
b
e
s
o
lv
ed
.
Gr
ad
u
ally
t
h
e
s
tu
d
e
n
t
m
a
y
s
tick
in
s
u
ch
a
s
itu
atio
n
an
d
tr
an
s
i
tio
n
to
f
r
u
s
tr
atio
n
.
if
th
is
f
r
u
s
tr
atio
n
p
e
r
s
is
ted
f
o
r
s
o
m
e
tim
e
f
u
r
th
er
lead
s
to
b
o
r
ed
o
m
.
A
g
o
o
d
teac
h
er
s
h
o
u
ld
b
e
ab
le
to
m
o
n
ito
r
th
e
c
h
an
g
es
in
th
e
m
en
tal
s
tates
o
f
th
e
s
tu
d
en
ts
d
u
r
i
n
g
lectu
r
in
g
,
th
u
s
s
h
e
ca
n
g
iv
e
p
er
s
o
n
alize
d
ass
is
tan
ce
to
th
e
s
tu
d
en
ts
wh
o
f
elt
co
n
f
u
s
io
n
,
f
r
u
s
tr
atio
n
,
o
r
an
y
o
th
er
n
eg
ativ
e
e
m
o
tio
n
s
.
B
y
id
en
tify
in
g
th
e
p
r
o
b
lem
s
f
ac
ed
d
u
r
in
g
th
e
d
is
cu
s
s
io
n
,
th
e
teac
h
er
ca
n
a
b
le
to
g
iv
e
f
u
r
th
er
e
x
p
lan
atio
n
/ch
an
g
e
th
e
way
o
f
teac
h
in
g
i
n
s
u
ch
a
way
th
at
th
ey
ca
n
u
n
d
er
s
tan
d
ea
s
ily
,
t
h
er
eb
y
m
a
x
im
izin
g
th
e
s
tu
d
e
n
ts
lear
n
in
g
o
u
tco
m
e.
Ma
s
s
iv
e
o
p
en
o
n
lin
e
c
o
u
r
s
es
(
MO
OC
s
)
h
av
e
b
r
o
u
g
h
t
a
r
e
v
o
lu
tio
n
i
n
h
ig
h
er
e
d
u
ca
tio
n
b
y
allo
win
g
in
ter
ested
s
tu
d
e
n
ts
to
p
u
r
s
u
e
th
eir
ed
u
ca
tio
n
at
th
eir
o
wn
p
ac
e
a
n
d
co
n
v
en
ien
ce
.
T
h
e
co
n
ten
t
d
eliv
er
y
s
u
cc
ess
f
u
l
o
n
ly
w
h
en
it
is
m
o
d
u
lated
b
y
r
ea
l
-
tim
e
s
tu
d
en
t
f
ee
d
b
ac
k
.
T
h
at
k
ey
f
ac
to
r
is
m
is
s
in
g
in
e
-
lear
n
in
g
en
v
ir
o
n
m
en
ts
.
T
h
e
au
to
m
ated
en
g
ag
em
e
n
t m
o
n
ito
r
in
g
m
eth
o
d
s
ca
n
ea
s
ily
b
e
em
p
l
o
y
ed
i
n
s
u
ch
e
-
lear
n
in
g
p
latf
o
r
m
s
[
3
]
.
Af
f
ec
tiv
e
co
m
p
u
tin
g
in
ed
u
ca
t
io
n
is
a
g
r
o
win
g
to
p
ic
th
at
is
in
cr
ea
s
in
g
in
p
o
p
u
lar
ity
as
tim
e
g
o
es
o
n
.
R
esear
ch
er
s
in
th
is
d
o
m
ain
e
m
p
lo
y
v
ar
io
u
s
m
eth
o
d
o
lo
g
ies
an
d
tech
n
i
q
u
es
to
ca
p
tu
r
e
a
n
d
in
ter
p
r
et
em
o
tio
n
s
in
ed
u
ca
tio
n
al
s
ettin
g
s
[
2
]
.
C
o
m
m
o
n
ly
u
s
ed
m
ac
h
in
e
le
ar
n
in
g
m
o
d
els
i
n
clu
d
e
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
i
n
e
s
(
SVM)
[
4
]
,
c
o
n
v
o
lu
ti
o
n
al
n
eu
r
al
n
etwo
r
k
s
(
C
NN)
[
5
]
,
an
d
o
th
er
d
ee
p
lea
r
n
in
g
a
lg
o
r
ith
m
s
[
6
]
,
[
7
]
.
Mu
ltimo
d
al
ap
p
r
o
ac
h
es
c
o
m
b
in
e
d
if
f
er
e
n
t
d
ata
ty
p
es,
s
u
ch
as
f
ac
ial
ex
p
r
ess
io
n
s
,
p
h
y
s
io
lo
g
ical
s
ig
n
als
[
8
]
,
[
9
]
,
an
d
tex
t
-
b
ased
in
ter
ac
tio
n
s
[
1
0
]
,
t
o
ac
h
ie
v
e
m
o
r
e
ac
cu
r
ate
em
o
tio
n
d
etec
tio
n
an
d
an
aly
s
is
.
I
n
s
ig
h
ts
g
ain
ed
f
r
o
m
em
o
tio
n
r
ec
o
g
n
iti
o
n
o
f
f
e
r
v
alu
ab
le
u
n
d
er
s
tan
d
i
n
g
o
f
s
tu
d
en
t
b
e
h
av
io
u
r
an
d
le
ar
n
in
g
ex
p
er
ien
ce
s
ac
r
o
s
s
v
ar
io
u
s
ed
u
ca
tio
n
al
s
ettin
g
s
,
in
clu
d
in
g
e
-
lear
n
in
g
,
o
f
f
lin
e
class
r
o
o
m
s
,
v
ir
tu
al
class
r
o
o
m
s
,
an
d
co
m
p
u
ter
-
en
ab
led
class
r
o
o
m
s
.
E
m
o
tio
n
r
ec
o
g
n
itio
n
tec
h
n
iq
u
es
h
o
ld
t
h
e
p
o
ten
tial
to
tailo
r
in
s
tr
u
ctio
n
al
s
tr
ateg
ies,
o
f
f
er
p
er
s
o
n
alize
d
f
ee
d
b
ac
k
,
an
d
cr
ea
te
m
o
r
e
en
g
ag
in
g
ed
u
ca
tio
n
al
en
v
ir
o
n
m
en
ts
.
I
n
teg
r
atin
g
af
f
ec
tiv
e
co
m
p
u
tin
g
tech
n
iq
u
e
s
,
s
u
ch
as
em
o
tio
n
r
ec
o
g
n
itio
n
an
d
s
en
tim
en
t
an
aly
s
is
,
in
to
in
tellig
en
t
tu
to
r
in
g
s
y
s
tem
s
en
ab
le
ad
ap
tiv
e
in
s
t
r
u
ctio
n
b
ased
o
n
s
tu
d
en
ts
’
a
f
f
ec
tiv
e
s
tates,
th
er
eb
y
en
h
an
cin
g
en
g
a
g
em
en
t,
m
o
tiv
atio
n
,
an
d
o
v
er
all
lear
n
in
g
o
u
tco
m
es.
Ad
d
itio
n
ally
,
af
f
ec
tiv
e
co
m
p
u
tin
g
p
lay
s
a
cr
u
cial
r
o
le
in
d
esi
g
n
in
g
em
o
tio
n
ally
r
esp
o
n
s
iv
e
o
n
lin
e
lear
n
in
g
p
latf
o
r
m
s
,
wh
ich
co
n
s
id
er
s
tu
d
en
ts
’
em
o
tio
n
al
ex
p
e
r
ien
ce
s
to
cr
ea
te
m
o
r
e
ef
f
ec
tiv
e
lear
n
in
g
e
n
v
ir
o
n
m
en
ts
.
T
h
e
ex
is
tin
g
liter
atu
r
e
also
f
o
cu
s
es
o
n
s
p
ec
if
ic
asp
ec
ts
o
f
em
o
tio
n
s
,
s
u
ch
as
ac
ad
em
ic
e
m
o
tio
n
s
,
en
g
ag
em
e
n
t
lev
els,
d
is
tr
ac
tio
n
,
f
atig
u
e,
an
d
lear
n
i
n
g
-
ce
n
tr
ed
em
o
tio
n
s
.
Fo
r
in
s
tan
ce
,
San
eir
o
et
a
l.
[
1
1
]
co
n
d
u
cted
a
s
tu
d
y
o
n
f
ac
ia
l
em
o
tio
n
r
ec
o
g
n
itio
n
(
FER)
to
p
r
ed
ict
ac
a
d
em
ic
p
er
f
o
r
m
an
ce
,
h
ig
h
lig
h
tin
g
t
h
e
p
o
ten
tial
o
f
af
f
ec
tiv
e
co
m
p
u
tin
g
in
e
d
u
ca
tio
n
al
ass
e
s
s
m
en
t.
Sy
s
tem
atic
r
ev
iews,
lik
e
th
e
o
n
e
c
o
n
d
u
cted
b
y
Alam
ed
a
-
Pin
ed
a
et
a
l.
[
1
2
]
,
em
p
h
asize
th
e
b
en
ef
its
o
f
in
teg
r
atin
g
af
f
ec
tiv
e
co
m
p
u
tin
g
in
lear
n
in
g
a
n
al
y
tics
,
p
r
o
v
id
in
g
v
a
lu
ab
le
in
s
ig
h
ts
f
o
r
tailo
r
in
g
in
ter
v
en
tio
n
s
an
d
s
u
p
p
o
r
tin
g
s
tu
d
e
n
t
well
-
b
ein
g
.
Nei
et
a
l
.
[
1
3
]
f
o
cu
s
ed
o
n
an
aly
zi
n
g
s
tu
d
en
ts
’
em
o
tio
n
al
s
tates
in
o
n
lin
e
lear
n
in
g
en
v
ir
o
n
m
en
ts
u
s
in
g
tex
t
-
m
in
in
g
tec
h
n
iq
u
es.
B
y
an
aly
zin
g
s
tu
d
e
n
ts
’
wr
itte
n
in
ter
ac
tio
n
s
,
th
e
r
esear
ch
er
s
g
ain
ed
i
n
s
ig
h
ts
in
to
s
tu
d
en
ts
’
em
o
tio
n
al
r
esp
o
n
s
es,
co
n
tr
ib
u
tin
g
to
a
d
ee
p
er
u
n
d
er
s
tan
d
in
g
o
f
th
eir
em
o
tio
n
al
ex
p
e
r
ien
ce
s
in
o
n
li
n
e
lear
n
in
g
.
T
h
e
in
c
o
r
p
o
r
atio
n
o
f
af
f
ec
tiv
e
co
m
p
u
tin
g
a
n
d
em
o
tio
n
r
ec
o
g
n
itio
n
tech
n
iq
u
es
h
as
th
e
p
o
ten
tial
to
en
h
an
ce
ed
u
ca
tio
n
al
ex
p
er
i
en
ce
s
,
im
p
r
o
v
e
lear
n
in
g
o
u
tc
o
m
es,
an
d
p
r
o
v
id
e
p
er
s
o
n
alize
d
s
u
p
p
o
r
t
b
ased
o
n
s
tu
d
en
ts
’
em
o
tio
n
al
s
tates.
B
y
lev
er
ag
in
g
th
ese
tech
n
iq
u
es,
ed
u
ca
to
r
s
ca
n
g
ain
a
d
ee
p
er
u
n
d
er
s
tan
d
in
g
o
f
s
tu
d
en
ts
’
em
o
tio
n
al
p
atter
n
s
,
m
o
n
i
to
r
c
h
an
g
es
i
n
af
f
ec
tiv
e
s
tates,
an
d
r
esp
o
n
d
p
r
o
ac
tiv
ely
to
s
u
p
p
o
r
t th
eir
lea
r
n
in
g
n
ee
d
s
.
T
r
ad
itio
n
al
m
eth
o
d
s
f
o
r
en
g
a
g
em
en
t
m
o
n
ito
r
in
g
o
f
ten
r
ely
o
n
m
an
u
al
o
b
s
er
v
atio
n
,
wh
i
ch
is
b
o
th
s
u
b
jectiv
e
an
d
tim
e
-
c
o
n
s
u
m
i
n
g
.
H
o
wev
er
,
r
ec
en
t
b
r
ea
k
th
r
o
u
g
h
s
in
d
ee
p
lea
r
n
in
g
m
et
h
o
d
s
,
n
o
tab
ly
C
NNs
[
1
4
]
,
h
av
e
d
em
o
n
s
tr
ated
s
ig
n
if
ican
t
p
o
ten
tial
i
n
au
to
m
atica
lly
an
aly
zin
g
f
ac
ial
e
x
p
r
ess
io
n
s
an
d
in
f
er
r
in
g
en
g
ag
em
e
n
t
lev
els.
I
n
th
is
s
tu
d
y
,
we
p
r
o
p
o
s
e
a
C
NN
-
b
ased
ap
p
r
o
ac
h
f
o
r
f
ac
ial
en
g
ag
em
en
t
m
o
n
ito
r
in
g
,
u
tili
zin
g
a
n
o
v
el
d
ata
s
et
n
am
ed
m
u
lti
-
s
o
u
r
ce
ac
a
d
em
ic
af
f
ec
tiv
e
en
g
ag
em
e
n
t
d
ataset
(
MA
AE
D)
.
Ou
r
C
NN
m
o
d
el
is
tr
ain
e
d
o
n
MA
AE
D
to
class
if
y
f
ac
ial
ex
p
r
ess
io
n
s
in
to
d
if
f
e
r
en
t
e
n
g
ag
em
e
n
t
ca
teg
o
r
ies,
s
u
ch
as
b
o
r
ed
o
m
,
co
n
f
u
s
io
n
,
f
r
u
s
tr
ati
o
n
,
an
d
y
awn
in
g
.
T
h
e
m
o
d
el
ca
n
m
ak
e
ac
cu
r
ate
p
r
e
d
ictio
n
s
ab
o
u
t
s
tu
d
en
ts
’
en
g
ag
em
e
n
t
lev
els
b
y
au
to
m
atica
lly
ex
tr
ac
tin
g
d
is
cr
im
in
ativ
e
f
ea
tu
r
es
f
r
o
m
f
ac
ial
im
ag
es.
T
h
e
k
ey
co
n
tr
ib
u
tio
n
s
o
f
th
is
s
tu
d
y
lie
in
th
e
cr
ea
tio
n
o
f
MA
AE
D,
a
u
n
iq
u
e
a
n
d
r
ich
d
ataset
th
at
r
ef
lects
r
ea
l
-
wo
r
ld
ac
ad
em
ic
en
g
ag
em
en
t,
an
d
t
h
e
d
e
v
elo
p
m
e
n
t
o
f
an
ef
f
ici
en
t
C
NN
-
b
ased
ap
p
r
o
ac
h
f
o
r
f
ac
ial
en
g
ag
em
e
n
t
m
o
n
ito
r
in
g
.
B
y
h
ar
n
ess
in
g
th
e
p
o
wer
o
f
d
ee
p
lear
n
in
g
,
t
h
e
p
r
o
p
o
s
ed
a
p
p
r
o
ac
h
aim
s
to
p
r
o
v
id
e
ed
u
ca
to
r
s
with
an
o
b
jectiv
e
an
d
ef
f
icien
t
m
e
th
o
d
to
ass
ess
s
tu
d
en
ts
’
en
g
a
g
em
en
t
d
u
r
in
g
ac
ad
em
ic
ac
tiv
ities
.
