I
nte
rna
t
io
na
l J
o
urna
l o
f
Rec
o
nfig
ura
ble a
nd
E
m
be
dd
e
d Sy
s
t
e
m
s
(
I
J
R
E
S)
Vo
l.
1
5
,
No
.
1
,
Ma
r
ch
20
2
6
,
p
p
.
42
~
53
I
SS
N:
2089
-
4864
,
DOI
:
1
0
.
1
1
5
9
1
/i
j
r
es
.
v
1
5
.
i
1
.
pp
42
-
53
42
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
r
es.ia
esco
r
e.
co
m
Deplo
y
m
en
t
a
nd
ev
a
lua
tion o
f
faci
a
l expres
sio
n reco
g
nition o
n
Andro
id and
Te
mi V3
in
co
n
trolled
settings
M
o
ha
m
a
d H
a
riz
Na
za
m
i
d
1
,
Ro
zit
a
J
a
ila
ni
1
,
Nur
K
ha
lid
a
h
Z
a
k
a
r
ia
1
,
Anw
a
r
P
.
P
.
Abdu
l M
a
j
ee
d
2
1
F
a
c
u
l
t
y
o
f
El
e
c
t
r
i
c
a
l
En
g
i
n
e
e
r
i
n
g
,
U
n
i
v
e
r
si
t
i
T
e
k
n
o
l
o
g
i
M
A
R
A
,
S
h
a
h
A
l
a
m,
M
a
l
a
y
si
a
2
F
a
c
u
l
t
y
o
f
En
g
i
n
e
e
r
i
n
g
a
n
d
T
e
c
h
n
o
l
o
g
y
,
S
u
n
w
a
y
U
n
i
v
e
r
si
t
y
,
B
a
n
d
a
r
S
u
n
w
a
y
,
M
a
l
a
y
si
a
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
an
2
1
,
2
0
2
5
R
ev
i
s
ed
Dec
1
1
,
2
0
2
5
A
cc
ep
ted
J
an
1
1
,
2
0
2
6
F
a
c
ial
e
x
p
re
ss
io
n
re
c
o
g
n
it
io
n
(
F
ER)
is
v
it
a
l
f
o
r
i
m
p
ro
v
in
g
h
u
m
a
n
-
ro
b
o
t
in
tera
c
ti
o
n
(HRI).
T
h
is
stu
d
y
p
re
se
n
ts
th
e
d
e
p
lo
y
m
e
n
t
a
n
d
e
v
a
lu
a
t
io
n
o
f
a
n
o
p
ti
m
ize
d
F
ER
m
o
d
e
l
o
n
a
n
d
ro
i
d
d
e
v
ice
s,
sp
e
c
i
f
ica
ll
y