T
h
is
h
as
th
e
p
o
ten
tial
to
e
n
h
an
ce
ed
u
ca
t
io
n
al
en
v
ir
o
n
m
e
n
ts
,
p
er
s
o
n
a
lize
lear
n
in
g
e
x
p
er
ien
ce
s
,
a
n
d
en
a
b
le
tim
ely
in
ter
v
en
tio
n
s
to
im
p
r
o
v
e
s
tu
d
e
n
t o
u
tco
m
es.
E
v
en
th
o
u
g
h
f
ac
ial
e
x
p
r
ess
io
n
an
aly
s
is
is
a
p
o
wer
f
u
l
m
eth
o
d
f
o
r
d
etec
tin
g
em
o
tio
n
al
r
esp
o
n
s
es
an
d
ass
es
s
in
g
s
tu
d
en
t
en
g
ag
em
en
t
,
it
is
b
en
ef
icial
to
in
teg
r
ate
m
u
ltip
le
m
o
d
alities
in
to
en
g
a
g
em
en
t
m
o
n
ito
r
in
g
s
y
s
tem
s
,
to
g
ain
a
m
o
r
e
h
o
lis
tic
u
n
d
er
s
tan
d
in
g
.
T
h
ese
ad
d
itio
n
al
m
o
d
alities
in
clu
d
e
ey
e
tr
ac
k
in
g
,
wh
ich
r
ev
ea
ls
v
alu
ab
le
in
f
o
r
m
atio
n
ab
o
u
t
s
tu
d
en
ts
’
atten
tio
n
,
f
o
c
u
s
,
an
d
in
f
o
r
m
atio
n
p
r
o
ce
s
s
in
g
d
u
r
in
g
lear
n
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
Leve
r
a
g
in
g
C
N
N
to
a
n
a
lyze fa
cia
l e
xp
r
ess
io
n
s
fo
r
a
ca
d
emic
en
g
a
g
eme
n
t
mo
n
ito
r
in
g
w
ith
…
(
N
o
o
r
a
C
.
T.
)
979
ac
tiv
ities
[
1
5
]
.
Sp
ee
ch
an
aly
s
is
o
f
f
er
s
in
s
ig
h
ts
in
to
s
tu
d
en
ts
’
lev
el
o
f
en
g
ag
em
en
t
an
d
e
m
o
tio
n
al
s
tates
b
y
ex
am
in
in
g
s
p
ee
ch
p
atter
n
s
,
to
n
e,
an
d
v
o
ca
l
cu
es
[
1
6
]
.
Natu
r
al
lan
g
u
a
g
e
p
r
o
ce
s
s
in
g
(
NL
P)
tech
n
iq
u
es
ap
p
lie
d
to
s
tu
d
en
ts
’
s
p
ee
ch
tr
an
s
cr
ip
ti
o
n
s
o
r
r
ec
o
r
d
i
n
g
s
ca
n
d
etec
t
s
en
tim
en
t,
en
g
a
g
em
en
t,
a
n
d
c
o
n
ten
t
u
n
d
er
s
tan
d
in
g
[
1
7
]
.
Ad
d
itio
n
ally
,
it
is
cr
u
cial
to
co
n
s
id
er
p
h
y
s
io
lo
g
ical
s
ig
n
als,
en
co
m
p
ass
in
g
h
ea
r
t
r
a
te
v
ar
iab
il
ity
,
s
k
in
co
n
d
u
cta
n
ce
,
r
esp
ir
atio
n
r
ate,
an
d
b
o
d
y
tem
p
er
atu
r
e.
Ap
ar
t
f
r
o
m
E
E
G,
wh
ich
ca
n
o
f
f
er
in
s
ig
h
ts
in
to
s
tu
d
en
ts
’
em
o
tio
n
al
s
tates,
s
tr
ess
s
lev
el
s
,
an
d
co
g
n
itiv
e
lo
a
d
[
1
8
]
,
[
1
9
]
.
An
aly
s
in
g
s
tu
d
e
n
ts
’
in
ter
ac
tio
n
s
with
d
ig
ital
lear
n
in
g
p
latf
o
r
m
s
,
in
clu
d
in
g
click
s
,
m
o
u
s
e
m
o
v
em
en
ts
,
an
d
to
u
ch
g
estu
r
es,
p
r
o
v
id
es
v
alu
ab
le
d
ata
o
n
en
g
ag
em
e
n
t
an
d
n
av
ig
ati
o
n
p
a
tter
n
s
[
2
0
]
,
[
2
1
]
.
A
d
d
itio
n
ally
,
p
r
o
x
im
ity
s
en
s
o
r
s
m
o
n
ito
r
in
g
s
tu
d
en
ts
’
p
h
y
s
ical
p
r
esen
ce
in
th
e
lear
n
in
g
e
n
v
ir
o
n
m
en
t
o
f
f
er
i
n
s
ig
h
ts
in
to
th
ei
r
lev
el
o
f
e
n
g
ag
e
m
en
t a
n
d
p
ar
t
icip
atio
n
.
Stu
d
en
t
en
g
ag
em
e
n
t
is
a
m
u
ltifa
ce
ted
co
n
ce
p
t,
e
n
co
m
p
ass
in
g
v
ar
io
u
s
d
im
en
s
io
n
s
s
u
ch
as
b
eh
av
io
r
al,
em
o
tio
n
al,
co
g
n
itiv
e,
s
o
cial,
cu
ltu
r
al,
an
d
co
n
tex
tu
al
en
g
a
g
em
en
t.
B
eh
av
i
o
u
r
al
en
g
ag
em
en
t
in
v
o
lv
es
o
b
s
er
v
ab
le
ac
tio
n
s
lik
e
p
ar
ti
cip
atio
n
an
d
co
m
p
letio
n
o
f
ass
ig
n
m
en
ts
,
wh
ile
em
o
tio
n
al
en
g
ag
em
e
n
t
f
o
cu
s
es
o
n
s
tu
d
en
ts
’
f
ee
lin
g
s
an
d
attitu
d
es
to
war
d
lear
n
in
g
.
C
o
g
n
itiv
e
en
g
ag
em
e
n
t
r
elate
s
to
m
en
tal
ef
f
o
r
ts
an
d
in
v
o
lv
em
en
t
in
cr
itical
th
in
k
in
g
a
n
d
p
r
o
b
le
m
-
s
o
lv
in
g
.
So
cial
en
g
ag
em
e
n
t
em
p
h
a
s
ize
s
in
ter
ac
tio
n
s
with
p
ee
r
s
an
d
teac
h
er
s
,
p
r
o
m
o
tin
g
co
llab
o
r
atio
n
an
d
co
m
m
u
n
icatio
n
.
C
u
ltu
r
al
an
d
co
n
tex
tu
al
en
g
ag
em
e
n
t
co
n
s
id
er
s
th
e
in
f
lu
en
ce
o
f
cu
lt
u
r
al
f
ac
to
r
s
an
d
th
e
lea
r
n
in
g
e
n
v
ir
o
n
m
en
t
o
n
s
tu
d
e
n
t
en
g
a
g
e
m
en
t,
em
p
h
asizin
g
in
clu
s
iv
ity
an
d
s
u
p
p
o
r
tiv
e
n
ess
.
E
n
g
ag
em
e
n
t
lev
els
ca
n
v
ar
y
f
r
o
m
lo
w
to
m
e
d
iu
m
to
h
i
g
h
,
with
ea
ch
lev
el
in
d
icatin
g
d
if
f
er
en
t
lev
els
o
f
in
ter
est,
m
o
tiv
atio
n
,
a
n
d
p
a
r
ticip
atio
n
.
R
ec
o
g
n
izin
g
n
eg
ativ
e
af
f
ec
tiv
e
s
tates
lik
e
y
awn
in
g
,
c
o
n
f
u
s
io
n
,
b
o
r
ed
o
m
,
an
d
f
r
u
s
tr
atio
n
is
cr
u
cial,
as
th
ey
ca
n
n
e
g
ativ
ely
im
p
ac
t
s
t
u
d
en
t
en
g
ag
em
en
t,
m
o
tiv
atio
n
,
an
d
well
-
b
ein
g
.
Ad
d
r
ess
in
g
th
ese
em
o
tio
n
s
is
ess
en
tial
f
o
r
cr
ea
tin
g
a
s
u
p
p
o
r
tiv
e
an
d
ef
f
ec
tiv
e
lear
n
in
g
e
n
v
ir
o
n
m
en
t.
B
y
u
n
d
er
s
tan
d
in
g
a
n
d
ca
ter
i
n
g
to
i
n
d
iv
id
u
al
s
tu
d
en
ts
’
n
ee
d
s
,
e
d
u
ca
to
r
s
ca
n
f
o
s
ter
a
p
o
s
itiv
e
class
r
o
o
m
atm
o
s
p
h
er
e
an
d
p
r
o
m
o
te
ac
ad
em
ic
s
u
cc
ess
,
h
elp
in
g
s
tu
d
en
ts
to
o
v
er
co
m
e
ch
allen
g
es
an
d
th
r
iv
e
in
t
h
eir
lear
n
in
g
e
x
p
er
ien
ce
s
.
Stra
teg
ies
s
u
ch
as
clar
if
icatio
n
,
a
d
d
itio
n
al
s
u
p
p
o
r
t,
r
elev
an
t
a
n
d
s
tim
u
latin
g
co
n
ten
t,
an
d
ef
f
e
ctiv
e
teac
h
in
g
ap
p
r
o
ac
h
es
p
la
y
a
v
ital
r
o
le
in
ad
d
r
ess
in
g
t
h
ese
em
o
tio
n
s
an
d
en
h
an
cin
g
s
tu
d
en
t e
n
g
ag
em
e
n
t.
Dee
p
lear
n
in
g
tech
n
iq
u
es,
s
u
c
h
as
C
NN,
h
av
e
g
ain
ed
g
r
ea
ter
p
o
p
u
lar
ity
in
r
ec
e
n
t
s
tu
d
ies
b
ec
au
s
e
o
f
its
o
u
ts
tan
d
in
g
r
esu
lts
.
Ash
win
an
d
Gu
d
d
eti
[
2
2
]
estab
lis
h
ed
a
s
tr
at
eg
y
b
ased
o
n
C
NN
f
o
r
h
id
d
e
n
e
n
g
ag
em
e
n
t
an
aly
s
is
u
s
in
g
n
o
n
-
v
e
r
b
al
c
u
e
s
.
Su
m
e
r
et
a
l
.
[
4
]
u
s
e
Atten
ti
o
n
-
Net
f
o
r
h
ea
d
p
o
s
e
esti
m
atio
n
an
d
Af
f
ec
t
-
Net
f
o
r
f
ac
ial
ex
p
r
ess
io
n
d
etec
tio
n
b
y
f
ac
ial
v
id
eo
an
al
y
s
is
.
B
id
well
et
a
l
.
[
2
3
]
estab
lis
h
ed
an
au
to
m
ated
b
eh
av
io
r
al
an
aly
s
is
s
y
s
tem
in
2
0
1
1
to
en
ab
le
teac
h
er
s
to
ef
f
ec
tiv
ely
ev
alu
ate
s
tu
d
e
n
t
b
eh
av
io
r
.
Stu
d
en
t
en
g
ag
em
e
n
t
is
m
o
d
eled
a
n
d
c
ateg
o
r
ized
u
s
in
g
m
an
y
ca
m
er
as
d
ep
lo
y
ed
i
n
a
th
ir
d
-
g
r
a
d
e
c
lass
r
o
o
m
to
ca
p
tu
r
e
s
tu
d
en
t
ey
e
m
o
v
e
m
en
t p
atter
n
s
.
Fo
r
th
i
s
p
u
r
p
o
s
e,
f
iv
e
co
lo
r
ca
m
er
as
an
d
f
o
u
r
Mic
r
o
s
o
f
t
K
in
ec
t
d
ep
th
-
s
en
s
in
g
ca
m
er
as,
SDK
k
n
o
wn
as
Pit
ts
b
u
r
g
h
Patter
n
R
ec
o
g
n
itio
n
(
Pit
tPatt),
wer
e
em
p
lo
y
ed
.
T
h
e
SDK
is
u
s
ed
to
co
m
p
u
te
h
ea
d
o
r
ien
tatio
n
s
a
n
d
g
az
e
tar
g
ets.
T
h
e
h
i
d
d
en
m
ar
k
o
v
m
o
d
el
(
HM
M)
is
u
tili
ze
d
t
o
ca
teg
o
r
ize
r
etr
iev
ed
s
eq
u
en
ce
s
o
f
in
d
iv
i
d
u
al
s
tu
d
en
t
g
az
e
tar
g
ets
as
en
g
a
g
ed
,
atten
tiv
e,
o
r
tr
a
n
s
itio
n
al.
T
h
e
ex
p
er
t
o
b
s
er
v
atio
n
d
ata
was
u
tili
ze
d
to
tr
ain
an
d
ev
alu
ate
th
e
HM
M
-
b
ased
m
o
d
el
f
o
r
au
t
o
m
atic
en
g
ag
em
e
n
t
d
etec
tio
n
.
T
h
e
wo
r
k
r
elied
ju
s
t
o
n
th
e
s
tu
d
en
ts
’
g
az
e
tar
g
ets
,
wh
ich
was
in
s
u
f
f
icien
t
to
f
u
l
ly
co
m
p
r
eh
en
d
th
e
b
eh
av
io
r
.
T
h
e
p
r
i
m
ar
y
o
b
jectiv
e
o
f
th
e
s
tu
d
y
is
to
in
v
esti
g
ate
t
h
e
r
o
le
o
f
f
ac
ial
e
x
p
r
ess
io
n
s
in
ev
alu
atin
g
s
tu
d
en
t
en
g
a
g
em
en
t
with
in
a
class
r
o
o
m
en
v
ir
o
n
m
en
t.
Facial
ex
p
r
ess
io
n
s
p
r
o
v
id
e
a
n
im
m
ed
iate
an
d
v
is
ib
le
m
ea
n
s
f
o
r
ed
u
ca
t
o
r
s
to
ass
ess
th
e
lev
el
o
f
s
tu
d
en
t
in
v
o
l
v
em
e
n
t.
Ho
wev
er
,
wh
en
d
ea
lin
g
with
lar
g
er
g
r
o
u
p
s
o
f
s
tu
d
en
ts
,
th
is
m
eth
o
d
f
ac
es
s
i
g
n
if
ican
t
c
h
allen
g
es.
As
th
e
n
u
m
b
er
o
f
s
tu
d
en
ts
in
a
class
r
o
o
m
i
n
cr
ea
s
es,
th
e
ab
ilit
y
to
ac
cu
r
ately
in
ter
p
r
et
an
d
an
aly
ze
f
ac
ial
ex
p
r
ess
io
n
s
b
ec
o
m
es
m
o
r
e
co
m
p
lex
an
d
less
r
eliab
le
[
2
4
]
.
T
h
e
d
iv
er
s
ity
o
f
ex
p
r
ess
io
n
s
,
co
m
b
in
ed
with
th
e
d
if
f
icu
lty
o
f
o
b
s
er
v
in
g
ev
e
r
y
s
tu
d
en
t
at
th
e
s
am
e
tim
e,
lim
i
ts
th
e
u
s
e
o
f
f
ac
ial
cu
es
to
ass
e
s
s
en
g
ag
em
en
t.
As
a
r
e
s
u
lt,
th
er
e
is
a
p
r
ess
in
g
n
ee
d
to
in
v
esti
g
ate
an
d
im
p
lem
en
t
alter
n
ativ
e,
m
o
r
e
c
o
m
p
r
e
h
en
s
iv
e
m
eth
o
d
s
o
f
an
aly
s
is
th
at
c
an
ass
is
t
teac
h
er
s
in
b
etter
u
n
d
er
s
tan
d
in
g
s
tu
d
en
t
en
g
ag
em
e
n
t.
Seek
in
g
in
n
o
v
ati
v
e
ap
p
r
o
ac
h
es
th
at
g
o
b
ey
o
n
d
f
ac
ial
ex
p
r
ess
io
n
o
b
s
er
v
atio
n
alo
n
e,
co
n
s
id
er
i
n
g
th
e
d
y
n
am
ics
o
f
lar
g
er
class
s
izes,
b
ec
o
m
es
cr
u
cial
in
d
ev
is
in
g
a
m
o
r
e
r
o
b
u
s
t
an
d
d
ep
e
n
d
ab
le
f
r
am
ewo
r
k
f
o
r
aid
in
g
ed
u
ca
to
r
s
in
ac
cu
r
ately
g
au
g
in
g
an
d
en
h
a
n
cin
g
s
tu
d
e
n
t e
n
g
ag
em
e
n
t le
v
els.
C
u
ltu
r
al
d
if
f
er
en
ce
s
p
o
s
e
a
s
ig
n
if
ican
t
ch
allen
g
e
f
o
r
th
is
s
tu
d
y
.
E
k
m
an
[
2
5
]
cr
o
s
s
-
cu
ltu
r
al
r
esear
ch
in
d
icate
s
,
s
o
m
e
cu
ltu
r
es
o
p
en
ly
ex
p
r
ess
em
o
tio
n
s
,
wh
ile
o
t
h
er
s
co
n
ce
al
th
ei
r
f
ee
lin
g
s
.
A
ck
n
o
wled
g
i
n
g
an
d
u
n
d
er
s
tan
d
i
n
g
th
ese
v
ar
iatio
n
s
ca
n
ass
is
t
teac
h
er
s
in
cr
ea
ti
n
g
in
clu
s
iv
e
en
v
ir
o
n
m
en
ts
.
Ho
wev
er
,
f
ew
p
r
io
r
s
tu
d
ies
h
av
e
ad
d
r
ess
ed
cu
ltu
r
al
d
iv
er
s
ity
in
em
o
tio
n
al
ex
p
r
ess
io
n
.
T
o
ac
co
m
m
o
d
ate
th
ese
v
ar
iatio
n
s
an
d
s
tu
d
en
ts
’
em
o
tio
n
al
ex
p
e
r
ien
c
es,
a
n
ew
d
ataset
was
g
en
er
ated
b
y
m
er
g
i
n
g
f
iv
e
p
u
b
licly
av
ailab
le
d
atasets
.
T
h
is
ap
p
r
o
ac
h
aim
s
to
d
e
v
elo
p
a
u
n
i
v
er
s
al
m
eth
o
d
f
o
r
r
ec
o
g
n
izin
g
em
o
tio
n
s
an
d
m
o
n
ito
r
in
g
en
g
a
g
em
en
t
in
class
r
o
o
m
s
ettin
g
s
,
lev
er
ag
in
g
cu
ltu
r
al
d
if
f
er
e
n
ce
s
to
en
h
an
ce
ac
cu
r
ac
y
an
d
in
clu
s
iv
ity
.
An
o
th
er
p
er
tin
e
n
t
asp
ec
t
o
f
th
e
s
tu
d
y
in
v
o
lv
es
th
e
au
to
m
ated
ex
tr
ac
tio
n
o
f
f
r
am
es
f
r
o
m
class
r
o
o
m
v
id
e
o
s
.
T
h
is
ex
tr
ac
tio
n
p
r
o
ce
s
s
u
tili
ze
d
a
tr
ain
ed
m
o
d
el
s
p
ec
if
ically
d
esig
n
ed
f
o
r
en
g
ag
em
e
n
t
an
aly
s
is
.
Fro
m
th
ese
v
id
eo
f
r
am
es,
o
n
ly
t
h
e
two
m
o
s
t
s
ig
n
if
ican
t
em
o
tio
n
s
f
r
o
m
ea
ch
ca
teg
o
r
y
wer
e
s
elec
ted
,
s
tr
ea
m
lin
in
g
th
e
f
r
am
e
ex
tr
ac
tio
n
p
r
o
ce
d
u
r
e
to
en
h
a
n
ce
b
o
th
ac
cu
r
ac
y
a
n
d
o
v
er
all
p
er
f
o
r
m
an
ce
.
R
esear
ch
er
s
in
v
esti
g
ated
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
41
,
No
.
3
,
Ma
r
ch
20
2
6
:
9
7
7
-
999
980
v
ar
io
u
s
p
r
e
-
tr
ain
ed
m
o
d
els
f
o
r
ac
h
iev
in
g
h
ig
h
ac
cu
r
ac
y
an
d
lo
w
lo
s
s
v
alu
es
in
a
n
ewly
b
u
ilt
d
ataset
ca
lled
MA
AE
D.
T
h
ey
t
r
ain
ed
an
d
c
o
m
p
ar
ed
f
o
u
r
p
o
p
u
lar
m
o
d
els
:
v
is
u
al
g
e
o
m
etr
y
g
r
o
u
p
1
6
(
VGG1
6
)
,
R
esNet
-
5
0
,
R
esNet
-
1
0
1
,
an
d
I
n
ce
p
tio
n
V
3
.
I
n
ad
d
itio
n
to
ac
cu
r
ac
y
an
d
lo
s
s
,
th
ey
also
ev
al
u
ated
F1
-
s
co
r
e,
p
r
ec
is
io
n
,
an
d
r
ec
all
to
p
r
o
v
id
e
a
co
m
p
r
eh
e
n
s
iv
e
u
n
d
er
s
tan
d
i
n
g
o
f
ea
ch
m
o
d
el
’
s
p
er
f
o
r
m
an
ce
.
T
h
is
a
p
p
r
o
ac
h
u
tili
ze
s
a
u
n
iq
u
e
d
ataset,
MA
AE
D,
t
o
ca
teg
o
r
ize
f
ac
ial
ex
p
r
ess
io
n
s
in
to
d
if
f
er
en
t
en
g
ag
em
en
t
ca
teg
o
r
ies
s
u
ch
as
b
o
r
ed
o
m
,
co
n
f
u
s
io
n
,
f
r
u
s
tr
a
tio
n
,
y
awn
in
g
,
an
d
c
o
n
ce
n
tr
ated
aid
in
g
ed
u
ca
to
r
s
in
ass
es
s
in
g
s
tu
d
en
t
en
g
ag
em
e
n
t
o
b
jectiv
ely
an
d
e
f
f
icien
tly
r
ath
er
th
a
n
it
s
o
lv
es
th
e
p
r
o
b
lem
o
f
m
a
n
u
al
s
elec
tio
n
o
f
p
ea
k
f
r
am
es
in
ea
ch
ca
teg
o
r
y
o
f
em
o
tio
n
s
b
y
c
o
n
s
id
er
in
g
v
a
r
iety
o
f
ex
p
r
ess
io
n
ar
o
u
n
d
th
e
wo
r
l
d
with
th
e
h
elp
o
f
MA
E
E
D.
T
h
e
lo
s
s
v
alu
e
h
av
e
b
ee
n
g
r
a
d
u
ally
d
ec
r
ea
s
ed
in
th
e
p
r
o
p
o
s
ed
m
o
d
el
as th
e
ac
cu
r
ac
y
im
p
r
o
v
es.
T
h
e
f
o
u
n
d
atio
n
o
f
t
h
is
p
ap
er
is
r
o
o
ted
in
its
m
eth
o
d
o
lo
g
y
,
o
u
tlin
in
g
th
e
m
eth
o
d
s
em
p
lo
y
ed
in
th
e
s
tu
d
y
.
Fo
llo
win
g
t
h
is
in
tr
o
d
u
c
tio
n
,
s
u
b
s
eq
u
e
n
t
s
ec
tio
n
s
d
elv
e
in
to
d
etailed
p
r
o
ce
s
s
es,
an
aly
s
es,
an
d
o
u
tc
o
m
es
ex
p
lo
r
ed
with
in
th
e
r
esear
c
h
.
T
h
e
m
eth
o
d
o
lo
g
y
s
ec
tio
n
o
u
tlin
es
h
o
w
th
e
s
tu
d
y
was
co
n
d
u
cted
,
d
etailin
g
t
h
e
to
o
ls
an
d
tech
n
iq
u
es
ap
p
lied
.
Fo
llo
win
g
th
is
,
th
e
liter
atu
r
e
r
ev
iew
cr
itically
ev
alu
ates
an
d
co
n
s
o
lid
ates
p
er
tin
en
t
ex
is
tin
g
r
esear
ch
,
e
m
p
h
asizin
g
k
e
y
in
s
ig
h
ts
.
I
t
th
en
m
o
v
es
o
n
to
th
e
in
te
g
r
atio
n
an
d
a
n
aly
s
is
o
f
th
e
MA
AE
D
d
a
taset,
ex
p
lain
in
g
f
r
am
e
p
r
ep
r
o
ce
s
s
in
g
,
m
o
d
el
tr
ain
in
g
,
a
n
d
s
u
b
s
eq
u
en
t
e
v
alu
atio
n
u
s
in
g
d
iv
er
s
e
m
etr
ics,
alo
n
g
with
d
is
cu
s
s
in
g
au
g
m
e
n
tatio
n
m
eth
o
d
s
f
o
r
e
n
h
an
ce
d
p
er
f
o
r
m
an
ce
.
A
d
eta
iled
an
aly
s
is
o
f
th
e
f
in
d
in
g
s
u
n
f
o
ld
s
,
ex
p
l
o
r
in
g
th
e
o
v
er
all
p
er
f
o
r
m
an
ce
m
et
r
ics
o
f
th
e
s
tu
d
y
.
L
astl
y
,
p
o
ten
tial
ch
allen
g
es
im
p
ac
tin
g
th
e
s
tu
d
y
’
s
o
u
tco
m
es a
r
e
r
ev
iewe
d
in
th
e
co
n
clu
s
io
n
.
2.
M
E
T
H
O
D
I
n
th
is
s
tu
d
y
,
we
p
r
o
p
o
s
e
a
m
eth
o
d
o
l
o
g
y
f
o
r
f
ac
ial
en
g
a
g
em
en
t
m
o
n
ito
r
i
n
g
in
ed
u
ca
tio
n
al
s
ettin
g
s
u
s
in
g
a
C
NN.
T
h
e
m
eth
o
d
o
l
o
g
y
in
co
r
p
o
r
ates
th
e
cr
ea
tio
n
o
f
th
e
MA
AE
D,
wh
ich
co
m
b
in
es
d
iv
er
s
e
f
ac
ial
ex
p
r
ess
io
n
d
atasets
co
v
er
in
g
a
wid
e
r
an
g
e
o
f
em
o
tio
n
s
an
d
en
g
ag
e
m
en
t
lev
els
r
elev
an
t
to
s
tu
d
en
ts
.
T
h
e
co
llected
d
atasets
ar
e
p
r
ep
r
o
c
ess
ed
to
en
s
u
r
e
u
n
if
o
r
m
ity
a
n
d
co
n
s
is
ten
cy
.
T
h
e
d
esig
n
ed
C
NN
ar
ch
itectu
r
e
co
n
s
is
ts
o
f
co
n
v
o
lu
tio
n
al
lay
er
s
,
p
o
o
lin
g
la
y
er
s
,
an
d
f
u
lly
co
n
n
ec
ted
lay
er
s
,
wh
ic
h
ar
e
tr
ain
ed
o
n
th
e
MA
AE
D
u
s
in
g
s
u
itab
le
o
p
tim
izatio
n
alg
o
r
ith
m
s
an
d
lo
s
s
f
u
n
ctio
n
s
f
o
r
m
u
lti
-
class
clas
s
if
icatio
n
.
T
h
e
tr
ain
ed
C
NN
m
o
d
el
is
ev
alu
ated
o
n
a
test
s
et
u
s
in
g
s
tan
d
ar
d
e
v
alu
atio
n
m
e
tr
ics,
s
u
ch
as
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
s
co
r
e,
as we
ll a
s
a
co
n
f
u
s
io
n
m
atr
ix
to
ass
ess
its
p
er
f
o
r
m
an
ce
.
C
NNs
ar
e
a
ca
teg
o
r
y
o
f
d
ee
p
lear
n
i
n
g
m
o
d
els
e
x
ten
s
iv
el
y
em
p
l
o
y
ed
in
d
iv
e
r
s
e
co
m
p
u
ter
v
is
io
n
ap
p
licatio
n
s
,
in
clu
d
in
g
FER
.
C
NNs
ar
e
s
tr
u
ctu
r
ed
t
o
a
u
to
m
atica
lly
ac
q
u
ir
e
an
d
e
x
tr
ac
t
m
ea
n
in
g
f
u
l
f
ea
t
u
r
es
f
r
o
m
in
p
u
t
d
ata,
wh
ich
g
r
ea
tly
en
h
a
n
ce
s
th
eir
ef
f
ec
tiv
e
n
es
s
in
th
e
f
ield
o
f
im
ag
e
an
aly
s
i
s
.
I
n
th
e
co
n
tex
t
o
f
FER
,
C
N
N
ca
n
b
e
tr
ain
ed
to
d
etec
t
an
d
class
if
y
d
if
f
er
e
n
t
f
a
cial
ex
p
r
ess
io
n
s
b
y
lear
n
in
g
p
atter
n
s
an
d
f
ea
tu
r
es
f
r
o
m
f
ac
ial
im
a
g
es.
T
h
e
n
etw
o
r
k
is
tr
ain
ed
to
r
ec
o
g
n
ize
c
r
u
cial
f
ac
ial
lan
d
m
ar
k
s
,
in
clu
d
in
g
th
e
ey
es,
n
o
s
e,
an
d
m
o
u
t
h
,
an
d
th
eir
s
p
atial
r
elatio
n
s
h
ip
s
to
ca
p
tu
r
e
th
e
d
is
tin
ct
f
ea
tu
r
es
ass
o
ciate
d
with
d
if
f
er
en
t
em
o
tio
n
s
.
T
h
e
co
r
e
d
esig
n
o
f
a
C
NN
in
clu
d
es
s
ev
er
al
lay
er
s
,
s
u
c
h
as
co
n
v
o
l
u
tio
n
al
la
y
er
s
,
p
o
o
li
n
g
lay
e
r
s
,
an
d
f
u
lly
co
n
n
ec
ted
lay
er
s
.
C
o
n
v
o
l
u
ti
o
n
al
lay
er
s
p
lay
a
p
iv
o
tal
r
o
le
in
f
ea
tu
r
e
ex
tr
ac
tio
n
b
y
a
p
p
ly
i
n
g
f
ilter
s
to
th
e
in
p
u
t
im
ag
e,
en
a
b
lin
g
th
e
d
etec
tio
n
o
f
lo
ca
l p
atter
n
s
an
d
f
ea
tu
r
es.
T
h
e
co
n
v
o
lu
tio
n
p
r
o
ce
s
s
en
tails
th
e
m
o
v
em
e
n
t
o
f
a
f
ilter
(
K)
ac
r
o
s
s
th
e
i
n
p
u
t
im
ag
e
(
I
)
,
wh
e
r
e
it
co
n
d
u
c
ts
elem
en
t
-
wis
e
m
u
ltip
licatio
n
s
an
d
s
u
b
s
eq
u
en
tly
ag
g
r
eg
ates th
e
o
u
tco
m
es b
y
s
u
m
m
in
g
th
em
u
p
.
(
∗
)
(
)
=
(
)
(
−
)
(
1
)
T
h
is
ess
en
tially
m
ea
n
s
th
at
e
v
er
y
p
ix
el
in
th
e
o
u
tp
u
t
is
g
e
n
er
ated
b
y
a
d
d
in
g
to
g
eth
er
th
e
in
p
u
t
p
ix
els,
ea
c
h
m
u
ltip
lied
b
y
its
r
esp
ec
tiv
e
we
ig
h
t d
ef
in
e
d
b
y
th
e
k
e
r
n
el.
I
n
t
wo
d
im
en
s
io
n
s
,
th
is
wo
u
ld
b
e
,
(
∗
)
(
,
)
=
(
,
)
(
−
,
−
)
(
2
)
I
n
th
is
p
r
o
ce
s
s
,
th
e
k
er
n
el
u
n
d
er
g
o
es
elem
en
t
-
wis
e
m
u
ltip
licatio
n
with
th
e
im
ag
e
m
atr
ix
,
an
d
af
ter
war
d
s
,
th
e
o
u
tco
m
es
ar
e
s
u
m
m
ed
u
p
.
Po
o
lin
g
lay
e
r
s
s
er
v
e
to
d
o
wn
s
am
p
le
th
e
f
ea
tu
r
e
m
ap
s
,
r
ed
u
cin
g
th
eir
s
p
atial
d
im
en
s
io
n
s
wh
i
le
p
r
eser
v
in
g
cr
itical
in
f
o
r
m
atio
n
.
T
wo
f
r
eq
u
e
n
tly
u
s
ed
tech
n
iq
u
es
f
o
r
th
is
p
u
r
p
o
s
e
ar
e
m
ax
p
o
o
lin
g
an
d
av
er
ag
e
p
o
o
lin
g
.
Ma
x
p
o
o
lin
g
,
as
a
tech
n
iq
u
e,
f
o
cu
s
es
o
n
ex
tr
ac
tin
g
th
e
h
ig
h
est
v
alu
e
with
in
ea
ch
win
d
o
w
o
f
th
e
f
ea
tu
r
e
m
ap
,
th
er
e
b
y
h
ig
h
lig
h
tin
g
th
e
m
o
s
t
p
r
o
m
in
en
t
f
ea
tu
r
es
with
in
th
e
d
ata.
I
n
co
n
tr
ast,
av
er
ag
e
p
o
o
lin
g
co
m
p
u
tes
th
e
av
er
a
g
e
v
alu
e
f
o
r
ea
c
h
win
d
o
w,
g
iv
in
g
an
eq
u
a
l
r
ep
r
esen
tatio
n
o
f
all
f
ea
tu
r
es
with
in
th
e
win
d
o
w.
T
h
is
s
p
atial
r
ed
u
ctio
n
ca
r
r
ied
o
u
t
b
y
th
e
p
o
o
lin
g
lay
e
r
s
en
s
u
r
es th
at
th
e
m
o
s
t salien
t f
ea
tu
r
es a
r
e
r
etain
ed
wh
ile
th
e
o
v
er
all
d
ata
s
ize
is
co
n
d
en
s
ed
,
m
ak
in
g
s
u
b
s
eq
u
e
n
t
lay
er
s
o
f
th
e
C
NN
m
o
r
e
co
m
p
u
tatio
n
a
lly
m
an
a
g
ea
b
le.
Fin
a
lly
,
th
e
f
u
lly
co
n
n
ec
te
d
lay
er
s
ar
e
r
esp
o
n
s
ib
le
f
o
r
class
if
icatio
n
,
m
ap
p
in
g
t
h
e
e
x
tr
ac
ted
f
ea
tu
r
es
to
s
p
ec
if
ic
em
o
tio
n
ca
teg
o
r
ies.
Af
ter
th
e
f
ea
tu
r
e
e
x
tr
ac
tio
n
p
r
o
ce
s
s
in
v
o
lv
in
g
th
e
c
o
n
v
o
lu
tio
n
al
an
d
p
o
o
lin
g
lay
er
s
,
th
e
f
u
lly
co
n
n
ec
ted
lay
e
r
s
co
m
e
i
n
to
p
lay
,
m
ap
p
i
n
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
Leve
r
a
g
in
g
C
N
N
to
a
n
a
lyze fa
cia
l e
xp
r
ess
io
n
s
fo
r
a
ca
d
emic
en
g
a
g
eme
n
t
mo
n
ito
r
in
g
w
ith
…
(
N
o
o
r
a
C
.
T.
)
981
th
e
ex
tr
ac
ted
f
ea
t
u
r
es
to
s
p
ec
i
f
ic
class
ca
teg
o
r
ies.
Ma
th
em
atica
lly
,
a
f
u
lly
co
n
n
ec
ted
lay
e
r
ex
ec
u
tes
a
lin
ea
r
tr
an
s
f
o
r
m
atio
n
an
d
s
u
b
s
eq
u
e
n
tly
ap
p
lies
an
ac
tiv
atio
n
f
u
n
ctio
n
.
I
f
we
d
e
n
o
te
th
e
i
n
p
u
t
to
th
e
la
y
er
as
x
(
wh
ich
wo
u
l
d
b
e
a
f
latten
e
d
v
er
s
io
n
o
f
th
e
o
u
tp
u
t
f
r
o
m
th
e
p
r
ev
io
u
s
la
y
er
)
,
th
e
weig
h
ts
as
W
,
th
e
b
iases
as
b
,
an
d
th
e
o
u
tp
u
t a
s
y
,
th
en
t
h
e
lin
ea
r
tr
an
s
f
o
r
m
atio
n
ca
n
b
e
wr
i
tten
as:
=
+
(
3
)
t
h
e
weig
h
ts
m
atr
ix
W
an
d
th
e
b
ias
v
ec
to
r
b
r
e
p
r
esen
t
th
e
p
ar
am
eter
s
o
f
th
e
f
u
lly
co
n
n
e
cted
lay
er
th
at
a
r
e
lear
n
ed
d
u
r
in
g
tr
ain
in
g
.
Af
te
r
th
e
lin
ea
r
tr
an
s
f
o
r
m
atio
n
,
an
ac
tiv
atio
n
f
u
n
ctio
n
is
ap
p
lied
elem
en
t
-
wis
e.
T
h
e
r
ec
tifie
d
lin
ea
r
u
n
it
(
R
eL
U)
is
a
co
m
m
o
n
l
y
ad
o
p
ted
ac
t
iv
atio
n
f
u
n
ctio
n
.
(
)
=
(
0
,
)
(
4)
I
f
th
e
f
u
lly
c
o
n
n
ec
ted
lay
er
i
s
p
lace
d
as
th
e
f
in
al
lay
er
i
n
th
e
n
etwo
r
k
an
d
is
u
s
ed
f
o
r
m
u
lti
-
class
class
if
icatio
n
,
th
en
a
So
f
tMa
x
f
u
n
ctio
n
is
ty
p
ically
ap
p
lied
t
o
th
e
o
u
tp
u
t
o
f
th
e
lay
e
r
to
g
e
n
er
ate
p
r
o
b
ab
ilit
ies
f
o
r
ea
ch
class
.
(
)
=
(
)
⁄
(
)
(
5
)
wh
er
e
th
e
v
ec
to
r
z
s
er
v
es
as
th
e
in
p
u
t
to
th
e
So
f
tMa
x
f
u
n
ctio
n
,
zi
is
th
e
i
th
elem
en
t
o
f
z,
an
d
th
e
d
en
o
m
in
a
to
r
is
th
e
s
u
m
o
f
ex
p
(
zj)
o
v
er
all
j.
C
NN,
a
s
p
ec
ial
ized
class
o
f
n
eu
r
al
n
etwo
r
k
s
ca
n
b
e
s
ee
in
Fig
u
r
e
1
,
h
av
e
g
ai
n
ed
p
r
o
m
in
en
ce
d
u
e
to
th
eir
ab
ilit
y
to
a
u
to
m
atica
lly
e
x
tr
ac
t
m
ea
n
i
n
g
f
u
l
f
ea
tu
r
es
f
r
o
m
im
ag
es.
I
n
th
e
r
ea
lm
o
f
FE
R
,
th
is
ch
ar
ac
ter
is
tic
p
r
o
v
es
in
v
alu
a
b
le
as
it
elim
in
ates
th
e
n
ee
d
f
o
r
m
an
u
al
f
e
atu
r
e
ex
tr
ac
tio
n
,
allo
win
g
th
e
m
o
d
el
to
d
is
ce
r
n
in
tr
ica
te
p
atter
n
s
a
n
d
s
u
b
tle
n
u
an
ce
s
in
h
e
r
en
t
in
f
ac
ial
ex
p
r
e
s
s
io
n
s
.
T
r
ain
in
g
with
a
C
NN
m
o
d
el
(
Fig
u
r
e
1
(
a)
)
,
s
ev
er
al
s
ig
n
if
ican
t
o
b
s
tacle
s
lik
e
o
v
er
f
itti
n
g
,
v
an
is
h
in
g
g
r
ad
ien
ts
,
an
d
class
im
b
alan
ce
in
th
e
d
ataset
ca
n
n
eg
ativ
ely
im
p
ac
t th
e
m
o
d
el
’
s
p
er
f
o
r
m
a
n
ce
.
T
o
m
itig
ate
th
e
s
e
is
s
u
es,
it b
ec
o
m
es n
ec
es
s
ar
y
to
ca
r
ef
u
lly
ad
ju
s
t
th
e
m
o
d
el
’
s
h
y
p
er
-
p
a
r
am
eter
s
an
d
ap
p
ly
r
eg
u
lar
izatio
n
tech
n
iq
u
es.
Hy
p
er
-
p
ar
am
eter
s
ar
e
e
lem
en
ts
th
at
g
u
id
e
th
e
lear
n
in
g
p
r
o
ce
s
s
o
f
th
e
m
o
d
el.
T
h
ese
in
clu
d
e
asp
ec
ts
s
u
ch
as
b
atch
s
ize,
k
er
n
el
s
ize
,
th
e
ch
o
ice
o
f
lo
s
s
f
u
n
ctio
n
,
an
d
th
e
o
p
tim
izatio
n
alg
o
r
ith
m
.
Ad
ju
s
tin
g
th
ese
ca
n
h
elp
m
an
ag
e
th
e
af
o
r
em
e
n
tio
n
ed
c
h
allen
g
es
ef
f
ec
tiv
ely
.
R
eg
u
lar
izatio
n
te
ch
n
iq
u
es
ar
e
a
g
r
o
u
p
o
f
s
tr
at
eg
ies
aim
ed
at
p
r
e
v
en
tin
g
o
v
er
f
itti
n
g
,
a
s
ce
n
ar
i
o
wh
er
e
th
e
m
o
d
el
e
x
ce
ls
with
tr
ain
in
g
d
ata
b
u
t
s
tr
u
g
g
les
wh
en
f
ac
ed
with
u
n
f
am
iliar
o
r
u
n
s
ee
n
d
ata.
L
1
a
n
d
L
2
r
eg
u
lar
izatio
n
,
d
r
o
p
o
u
t,
d
ata
au
g
m
en
tatio
n
,
an
d
ea
r
l
y
s
to
p
p
in
g
,
ar
e
s
o
m
e
o
f
th
e
co
m
m
o
n
r
eg
u
la
r
izatio
n
tech
n
iq
u
es
u
s
ed
.
I
n
tr
ain
i
n
g
th
e
FER
m
o
d
el,
th
ese
ch
a
llen
g
es
w
er
e
m
an
ag
ed
ef
f
ec
tiv
ely
,
r
esu
ltin
g
in
co
m
m
en
d
a
b
le
class
if
icatio
n
ac
cu
r
ac
y
.
Hy
p
er
-
p
a
r
am
eter
tu
n
in
g
an
d
r
e
g
u
lar
izatio
n
tech
n
iq
u
es
wer
e
em
p
lo
y
ed
to
o
p
tim
ize
th
e
m
o
d
el,
t
h
er
e
b
y
lead
in
g
to
a
m
o
r
e
b
alan
ce
d
an
d
ac
c
u
r
ate
class
if
icatio
n
o
f
f
ac
ial
em
o
ti
o
n
s
.
T
h
e
in
p
u
t
im
a
g
e
u
n
d
er
g
o
es
a
s
er
ies
o
f
co
n
v
o
l
u
tio
n
al
lay
er
s
,
f
o
llo
wed
b
y
p
o
o
lin
g
la
y
er
s
,
wh
ich
p
r
o
g
r
ess
iv
ely
r
ed
u
ce
th
e
s
p
atial
d
im
e
n
s
io
n
s
.
T
h
e
f
latten
ed
r
esu
lt
f
r
o
m
t
h
e
last
p
o
o
lin
g
la
y
er
is
f
e
d
in
to
f
u
lly
co
n
n
ec
ted
la
y
e
r
s
,
wh
er
e
class
if
icatio
n
is
ca
r
r
ied
o
u
t
u
s
in
g
th
e
e
x
tr
ac
ted
f
ea
tu
r
es.
T
h
e
f
in
al
o
u
tp
u
t
la
y
er
r
ep
r
esen
ts
th
e
d
if
f
er
e
n
t
em
o
tio
n
class
es,
s
u
ch
as
h
a
p
p
in
ess
,
s
ad
n
ess
,
an
d
an
g
e
r
.
T
h
e
k
ey
ad
v
an
tag
e
o
f
C
NN
in
FER
is
th
eir
in
h
er
en
t
ca
p
ab
ilit
y
to
lear
n
an
d
ex
t
r
ac
t
r
elev
an
t
f
ea
tu
r
es
au
to
m
atica
lly
f
r
o
m
f
ac
ial
im
ag
es,
th
er
e
b
y
elim
in
atin
g
th
e
r
eq
u
ir
em
e
n
t
f
o
r
m
a
n
u
al
f
ea
tu
r
e
en
g
i
n
ee
r
in
g
.
T
h
is
tr
ait
m
ak
es
C
NNs
h
ig
h
ly
p
r
o
f
icien
t
at
ca
p
tu
r
in
g
in
tr
icate
p
atter
n
s
an
d
s
u
b
tle
d
etails lin
k
ed
to
v
ar
i
o
u
s
em
o
tio
n
s
.
T
h
e
h
ea
r
t
o
f
t
h
is
s
tu
d
y
is
th
e
MA
AE
D
d
ataset,
wh
ich
co
m
b
i
n
es
f
ac
ial
ex
p
r
ess
io
n
d
ata
f
r
o
m
f
iv
e
p
u
b
licly
av
ailab
le
s
o
u
r
ce
s
wo
r
ld
wid
e.
T
h
e
d
ataset
cr
ea
tio
n
in
v
o
lv
es
s
ev
er
al
s
tep
s
,
f
r
o
m
g
ath
er
in
g
d
iv
er
s
e
ex
p
r
ess
io
n
s
to
r
ef
in
in
g
th
e
f
r
a
m
es.
T
h
e
p
iv
o
tal
s
u
cc
ess
in
th
is
p
r
o
ce
s
s
in
v
o
lv
es lev
er
ag
in
g
a
tr
ain
ed
m
o
d
el
f
o
r
a
u
to
m
ated
f
r
am
e
ex
tr
ac
tio
n
,
s
ig
n
if
ican
tly
r
ed
u
cin
g
tim
e
co
m
p
ar
ed
to
m
a
n
u
al
s
elec
tio
n
.
Alth
o
u
g
h
th
e
d
ataset
en
co
m
p
ass
es
f
iv
e
em
o
tio
n
s
,
th
is
s
tu
d
y
p
r
im
ar
ily
f
o
c
u
s
es
o
n
f
o
u
r
s
p
ec
if
ic
n
eg
ativ
e
e
m
o
tio
n
s
—
b
o
r
ed
o
m
,
f
r
u
s
tr
atio
n
,
y
awn
in
g
,
a
n
d
co
n
f
u
s
io
n
—
as
an
in
itial
s
tep
.
T
h
e
co
n
ce
n
tr
ated
f
r
a
m
es
co
r
r
esp
o
n
d
in
g
to
th
ese
em
o
tio
n
s
ar
e
n
o
t
with
in
th
e
s
co
p
e
o
f
th
is
s
tu
d
y
’
s
co
n
s
id
er
atio
n
s
.
I
n
Fig
u
r
e
1
(
b
)
illu
s
tr
ates
th
e
MA
AE
D
d
ataset
cr
ea
tio
n
m
eth
o
d
an
d
th
e
en
g
ag
em
en
t
class
if
icatio
n
p
r
o
ce
s
s
b
ased
o
n
th
e
M
AAE
D
d
ataset.
T
h
e
de
tailed
p
r
o
ce
s
s
o
f
d
ataset
cr
ea
tio
n
is
elab
o
r
ated
in
s
ec
tio
n
4
o
f
th
is
s
tu
d
y
.
Up
o
n
tr
ain
i
n
g
with
p
r
e
-
tr
ain
ed
m
o
d
els
lik
e
VGG1
6
,
R
esNet
5
0
,
R
esNet1
0
1
,
an
d
I
n
ce
p
tio
n
V3
,
th
e
p
r
o
p
o
s
ed
m
o
d
el
d
e
m
o
n
s
tr
ates
s
u
p
er
io
r
p
er
f
o
r
m
an
ce
ac
r
o
s
s
v
ar
io
u
s
ev
alu
atio
n
m
etr
ics in
cl
u
d
in
g
ac
c
u
r
ac
y
,
lo
s
s
,
an
d
o
th
er
r
elev
an
t
b
en
ch
m
ar
k
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
41
,
No
.
3
,
Ma
r
ch
20
2
6
:
9
7
7
-
999
982
(
a)
(
b
)
Fig
u
r
e
1
.
Sp
ec
ialized
class
o
f
n
eu
r
al
n
etwo
r
k
s
(
a)
a
r
ch
itectu
r
e
o
f
C
NN
an
d
(
b
)
MA
AE
D
cr
ea
tio
n
,
class
if
icatio
n
an
d
p
r
ed
ictio
n
3.
RE
L
AT
E
D
WO
RK
S
Un
d
er
s
tan
d
in
g
an
d
ass
ess
in
g
s
tu
d
en
ts
’
en
g
ag
e
m
en
t
lev
els
a
n
d
em
o
tio
n
al
s
tates
p
lay
a
cr
u
cial
r
o
le
in
s
h
ap
in
g
ef
f
ec
tiv
e
teac
h
in
g
s
tr
ateg
ies
an
d
p
r
o
m
o
tin
g
o
p
tim
al
lear
n
in
g
ex
p
e
r
ien
ce
s
.
E
ac
h
s
tu
d
y
i
n
T
a
b
le
1
ad
o
p
ts
v
ar
io
u
s
m
eth
o
d
o
lo
g
ies,
r
an
g
in
g
f
r
o
m
tr
ad
itio
n
al
m
ac
h
in
e
lear
n
in
g
m
o
d
els
to
s
o
p
h
is
t
icate
d
d
ee
p
lear
n
in
g
ar
c
h
itectu
r
es.
T
h
ese
ap
p
r
o
ac
h
es
ar
e
d
esig
n
ed
to
ca
p
tu
r
e
a
wid
e
ar
r
ay
o
f
n
o
n
-
v
e
r
b
al
cu
es,
in
clu
d
in
g
f
ac
ial
ex
p
r
ess
io
n
s
,
b
o
d
y
lan
g
u
ag
e,
an
d
ey
e
m
o
v
em
e
n
ts
.
T
h
ese
ap
p
r
o
ac
h
es
ar
e
d
esig
n
ed
t
o
d
ec
ip
h
er
s
tu
d
en
ts
’
en
g
ag
em
e
n
t
lev
els
an
d
e
m
o
tio
n
a
l
r
esp
o
n
s
es
d
u
r
in
g
t
h
e
lear
n
in
g
p
r
o
ce
s
s
.
Ad
v
an
ce
d
d
ee
p
lear
n
in
g
m
eth
o
d
s
in
ce
r
tain
s
tu
d
ies
ex
em
p
lif
y
o
n
g
o
in
g
p
r
o
g
r
ess
in
e
d
u
ca
tio
n
,
o
f
f
er
i
n
g
e
d
u
ca
to
r
s
d
ee
p
er
i
n
s
ig
h
ts
in
to
h
u
m
an
em
o
tio
n
s
an
d
e
n
g
ag
e
m
en
t.
W
h
iteh
ill
et
a
l.
[
2
6
]
co
n
d
u
c
ted
a
t
h
o
r
o
u
g
h
an
al
y
s
is
o
f
e
x
is
tin
g
co
m
p
u
ter
-
v
is
io
n
alg
o
r
ith
m
s
f
o
r
au
to
m
atic
s
tu
d
en
t
e
n
g
ag
e
m
en
t
an
aly
s
is
an
d
r
ec
o
g
n
itio
n
.
V
io
la
an
d
J
o
n
es
[
2
7
]
co
m
p
ar
e
d
f
ac
ial
f
ea
tu
r
es
o
f
f
ac
e
p
atch
es
f
r
o
m
v
ar
io
u
s
m
eth
o
d
s
s
u
ch
as
B
o
o
s
tB
F,
S
V
M,
Gab
o
r
,
an
d
C
E
R
T
to
o
lb
o
x
,
an
d
d
i
d
a
b
in
ar
y
class
if
icati
o
n
o
f
th
e
f
o
u
r
ty
p
e
s
o
f
en
g
ag
em
e
n
t
o
n
th
e
f
ac
ia
l
ex
p
r
ess
io
n
,
an
d
f
in
al
e
n
g
ag
em
en
t
is
esti
m
ated
f
r
o
m
a
r
eg
r
ess
io
n
m
o
d
el
u
s
in
g
th
e
b
in
ar
y
class
if
icatio
n
o
u
t
p
u
ts
.
Z
alete
lj
an
d
Ko
š
ir
[
28
]
d
ev
elo
p
a
f
ea
tu
r
e
s
et
d
ef
in
in
g
b
o
th
th
e
f
ac
e
an
d
b
o
d
ily
attr
ib
u
tes
o
f
a
s
tu
d
en
t,
in
cl
u
d
in
g
g
az
e
p
o
in
t
a
n
d
b
o
d
y
p
o
s
tu
r
e,
u
s
in
g
2
D
an
d
3
D
d
ata
r
ec
eiv
ed
b
y
th
e
Kin
ec
t
o
n
e
s
en
s
o
r
.
Ma
ch
in
e
lear
n
in
g
tech
n
i
q
u
es
ar
e
u
s
ed
t
o
tr
ain
class
if
ier
s
th
at
ass
es
s
a
s
tu
d
en
t
’
s
atten
tio
n
l
ev
els
at
v
ar
io
u
s
in
ter
v
als.
Kr
ith
ik
a
an
d
Priy
a
[
29
]
d
ev
el
o
p
ed
a
p
r
o
g
r
am
t
o
id
en
tify
th
e
em
o
tio
n
s
o
f
th
e
s
tu
d
en
ts
b
y
m
o
n
ito
r
in
g
th
eir
h
ea
d
,
lip
,
an
d
ey
e
m
o
v
em
en
t
s
in
th
e
e
-
lear
n
in
g
en
v
ir
o
n
m
en
t.
Sah
la
an
d
Ku
m
ar
[3
0
]
d
ev
el
o
p
ed
a
d
ee
p
C
NN
tech
n
i
q
u
e
f
o
r
class
r
o
o
m
em
o
tio
n
d
etec
tio
n
.
A
clo
u
d
-
b
ased
f
ac
ial
em
o
tio
n
an
aly
s
is
was
co
n
d
u
cted
(
2
0
1
9
)
b
y
B
o
o
n
r
o
u
n
g
r
u
t
et
a
l
.
[3
1
]
in
f
ac
ial
em
o
tio
n
an
aly
s
is
to
f
in
d
s
tu
d
en
ts
’
em
o
tio
n
s
in
th
e
class
r
o
o
m
.
T
h
e
s
tu
d
y
was
co
n
d
u
cted
am
o
n
g
2
9
in
te
r
n
atio
n
al
s
tu
d
en
ts
b
y
ex
am
in
in
g
th
eir
m
o
o
d
ch
a
n
g
es.
Ay
v
az
et
a
l.
[3
2
]
em
p
lo
y
s
ev
er
al
class
if
icatio
n
alg
o
r
ith
m
s
s
u
ch
as
C
AR
T
,
R
F,
k
NN,
an
d
SVM
to
an
aly
ze
th
e
f
ac
ial
ex
p
r
ess
io
n
s
o
f
p
a
r
ticip
an
ts
in
an
e
-
lear
n
in
g
s
ess
io
n
h
el
d
o
v
er
Sk
y
p
e
s
o
f
twar
e
u
s
in
g
th
e
s
y
s
tem
th
ey
b
u
ilt.
T
h
e
y
ev
e
n
tu
ally
co
n
cl
u
d
ed
th
at
em
o
tio
n
s
s
u
ch
as
h
a
p
p
in
ess
,
f
ea
r
,
s
ad
n
ess
,
an
g
e
r
,
s
u
r
p
r
is
e,
an
d
d
is
g
u
s
t
ar
e
u
n
iv
e
r
s
ally
ac
k
n
o
wled
g
e
d
in
class
r
o
o
m
s
.
Am
o
n
g
,
th
e
SVM
alg
o
r
ith
m
o
u
tp
er
f
o
r
m
s
o
th
er
s
.
R
ec
en
t
n
eu
r
o
l
o
g
ical
ad
v
an
ce
m
en
ts
h
ig
h
lig
h
t
th
e
co
n
n
ec
tio
n
b
etwe
en
lear
n
in
g
an
d
em
o
tio
n
s
.
Ma
n
y
s
tu
d
ies em
p
h
asize
th
e
im
p
o
r
tan
ce
o
f
s
t
u
d
en
ts
’
em
o
tio
n
s
d
u
r
i
n
g
lectu
r
es.
Ack
n
o
wled
g
in
g
th
is
s
tr
o
n
g
lin
k
b
etwe
en
e
m
o
tio
n
s
an
d
lear
n
in
g
,
it
’
s
cr
u
cial
to
in
te
g
r
ate
em
o
ti
o
n
s
in
to
ed
u
ca
tio
n
.
Un
d
er
s
tan
d
in
g
an
d
s
u
p
p
o
r
tin
g
s
tu
d
en
ts
’
f
ee
lin
g
s
lead
s
t
o
im
p
r
o
v
ed
,
cu
s
to
m
ized
lear
n
in
g
e
x
p
er
ie
n
ce
s
,
en
h
an
cin
g
b
o
t
h
ac
ad
em
ic
p
er
f
o
r
m
an
ce
a
n
d
s
tu
d
e
n
ts
’
g
en
er
al
well
-
b
ein
g
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
Leve
r
a
g
in
g
C
N
N
to
a
n
a
lyze fa
cia
l e
xp
r
ess
io
n
s
fo
r
a
ca
d
emic
en
g
a
g
eme
n
t
mo
n
ito
r
in
g
w
ith
…
(
N
o
o
r
a
C
.
T.
)
983
T
ab
le
1
.
A
co
m
p
ar
ativ
e
a
n
aly
s
is
f
o
r
u
n
d
er
s
tan
d
i
n
g
p
er
f
o
r
m
a
n
ce
in
d
iv
e
r
s
e
lear
n
in
g
s
ettin
g
s
S
t
u
d
y
M
e
t
h
o
d
o
l
o
g
y
C
a
t
e
g
o
r
i
e
s
A
c
c
u
r
a
c
y
E
-
l
e
a
r
n
i
n
g
/
C
l
a
s
sr
o
o
m
1.
W
h
i
t
e
h
i
l
l
e
t
a
l
.
[
2
6
]
S
V
M
w
i
t
h
G
a
b
o
u
r
f
e
a
t
u
r
e
s
N
o
t
e
n
g
a
g
e
d
a
t
a
l
l
,
n
o
m
i
n
a
l
l
y
e
n
g
a
g
e
d
,
e
n
g
a
g
e
d
i
n
t
a
s
k
,
v
e
r
y
e
n
g
a
g
e
d
,
u
n
c
l
e
a
r
.
7
6
.
3
2
E
-
l
e
a
r
n
i
n
g
2.
B
o
s
c
h
e
t
a
l
.
[
3
3
]
1
4
d
i
f
f
e
r
e
n
t
ma
c
h
i
n
e
l
e
a
r
n
i
n
g
m
o
d
e
l
s
l
i
k
e
S
V
M
-
u
s
i
n
g
f
a
c
i
a
l
e
x
p
r
e
ssi
o
n
s
B
o
r
e
d
,
c
o
n
f
u
se
d
,
d
e
l
i
g
h
t
e
d
,
e
n
g
a
g
e
d
,
f
r
u
st
r
a
t
e
d
I
n
d
i
v
i
d
u
a
l
c
l
a
ss
a
c
c
u
r
a
c
y
(
0
.
6
1
-
0
.
8
7
)
E
-
l
e
a
r
n
i
n
g
3.
K
r
i
t
h
i
k
a
e
t
a
l
.
[
2
9
]
F
a
c
i
a
l
f
e
a
t
u
r
e
s
l
i
k
e
a
b
n
o
r
mal
h
e
a
d
a
n
d
e
y
e
s m
o
v
e
me
n
t
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
f
e
a
t
u
r
e
s
a
n
d
2
D
,
a
n
d
3
D
f
e
a
t
u
r
e
s fr
o
m
K
i
n
e
c
t
o
n
e
c
a
mera
Ex
c
i
t
e
d
,
b
o
r
e
d
o
m,
y
a
w
n
i
n
g
,
d
r
o
w
s
i
n
e
ss
N
/
A
E
-
l
e
a
r
n
i
n
g
4.
Za
l
a
t
e
l
j
i
e
t
a
l
.
[
28
]
M
a
c
h
i
n
e
l
e
a
r
n
i
n
g
f
e
a
t
u
r
e
s
a
n
d
2
D
,
a
n
d
3
D
f
e
a
t
u
r
e
s fr
o
m K
i
n
e
c
t
o
n
e
c
a
mera
H
i
g
h
a
t
t
e
n
t
i
o
n
,
me
d
i
u
m
a
t
t
e
n
t
i
o
n
,
n
o
a
t
t
e
n
t
i
o
n
7
5
.
3
%
C
l
a
s
sr
o
o
m
5.
S
h
a
r
ma
e
t
a
l
.
[
1
5
]
C
N
N
S
t
u
d
e
n
t
’
s
b
a
s
i
c
f
a
c
i
a
l
e
x
p
r
e
ss
i
o
n
s
7
0
%
E
-
l
e
a
r
n
i
n
g
6.
C
h
e
n
e
t
a
l
.
[
3
4
]
Ey
e
t
r
a
c
k
i
n
g
w
i
t
h
a
r
u
l
e
-
b
a
s
e
d
s
y
st
e
m
H
a
p
p
y
f
a
c
e
,
n
e
u
t
r
a
l
f
a
c
e
N
/
A
C
l
a
s
sr
o
o
m
7.
S
h
a
r
ma
e
t
a
l
.
[
1
5
]
M
u
l
t
i
m
o
d
a
l
(
F
E
R
,
E
y
e
T
r
a
c
k
i
n
g
,
B
o
d
y
La
n
g
u
a
g
e
)
Emo
t
i
o
n
s
8
8
%
E
-
l
e
a
r
n
i
n
g
8.
M
u
k
h
o
p
a
d
h
y
a
y
e
t
a
l
.
[
35
]
C
N
N
F
ER
2
0
1
3
e
m
o
t
i
o
n
s
62
E
-
l
e
a
r
n
i
n
g
9.
B
h
a
r
d
w
a
j
e
t
a
l
.
[
6
]
D
e
e
p
l
e
a
r
n
i
n
g
A
n
g
r
y
,
d
i
sg
u
s
t
,
f
e
a
r
,
h
a
p
p
y
,
sad
,
s
u
r
p
r
i
s
e
a
n
d
n
e
u
t
r
a
l
9
3
.
6
%
E
-
l
e
a
r
n
i
n
g
1
0
.
Th
o
mas
a
n
d
Jay
a
g
o
p
i
[
36]
P
o
se,
G
a
z
e
,
A
U
s
En
g
a
g
e
d
,
d
i
st
r
a
c
t
e
d
8
9
%
C
l
a
s
sr
o
o
m
1
1
.
A
sh
w
i
n
e
t
a
l
.
[
2
2]
C
N
N
-
b
a
s
e
d
mo
d
e
l
t
o
a
n
a
l
y
s
e
n
o
n
-
v
e
r
b
a
l
c
u
e
s
-
f
a
c
e
,
h
a
n
d
g
e
st
u
r
e
s,
a
n
d
b
o
d
y
p
o
s
t
u
r
e
s
N
o
t
e
n
g
a
g
e
d
a
t
a
l
l
,
n
o
m
i
n
a
l
l
y
e
n
g
a
g
e
d
,
e
n
g
a
g
e
d
i
n
t
a
s
k
,
v
e
r
y
e
n
g
a
g
e
d
.
7
1
%
La
r
g
e
c
l
a
ssr
o
o
m
1
2
.
Zh
e
n
g
e
t
a
l
.
[
37
]
F
a
st
e
r
R
-
C
N
N
S
t
u
d
e
n
t
b
e
h
a
v
i
o
r
s
(
h
a
n
d
r
i
si
n
g
,
st
a
n
d
i
n
g
,
sl
e
e
p
i
n
g
)
5
7
.
6
(
mA
P
)
C
l
a
s
sr
o
o
m
1
3
.
G
u
p
t
a
e
t
a
l
.
[
3
8
]
M
u
l
t
i
m
o
d
a
l
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
A
n
g
e
r
,
f
e
a
r
,
h
a
p
p
i
n
e
s
s,
sad
n
e
ss
,
s
u
r
p
r
i
s
e
9
1
.
6
E
-
l
e
a
r
n
i
n
g
1
4
.
A
i
e
t
a
l
.
[
3
9
]
D
e
e
p
e
n
g
a
g
e
me
n
t
r
e
c
o
g
n
i
t
i
o
n
n
e
t
w
o
r
k
Emo
t
i
o
n
s
b
a
se
d
o
n
t
h
e
D
a
i
se
e
d
a
t
a
se
t
6
0
%
E
-
l
e
a
r
n
i
n
g
1
5
.
G
u
p
t
a
e
t
a
l
.
[
4
0
]
C
N
N
b
a
se
d
o
n
A
l
e
x
N
e
t
a
r
c
h
i
t
e
c
t
u
r
e
En
g
a
g
e
d
,
d
i
se
n
g
a
g
e
d
8
9
.
6
0
La
r
g
e
c
l
a
ssr
o
o
m
4.
M
UL
T
I
-
SO
URC
E
ACAD
E
M
I
C
AF
F
E
CT
I
VE
E
NG
A
G
E
M
E
NT
DA
T
A
SE
T
-
I
N
T
E
G
RA
T
I
O
N
A
ND
ANA
L
YS
I
S
T
h
er
e
’
s
a
s
ig
n
if
ican
t n
ee
d
f
o
r
ex
ten
s
iv
e
d
atasets
ca
p
tu
r
in
g
af
f
ec
tiv
e
s
tates r
elate
d
to
s
tu
d
en
t le
ar
n
in
g
,
m
ain
ly
b
ec
au
s
e
s
tu
d
ies
f
o
cu
s
in
g
s
p
ec
if
ically
o
n
th
ese
s
tates
ar
e
q
u
ite
r
ar
e.
Mo
s
t
r
esear
ch
t
en
d
s
to
co
n
ce
n
tr
ate
o
n
g
en
e
r
al
em
o
tio
n
s
,
o
v
er
l
o
o
k
in
g
th
e
s
p
ec
if
ic
em
o
tio
n
al
s
tates
co
n
n
ec
ted
to
h
o
w
s
tu
d
en
ts
lear
n
.
R
ec
o
g
n
izin
g
th
is
g
ap
,
esp
ec
ial
ly
in
n
eg
ativ
e
em
o
tio
n
s
,
r
esear
ch
er
s
n
o
ticed
th
e
n
ec
ess
ity
to
co
n
s
o
lid
at
e
m
u
ltip
le
d
atasets
.
T
h
e
s
ca
r
city
o
f
d
atasets
,
p
ar
t
icu
lar
ly
th
o
s
e
h
ig
h
lig
h
tin
g
n
e
g
ativ
e
em
o
tio
n
s
,
led
to
th
e
m
er
g
in
g
o
f
v
a
r
io
u
s
d
atasets
.
T
h
is
co
m
b
in
atio
n
allo
ws
u
s
to
in
clu
d
e
d
iv
e
r
s
e
cu
lt
u
r
al
v
iewp
o
i
n
ts
as
th
ese
d
atasets
o
r
ig
in
ate
f
r
o
m
d
if
f
er
en
t
p
ar
ts
o
f
th
e
wo
r
ld
.
H
o
wev
er
,
it
’
s
im
p
o
r
tan
t
t
o
n
o
te
th
at
t
h
is
co
m
p
ilatio
n
c
o
n
tain
s
b
o
t
h
s
p
o
n
tan
eo
u
s
an
d
ac
ted
em
o
tio
n
s
,
ac
h
iev
i
n
g
a
b
alan
ce
b
etwe
en
th
ese
two
asp
ec
ts
.
T
h
is
eq
u
ilib
r
iu
m
ac
k
n
o
wled
g
es
a
wid
e
r
an
g
e
o
f
em
o
tio
n
al
ex
p
r
ess
io
n
s
,
p
r
o
v
id
in
g
a
m
o
r
e
co
m
p
r
e
h
en
s
iv
e
u
n
d
er
s
tan
d
in
g
o
f
h
o
w
em
o
tio
n
s
r
elate
to
s
tu
d
en
ts
’
lear
n
in
g
e
x
p
er
ien
ce
s
.
T
h
e
MA
AE
D
d
ataset
cr
ea
tio
n
p
r
o
ce
s
s
in
v
o
lv
e
d
th
e
co
n
s
id
er
atio
n
o
f
f
i
v
e
p
u
b
licly
av
ailab
l
e
d
atasets
-
YawDD
[
4
1
]
,
Daisee
[
4
2
]
,
m
a
n
y
f
ac
es
o
f
c
o
n
f
u
s
io
n
in
th
e
w
ild
d
ataset
(
MFC
-
W
ild
)
[
4
3
]
,
FER
2
0
1
3
[
4
4
]
an
d
B
AUM
-
1
[
4
5
]
.
Alth
o
u
g
h
th
e
MA
AE
D
d
atasets
in
clu
d
ed
t
h
e
p
o
s
itiv
e
em
o
tio
n
“
C
o
n
ce
n
tr
atio
n
,
”
th
is
s
tu
d
y
p
r
im
ar
ily
f
o
cu
s
ed
o
n
n
e
g
ativ
e
em
o
tio
n
s
lik
e
“
Yaw
n
in
g
”
,
“
B
o
r
ed
o
m
”
,
“
Fru
s
tr
atio
n
”
,
an
d
“
C
o
n
f
u
s
io
n
”
.
T
h
e
s
tu
d
y
d
i
d
n
o
t
ac
tiv
ely
in
c
o
r
p
o
r
ate
o
r
an
aly
ze
th
e
p
o
s
itiv
e
em
o
tio
n
“
C
o
n
ce
n
tr
atio
n
”
with
in
i
ts
s
p
ec
if
ic
co
n
tex
t.
Mo
n
ito
r
in
g
n
e
g
ativ
e
em
o
tio
n
s
in
s
tu
d
en
t
’
s
af
f
ec
tiv
e
s
tates
is
ess
en
tial
f
o
r
th
e
wel
l
-
b
ein
g
an
d
m
en
tal
h
ea
lth
o
f
s
tu
d
en
ts
.
T
h
ese
em
o
tio
n
s
ca
n
s
ig
n
if
ican
tly
im
p
ac
t
th
eir
o
v
er
all
well
-
b
ein
g
an
d
h
in
d
e
r
ef
f
ec
tiv
e
lear
n
in
g
.
B
y
m
o
n
ito
r
in
g
n
e
g
ativ
e
em
o
tio
n
s
,
ed
u
c
ato
r
s
ca
n
id
en
ti
f
y
s
tu
d
en
ts
in
em
o
tio
n
al
d
i
s
tr
ess
an
d
p
r
o
v
id
e
ap
p
r
o
p
r
iate
s
u
p
p
o
r
t
an
d
in
ter
v
en
tio
n
s
.
I
t
also
en
ab
les
p
er
s
o
n
alize
d
ass
is
tan
ce
,
ea
r
ly
in
ter
v
en
tio
n
,
a
n
d
th
e
cr
ea
tio
n
o
f
a
p
o
s
itiv
e
lear
n
in
g
en
v
ir
o
n
m
en
t
th
at
p
r
o
m
o
tes
b
o
th
ac
a
d
em
ic
s
u
cc
ess
an
d
o
v
er
all
well
-
b
ein
g
.
Fo
r
d
etailed
s
p
ec
if
ic
s
ab
o
u
t
t
h
e
d
atasets
u
tili
ze
d
in
th
i
s
s
tu
d
y
,
in
clu
d
in
g
s
am
p
le
s
izes,
e
m
o
tio
n
class
es,
an
d
o
th
er
p
e
r
tin
en
t d
etails,
p
lease
r
ef
er
to
T
a
b
le
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
41
,
No
.
3
,
Ma
r
ch
20
2
6
:
9
7
7
-
999
984
T
ab
le
2
.
A
co
m
p
ar
is
o
n
o
f
em
o
tio
n
r
ec
o
g
n
itio
n
d
atasets
f
o
r
MA
AE
D
d
ataset
co
n
s
tr
u
ctio
n
D
a
t
a
s
e
t
N
u
mb
e
r
o
f
sam
p
l
e
s
Emo
t
i
o
n
c
l
a
sse
s
O
t
h
e
r
d
e
t
a
i
l
s
Y
a
w
D
D
3
5
1
v
i
d
e
o
c
l
i
p
s
N
o
r
mal
,
t
a
l
k
i
n
g
/
s
i
n
g
i
n
g
,
a
n
d
y
a
w
n
i
n
g
V
i
d
e
o
s
o
f
p
e
o
p
l
e
y
a
w
n
i
n
g
i
n
n
a
t
u
r
a
l
set
t
i
n
g
s
.
R
e
q
u
i
r
e
s m
a
n
u
a
l
a
n
n
o
t
a
t
i
o
n
o
f
y
a
w
n
i
n
g
i
n
st
a
n
c
e
s w
i
t
h
i
n
v
i
d
e
o
s
.
D
A
i
S
EE
9
,
0
6
8
v
i
d
e
o
sn
i
p
p
e
t
s
B
o
r
e
d
o
m,
c
o
n
f
u
si
o
n
,
e
n
g
a
g
e
me
n
t
,
f
r
u
st
r
a
t
i
o
n
F
o
u
r
l
e
v
e
l
s
o
f
l
a
b
e
l
s
f
o
r
e
a
c
h
a
f
f
e
c
t
i
v
e
st
a
t
e
.
M
F
C
-
w
i
l
d
1
,
0
0
0
v
i
d
e
o
c
l
i
p
s
C
o
n
f
u
s
i
o
n
,
a
n
g
e
r
,
d
i
s
g
u
st
F
a
c
i
a
l
e
x
p
r
e
ss
i
o
n
s
w
e
r
e
c
o
l
l
e
c
t
e
d
f
r
o
m
Y
o
u
T
u
b
e
a
n
d
G
i
p
h
y
.
E
n
s
u
r
e
s
r
e
p
r
e
s
e
n
t
a
t
i
o
n
o
f
d
i
f
f
e
r
e
n
t
e
t
h
n
i
c
g
r
o
u
p
s.
F
ER
2
0
1
3
3
5
,
8
8
7
i
ma
g
e
s
A
n
g
e
r
,
d
i
s
g
u
s
t
,
f
e
a
r
,
h
a
p
p
i
n
e
ss
,
sad
n
e
ss
,
s
u
r
p
r
i
s
e
,
n
e
u
t
r
a
l
F
a
c
i
a
l
e
x
p
r
e
ss
i
o
n
mo
n
i
t
o
r
i
n
g
i
n
r
e
a
l
-
w
o
r
l
d
v
a
r
i
a
b
i
l
i
t
y
.
B
A
U
M
-
1
1
,
5
1
9
v
i
d
e
o
s
H
a
p
p
i
n
e
ss
,
a
n
g
e
r
,
s
a
d
n
e
ss
,
d
i
s
g
u
st
,
f
e
a
r
,
su
r
p
r
i
s
e
,
b
o
r
e
d
o
m
,
c
o
n
t
e
mp
t
,
c
o
n
f
u
s
i
o
n
,
c
o
n
c
e
n
t
r
a
t
i
o
n
,
c
u
r
i
o
si
t
y
,
c
o
m
p
l
a
i
n
t
A
u
d
i
o
-
v
i
s
u
a
l
a
f
f
e
c
t
a
n
d
m
e
n
t
a
l
s
t
a
t
e
r
e
c
o
g
n
i
t
i
o
n
.
4
.
1
.
A
uto
m
a
t
ic
f
ra
m
e
ex
t
r
a
c
t
io
n a
nd
s
elec
t
io
n
I
n
b
u
ild
i
n
g
th
e
MA
AE
D
,
f
r
am
e
ex
tr
ac
tio
n
p
lay
s
a
cr
u
cial
r
o
le
in
ca
p
tu
r
i
n
g
r
ele
v
an
t
f
ac
ial
ex
p
r
ess
io
n
s
.
A
co
m
b
in
atio
n
o
f
f
iv
e
p
u
b
licly
av
ailab
le
d
atas
ets
was
s
elec
ted
to
p
r
ed
ict
s
tu
d
en
ts
’
en
g
ag
em
e
n
t
lev
els
in
ac
ad
em
ic
e
n
v
ir
o
n
m
e
n
ts
,
f
o
cu
s
in
g
o
n
n
eg
ativ
e
em
o
tio
n
s
.
T
h
e
B
AUM
-
1
,
Daisee,
Yaw
DD
an
d
MFC
d
atasets
p
r
o
v
id
ed
v
id
eo
s
co
n
tain
in
g
d
if
f
er
en
t
em
o
tio
n
s
r
el
ev
an
t
to
s
tu
d
en
ts
’
lear
n
in
g
a
f
f
ec
tiv
e
s
tates.
T
h
e
in
itial
s
tep
in
v
o
lv
ed
m
an
u
ally
s
elec
tin
g
p
ea
k
f
r
am
es
f
r
o
m
ea
ch
em
o
tio
n
ca
teg
o
r
y
with
in
th
e
d
atasets
.
T
h
is
en
s
u
r
ed
th
at
ea
c
h
f
r
am
e
ac
cu
r
ately
r
ep
r
esen
ted
a
s
p
ec
if
ic
af
f
ec
tiv
e
s
tate.
T
h
is
m
an
u
al
s
el
ec
tio
n
p
r
o
ce
s
s
was
tim
e
-
co
n
s
u
m
in
g
,
p
r
o
m
p
tin
g
th
e
n
ee
d
f
o
r
an
au
to
m
ate
d
m
eth
o
d
to
s
tr
ea
m
lin
e
f
r
am
e
ex
tr
ac
t
io
n
(
Alg
o
r
ith
m
1
)
.
T
o
ad
d
r
ess
th
is
ch
allen
g
e,
tw
o
p
r
e
-
tr
ain
ed
m
o
d
els
wer
e
u
ti
lized
f
o
r
d
is
tin
ct
p
u
r
p
o
s
es.
T
h
e
VGG
-
f
ac
e
m
o
d
el
was
em
p
lo
y
ed
f
o
r
f
ea
tu
r
e
e
x
tr
ac
tio
n
.
Sp
ec
if
ically
d
esig
n
e
d
f
o
r
f
ac
e
r
ec
o
g
n
itio
n
task
s
,
th
is
m
o
d
el
ex
tr
ac
ted
m
ea
n
in
g
f
u
l a
n
d
d
is
cr
im
in
ativ
e
f
ea
tu
r
es f
r
o
m
th
e
in
p
u
t f
r
a
m
e
s
.
Alg
o
r
ith
m
1
.
Fra
m
e
ex
tr
ac
tio
n
Input:
•
Video file path
•
VGGFace model
•
VGG16 pretrined emotion recognition model
•
Number of frames to select
num_frames_to_select
•
Output folder path
Output:
Selected frames are saved in the specified output folder
Begin
1.
Initialize peak_emotion_confidence = 0.0, frame_count = 0
2.
Open video capture object with video file path
3.
WHILE video i
s open:
a. Read frame
b. Resize frame to (224, 224)
c. Convert frame to RGB
d. Expand frame dimensions for model input
e. Predict facial features using VGGFace
f. Predict emotion using the custom model
g. IF emotion_confidence > peak_emotion_confidence THEN
-
Update peak_emotion_confidence and peak_emotion_frame_index
h. Increment frame_count
4.
Release video capture object
5.
Determine selected_fram
e_indices using np.linspace (0, frame_count
-
1,
num_frames_to_select)
6.
Open video capture object again
7.
FOR each index in selected_frame_indices:
a. Set video capture to the specific frame index
b. Read and resize frame
c. Save frame to output folder
8.
Release video capture object
9.
Return success message
End
T
h
e
f
ir
s
t
m
o
d
el
em
p
l
o
y
ed
w
as
th
e
p
r
e
-
tr
ain
ed
VGG
-
f
ac
e
m
o
d
el
[
4
6
]
.
T
h
e
VGG
-
f
ac
e
m
o
d
el
is
a
s
p
ec
ialized
C
NN
ar
ch
itectu
r
e
p
r
im
ar
ily
d
esig
n
e
d
f
o
r
f
ac
e
r
ec
o
g
n
itio
n
ap
p
licatio
n
s
.
B
u
ilt
u
p
o
n
th
e
VGG
ar
ch
i
tectu
r
e,
th
is
m
o
d
el
c
o
m
p
r
is
es
1
6
co
n
v
o
lu
tio
n
al
lay
e
r
s
f
o
llo
wed
b
y
f
u
lly
co
n
n
ec
te
d
lay
er
s
.
I
t
u
tili
ze
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
Leve
r
a
g
in
g
C
N
N
to
a
n
a
lyze fa
cia
l e
xp
r
ess
io
n
s
fo
r
a
ca
d
emic
en
g
a
g
eme
n
t
mo
n
ito
r
in
g
w
ith
…
(
N
o
o
r
a
C
.
T.
)
985
s
m
aller
-
s
ized
co
n
v
o
lu
tio
n
al
f
i
lter
s
(
3
x
3
)
c
o
n
s
is
ten
tly
ac
r
o
s
s
th
e
n
etwo
r
k
,
en
ab
lin
g
a
d
ee
p
ar
ch
itectu
r
e
wh
ile
m
ain
tain
in
g
a
m
a
n
ag
ea
b
le
n
u
m
b
er
o
f
p
ar
am
eter
s
.
T
r
ain
ed
o
n
an
ex
ten
s
iv
e
d
ataset
o
f
f
ac
ia
l im
ag
es,
VGG
-
f
ac
e
f
o
cu
s
es
o
n
lear
n
i
n
g
co
m
p
r
eh
e
n
s
iv
e
f
ac
ial
f
ea
tu
r
es
cr
itical
f
o
r
ac
cu
r
ate
f
ac
e
id
en
tific
atio
n
an
d
d
is
tin
ctio
n
.
I
ts
p
r
e
-
tr
ain
ed
n
at
u
r
e
a
n
d
tr
an
s
f
er
lear
n
in
g
ca
p
ab
ilit
ies
allo
w
f
o
r
ef
f
icien
t
ad
a
p
tat
io
n
to
s
p
ec
if
ic
f
ac
e
-
r
elate
d
task
s
with
lim
i
ted
tr
ain
in
g
d
a
ta,
s
p
ee
d
in
g
u
p
tr
ain
in
g
p
r
o
ce
s
s
es
an
d
en
h
an
cin
g
p
er
f
o
r
m
a
n
ce
.
Kn
o
wn
f
o
r
its
d
ep
th
an
d
h
ier
a
r
ch
ical
f
ea
tu
r
e
ex
tr
ac
tio
n
,
t
h
e
m
o
d
el
ex
tr
a
cts
in
tr
icate
f
ac
ial
d
etails
at
v
ar
y
in
g
ab
s
tr
ac
tio
n
lev
els,
in
c
lu
d
in
g
s
h
a
p
es,
tex
t
u
r
es,
an
d
n
u
an
ce
d
f
ac
ial
attr
ib
u
tes.
R
en
o
wn
ed
f
o
r
r
o
b
u
s
t
n
ess
,
it
s
h
o
wca
s
e
s
r
eliab
ilit
y
in
r
ec
o
g
n
izin
g
f
ac
es
am
id
s
t
d
iv
er
s
e
co
n
d
itio
n
s
s
u
ch
as
d
if
f
er
en
t
e
x
p
r
ess
io
n
s
,
p
o
s
es,
lig
h
tin
g
v
ar
iatio
n
s
,
an
d
b
ac
k
g
r
o
u
n
d
s
.
B
y
lev
er
ag
in
g
th
e
VGG
-
f
ac
e
m
o
d
el,
m
ea
n
in
g
f
u
l
a
n
d
d
is
cr
im
in
ativ
e
f
ea
t
u
r
es
wer
e
ex
tr
ac
ted
f
r
o
m
in
p
u
t f
r
a
m
es o
f
v
id
e
o
s
,
en
ab
lin
g
f
u
r
t
h
e
r
an
aly
s
is
an
d
p
r
o
ce
s
s
in
g
.
T
h
e
s
ec
o
n
d
m
o
d
el
u
s
ed
wa
s
a
s
ep
ar
ate
p
r
e
-
tr
ain
ed
em
o
tio
n
r
ec
o
g
n
itio
n
m
o
d
el
wit
h
VGG
1
6
ar
ch
itectu
r
e.
T
h
is
m
o
d
el
is
tr
a
in
ed
s
p
ec
if
ically
to
r
ec
o
g
n
ize
em
o
tio
n
s
in
im
a
g
es
an
d
ca
n
c
lass
if
y
im
ag
es
in
to
d
if
f
er
en
t
em
o
tio
n
ca
teg
o
r
ies.
T
h
e
f
ea
tu
r
es
ex
tr
ac
te
d
f
r
o
m
th
e
VGG
-
f
ac
e
m
o
d
el
wer
e
in
p
u
tted
in
to
th
is
em
o
tio
n
r
ec
o
g
n
itio
n
m
o
d
el.
T
o
ac
co
m
p
lis
h
th
is
,
a
m
o
d
if
ied
m
o
d
el
was
cr
ea
ted
b
y
co
m
b
i
n
in
g
th
e
VGG
-
f
ac
e
m
o
d
el
with
a
d
en
s
e
lay
er
r
esp
o
n
s
ib
le
f
o
r
e
m
o
tio
n
class
if
icatio
n
.
T
h
is
m
o
d
if
ie
d
m
o
d
el
p
r
ed
icts
th
e
em
o
tio
n
class
o
f
an
in
p
u
t
f
r
am
e
b
y
lev
er
ag
in
g
th
e
e
x
tr
ac
ted
f
ea
tu
r
es
f
r
o
m
th
e
VGG
-
f
ac
e
m
o
d
el.
Seq
u
en
tially
u
tili
zin
g
b
o
th
m
o
d
els
b
en
ef
its
f
r
o
m
th
e
VGG
-
f
ac
e
m
o
d
el’
s
ca
p
ab
ilit
y
to
ex
tr
ac
t
r
ich
f
ac
ial
f
ea
tu
r
es,
wh
ich
ar
e
h
ig
h
ly
r
elev
an
t
f
o
r
ca
p
tu
r
in
g
m
ea
n
i
n
g
f
u
l
p
atter
n
s
r
elate
d
t
o
em
o
tio
n
s
.
T
h
e
s
ep
ar
ate
p
r
e
-
t
r
ain
ed
e
m
o
tio
n
r
ec
o
g
n
itio
n
m
o
d
el
p
er
f
o
r
m
s
class
if
icatio
n
o
n
th
ese
e
x
tr
ac
ted
f
ea
tu
r
es
,
a
s
s
ig
n
in
g
em
o
tio
n
lab
els
to
th
e
in
p
u
t
f
r
am
es.
T
h
is
m
u
lti
-
s
tep
ap
p
r
o
ac
h
f
ac
il
itates
a
m
o
r
e
s
p
ec
ialized
an
d
a
cc
u
r
ate
em
o
tio
n
r
ec
o
g
n
itio
n
p
r
o
ce
s
s
th
an
u
s
in
g
a
s
in
g
le
m
o
d
el
alo
n
e
.
T
h
is
ap
p
r
o
ac
h
a
d
d
r
ess
es
th
e
lim
itatio
n
s
o
f
m
an
u
al
f
r
am
e
s
elec
tio
n
,
m
ak
in
g
it
a
v
alu
ab
le
co
n
tr
ib
u
tio
n
to
th
e
f
ield
o
f
em
o
tio
n
r
ec
o
g
n
itio
n
.
Utilizin
g
co
m
p
u
ter
v
is
io
n
tec
h
n
iq
u
es
v
ia
t
h
e
Op
e
n
C
V
lib
r
ar
y
,
s
y
s
tem
atica
lly
iter
ates
th
r
o
u
g
h
v
i
d
eo
s
,
r
esizin
g
f
r
am
es
to
m
atch
th
e
in
p
u
t
s
p
ec
if
icatio
n
s
o
f
a
p
r
e
-
tr
ain
ed
VGG
-
f
ac
e
m
o
d
el.
Su
b
s
eq
u
en
tly
,
it
em
p
lo
y
s
th
is
m
o
d
el
to
p
r
ed
ict
em
o
tio
n
s
with
in
ea
ch
f
r
am
e,
id
en
tify
in
g
th
e
h
ig
h
est
co
n
f
id
e
n
ce
lev
el
ass
o
ciate
d
w
ith
ea
ch
en
g
a
g
em
en
t
ca
teg
o
r
y
b
y
th
e
VGG
1
6
p
r
et
r
ain
ed
m
o
d
e
l.
T
h
e
alg
o
r
ith
m
p
r
ec
is
ely
tr
ac
k
s
th
e
f
r
a
m
e
in
d
e
x
s
h
o
wca
s
in
g
th
e
m
o
s
t
in
ten
s
e
e
m
o
tio
n
al
r
esp
o
n
s
e,
ex
tr
ac
tin
g
a
s
p
ec
if
ic
n
u
m
b
e
r
o
f
f
r
am
es
c
en
te
r
ed
a
r
o
u
n
d
th
is
h
ig
h
lig
h
ted
p
ea
k
.
T
h
ese
s
elec
ted
f
r
am
es,
e
n
ca
p
s
u
latin
g
t
h
e
p
in
n
ac
le
o
f
em
o
tio
n
al
r
esp
o
n
s
e
,
ar
e
th
e
n
s
av
ed
t
o
an
o
u
tp
u
t
d
ir
ec
to
r
y
f
o
r
f
u
r
th
e
r
in
v
esti
g
atio
n
o
r
an
aly
s
is
.
T
h
is
ex
tr
ac
tio
n
p
r
o
ce
s
s
n
o
t
o
n
ly
ca
p
tu
r
es
cr
u
cial
m
o
m
en
ts
r
ef
lectin
g
h
ei
g
h
ten
ed
en
g
a
g
em
en
t
ca
teg
o
r
y
with
in
v
id
e
o
s
b
u
t
also
s
tr
ea
m
li
n
es
th
e
s
u
b
s
eq
u
en
t
ex
p
lo
r
atio
n
an
d
i
n
ter
p
r
etatio
n
o
f
th
ese
em
o
tio
n
all
y
s
ig
n
if
ican
t
f
r
am
es
f
o
r
p
o
ten
tial
d
ee
p
er
in
s
ig
h
ts
o
r
d
iag
n
o
s
tic
p
u
r
p
o
s
es.
T
h
e
d
ataset
h
as
f
u
r
th
er
en
lar
g
ed
b
y
in
co
r
p
o
r
atin
g
FER
1
3
d
ataset.
T
h
is
b
r
in
g
s
t
h
e
to
tal
n
u
m
b
er
o
f
im
ag
es
to
1
6
,
9
2
4
f
o
r
tr
ai
n
in
g
,
4
,
7
2
5
f
o
r
test
in
g
,
a
n
d
3
,
6
4
1
f
o
r
v
alid
atio
n
.
T
o
e
n
s
u
r
e
th
e
ac
cu
r
ac
y
o
f
th
e
d
ataset,
a
m
an
u
al
an
aly
s
is
w
as
co
n
d
u
cted
to
v
er
if
y
th
e
co
r
r
ec
tn
ess
o
f
th
e
au
to
m
atic
f
r
a
m
e
ex
tr
ac
tio
n
an
d
p
r
ep
r
o
ce
s
s
in
g
s
tep
s
.
T
h
is
an
al
y
s
is
h
elp
ed
to
e
n
s
u
r
e
th
at
th
e
ex
tr
ac
ted
f
r
am
es
an
d
th
e
a
p
p
lied
p
r
ep
r
o
ce
s
s
in
g
tech
n
iq
u
es r
esu
lted
in
ac
c
u
r
at
e
an
d
r
eliab
le
d
ata
f
o
r
FER.
T
h
e
co
m
p
a
r
is
o
n
T
ab
le
3
p
r
o
v
id
es
an
o
v
er
v
iew
o
f
v
a
r
io
u
s
r
esear
ch
s
tu
d
ies
co
n
ce
n
tr
atin
g
o
n
f
ac
e
d
etec
tio
n
in
s
in
g
le
an
d
m
u
ltip
le
f
r
am
es,
ac
a
d
em
ic
af
f
ec
tiv
e
s
tates,
tech
n
iq
u
es
em
p
l
o
y
ed
(
s
u
ch
as
C
NN,
SVM,
an
d
KNN)
,
d
atasets
u
tili
ze
d
(
lik
e
C
K+
,
FER2
0
1
3
,
an
d
D
Ai
SEE
d
ataset)
,
an
d
f
r
a
m
e
e
x
tr
ac
tio
n
m
eth
o
d
s
(
s
u
ch
as
r
eg
u
lar
in
te
r
v
als,
f
ix
ed
-
r
ate
ex
tr
ac
tio
n
,
an
d
a
u
to
m
ated
p
ea
k
f
r
am
e
s
elec
t
io
n
)
.
E
ac
h
s
tu
d
y
d
em
o
n
s
tr
ates
d
is
tin
ct
f
o
cu
s
es,
tech
n
iq
u
es,
an
d
d
atasets
,
p
r
esen
tin
g
d
iv
er
s
e
ap
p
r
o
ac
h
es
to
d
etec
tin
g
f
ac
ial
ex
p
r
es
s
io
n
s
an
d
r
ec
o
g
n
izin
g
a
f
f
ec
tiv
e
s
tates
in
ac
ad
em
ic
co
n
tex
ts
.
No
tab
ly
,
th
e
p
r
o
p
o
s
ed
m
o
d
el
s
tan
d
s
o
u
t b
y
u
tili
zin
g
a
cu
s
to
m
d
ataset
(
M
AAE
D
)
an
d
im
p
lem
en
ti
n
g
au
to
m
ated
p
ea
k
f
r
am
e
s
elec
tio
n
m
eth
o
d
s
f
o
r
f
r
a
m
e
ex
tr
ac
tio
n
.
4
.
2
.
P
re
pro
ce
s
s
ing
Pre
p
r
o
ce
s
s
in
g
p
lay
s
a
p
iv
o
tal
r
o
le
in
r
ea
d
y
in
g
d
ata
f
o
r
a
n
aly
s
is
,
p
ar
ticu
lar
ly
in
task
s
li
k
e
f
ac
ial
an
aly
s
is
an
d
em
o
tio
n
r
ec
o
g
n
itio
n
.
Sp
ec
if
ically
tailo
r
ed
f
o
r
f
ac
ial
ex
p
r
ess
io
n
d
atasets
ex
tr
ac
ted
f
r
o
m
v
id
e
o
s
,
p
r
ep
r
o
ce
s
s
in
g
en
co
m
p
ass
es
s
ev
er
al
cr
u
cial
s
tep
s
th
at
r
ef
in
e
an
d
s
tan
d
ar
d
ize
th
e
in
p
u
t
f
r
am
es.
R
esizin
g
th
e
f
r
am
es
to
a
s
p
ec
if
ic
d
im
en
s
io
n
en
s
u
r
es
u
n
i
f
o
r
m
ity
,
f
ac
i
litatin
g
co
n
s
is
ten
t
an
aly
s
is
a
cr
o
s
s
th
e
d
ataset.
C
o
n
v
er
s
io
n
to
g
r
ay
s
ca
le
s
im
p
lifie
s
th
e
d
ata
wh
ile
p
r
eser
v
in
g
ess
en
tial
f
ac
ia
l
f
ea
tu
r
es,
r
ed
u
cin
g
co
m
p
u
tatio
n
al
co
m
p
lex
ity
with
o
u
t
c
o
m
p
r
o
m
is
in
g
cr
itical
v
is
u
al
in
f
o
r
m
ati
o
n
.
Ad
d
itio
n
ally
,
n
o
r
m
alizin
g
p
ix
el
v
alu
es
to
a
s
tan
d
ar
d
ized
r
a
n
g
e
o
p
tim
izes
d
ata
f
o
r
d
ee
p
lear
n
i
n
g
m
o
d
els,
en
h
an
cin
g
c
o
n
v
e
r
g
en
ce
d
u
r
in
g
t
r
ain
in
g
a
n
d
en
ab
lin
g
m
o
d
els
to
b
etter
g
en
er
alize
ac
r
o
s
s
d
if
f
er
e
n
t
f
a
cial
ex
p
r
ess
io
n
s
an
d
in
d
i
v
id
u
als.
Ov
er
all,
th
ese
p
r
ep
r
o
ce
s
s
in
g
tech
n
iq
u
es
ar
e
e
s
s
en
tial
f
o
r
r
e
f
in
in
g
r
aw
d
ata,
o
p
tim
izin
g
its
s
u
itab
ilit
y
f
o
r
s
u
b
s
eq
u
en
t
an
aly
s
is
,
an
d
b
o
ls
ter
in
g
th
e
ac
cu
r
ac
y
an
d
r
eliab
ilit
y
o
f
f
ac
ial
an
aly
s
is
an
d
em
o
tio
n
r
ec
o
g
n
itio
n
s
y
s
te
m
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
41
,
No
.
3
,
Ma
r
ch
20
2
6
:
9
7
7
-
999
986
T
ab
le
3
.
C
o
m
p
a
r
is
o
n
o
f
th
e
p
r
o
p
o
s
ed
m
o
d
el
to
p
r
ev
io
u
s
s
tu
d
ies in
ter
m
s
o
f
co
n
tr
i
b
u
tio
n
s
Li
t
e
r
a
t
u
r
e
F
a
c
e
d
e
t
e
c
t
i
o
n
i
n
a
si
n
g
l
e
f
r
a
m
e
A
c
a
d
e
mi
c
a
f
f
e
c
t
i
v
e
st
a
t
e
s
T
e
c
h
n
i
q
u
e
D
a
t
a
s
e
t
F
r
a
me
e
x
t
r
a
c
t
i
o
n
S
i
n
g
l
e
f
a
c
e
M
u
l
t
i
p
l
e
f
a
c
e
s
W
a
n
g
e
t
a
l
.
[4
7
]
✓
X
C
N
N
C
K
+
a
n
d
F
E
R
2
0
1
3
_
_
_
Y
u
a
n
[
4
8
]
✓
X
✓
M
T
C
N
N
C
l
a
s
sr
o
o
m
d
a
t
a
s
e
t
+
R
A
F
-
D
B
+
mas
k
e
d
d
a
t
a
se
t
_
_
_
G
u
p
t
a
e
t
a
l
.
[
40
]
✓
X
X
R
e
sN
e
t
-
50
R
A
F
-
D
B
+
F
E
R
-
2
0
1
3
+
O
W
N
D
a
t
a
se
t
+
C
K
(
+
)
e
x
t
r
a
c
t
s fr
a
m
e
s a
t
r
e
g
u
l
a
r
i
n
t
e
r
v
a
l
s
(
e
v
e
r
y
2
0
se
c
)
W
h
i
t
e
h
i
l
l
e
t
a
l
.
[
2
6
]
✓
X
✓
B
o
o
st
(
B
F
)
,
S
V
M
,
M
L
R
(
C
ER
T)
H
B
C
U
+
U
C
f
r
a
mes
w
e
r
e
e
x
t
r
a
c
t
e
d
a
t
r
e
g
u
l
a
r
i
n
t
e
r
v
a
l
s
a
n
d
h
u
m
a
n
l
a
b
e
l
l
i
n
g
G
o
n
g
a
n
d
W
e
i
[
49
]
✓
X
X
KNN
F
A
C
S
----
P
a
b
b
a
a
n
d
K
u
m
a
r
[
2
4
]
✓
✓
✓
C
N
N
En
l
a
r
g
e
d
C
S
F
ED
e
x
t
r
a
c
t
s fr
a
m
e
s a
t
r
e
g
u
l
a
r
i
n
t
e
r
v
a
l
s
(
4
f
r
a
mes
p
e
r
sec
o
n
d
)
A
l
a
me
d
a
-
P
i
n
e
d
a
e
t
a
l
.
[
1
2
]
✓
X
X
R
G
B
-
I
3
D
N
e
t
w
o
r
k
D
A
i
S
EE
d
a
t
a
se
t
1
5
f
r
a
mes
p
e
r
v
i
d
e
o
A
i
e
t
a
l
.
[
39
]
✓
X
X
C
N
N
D
A
i
S
EE
d
a
t
a
se
t
2
0
f
r
a
mes
i
n
e
a
c
h
v
i
d
e
o
K
a
m
a
t
h
e
t
a
l
.
[
5
0
]
✓
X
X
S
V
M
D
R
I
S
h
TI
W
A
C
V
2
0
1
6
F
i
x
e
d
-
r
a
t
e
f
r
a
m
e
e
x
t
r
a
c
t
i
o
n
M
u
k
h
o
p
a
d
h
y
a
y
e
t
a
l
.
[
35
]
✓
X
X
C
N
N
F
ER
2
0
1
3
1
f
r
a
m
e
/
se
c
Th
o
mas
a
n
d
Jay
a
g
o
p
i
[
36
]
✓
✓
X
S
V
M
C
u
s
t
o
m
d
a
t
a
se
t
b
u
i
l
t
b
y
r
e
se
a
r
c
h
e
r
s
2
5
f
r
a
mes
p
e
r
sec
o
n
d
P
r
o
p
o
se
d
m
o
d
e
l
✓
✓
✓
C
N
N
C
u
s
t
o
m
d
a
t
a
se
t
b
u
i
l
t
b
y
r
e
se
a
r
c
h
e
r
s
(
M
A
A
ED
)
A
u
t
o
ma
t
e
d
p
e
a
k
f
r
a
me
se
l
e
c
t
i
o
n
4
.
2
.
1
.
F
a
ce
det
ec
t
i
o
n
T
h
e
im
p
o
r
tan
ce
o
f
ch
o
o
s
in
g
th
e
r
ig
h
t
f
ac
e
d
etec
tio
n
alg
o
r
ith
m
is
cr
u
cial
i
n
r
esear
ch
f
o
cu
s
ed
o
n
f
ac
ial
an
aly
s
is
an
d
em
o
tio
n
r
e
co
g
n
itio
n
.
C
o
n
s
id
er
in
g
th
e
co
m
p
lex
ities
o
f
th
e
r
esear
c
h
,
w
e
f
o
u
n
d
th
at
m
u
lti
-
task
ca
s
ca
d
ed
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
s
(
MT
C
NN
)
,
with
its
m
u
lti
-
s
tag
e
h
ier
ar
ch
ical
ap
p
r
o
ac
h
,
o
u
tp
er
f
o
r
m
ed
s
ev
er
al
o
th
e
r
f
ac
e
d
etec
tio
n
alg
o
r
ith
m
s
.
I
ts
ab
ilit
y
to
ac
cu
r
ately
d
etec
t
f
a
ce
s
ac
r
o
s
s
v
ar
io
u
s
s
ca
les,
o
r
ien
tatio
n
s
,
an
d
ch
allen
g
in
g
co
n
d
itio
n
s
alig
n
s
p
er
f
ec
tly
with
o
u
r
p
r
o
ject
’
s
r
eq
u
ir
em
e
n
ts
.
B
y
l
ev
er
ag
in
g
s
tag
es
lik
e
p
r
o
p
o
s
al
n
etwo
r
k
(
P
-
Net)
,
r
ef
in
e
n
etwo
r
k
(
R
-
Net)
,
an
d
o
u
tp
u
t
n
etwo
r
k
(
O
-
Net
)
,
MT
C
NN
ex
ce
ls
in
id
en
tify
in
g
a
h
ig
h
er
n
u
m
b
er
o
f
f
ac
es
with
in
a
s
in
g
le
f
r
am
e,
c
r
u
cial
f
o
r
o
u
r
g
o
al
o
f
ac
cu
r
ately
r
ec
o
g
n
izin
g
an
d
an
aly
zin
g
f
ac
ial
ex
p
r
ess
io
n
s
.
T
h
is
ad
ap
tab
ilit
y
an
d
r
o
b
u
s
tn
ess
o
f
MT
C
N
N
p
lay
a
p
iv
o
tal
r
o
le
in
en
s
u
r
in
g
th
e
s
u
cc
ess
an
d
ef
f
ec
tiv
e
n
e
s
s
o
f
o
u
r
f
ac
ial
an
aly
s
is
an
d
em
o
tio
n
r
ec
o
g
n
itio
n
task
s
.
T
h
e
MT
C
NN
alg
o
r
ith
m
d
etec
ts
f
ac
es
u
s
in
g
ca
s
ca
d
ed
C
NN
s
.
I
t
p
r
o
v
id
es
b
o
u
n
d
in
g
b
o
x
c
o
o
r
d
in
ates
an
d
lan
d
m
a
r
k
s
f
o
r
ea
ch
d
etec
t
ed
f
ac
e,
allo
win
g
c
r
o
p
p
in
g
a
n
d
r
esizin
g
to
2
2
4
×2
2
4
p
ix
els
f
o
r
f
o
c
u
s
ed
an
aly
s
is
.
T
h
e
MT
C
NN
alg
o
r
ith
m
f
o
r
f
ac
e
d
etec
tio
n
o
p
er
ates
in
th
r
ee
m
ajo
r
s
tag
es:
th
e
P
-
Net,
th
e
R
-
Net,
an
d
th
e
O
-
Net.
I
n
th
e
P
-
Net
s
tag
e,
th
e
co
n
v
o
l
u
tio
n
al
lay
er
is
f
o
r
m
u
la
ted
as
:
=
∑
,
∋
(
+
)
(
+
)
+
(
6
)
an
d
t
h
e
R
eL
U
ac
tiv
atio
n
f
u
n
ct
io
n
ca
n
b
e
r
ep
r
esen
ted
as:
(
)
=
(
0
,
)
(
7
)
T
h
e
m
ax
-
p
o
o
lin
g
o
p
er
atio
n
is
g
iv
en
b
y
,
=
(
:
+
,
:
+
)
(
8
)
wh
ile
th
e
So
f
tMa
x
f
u
n
ctio
n
f
o
r
class
if
icatio
n
is
;
Evaluation Warning : The document was created with Spire.PDF for Python.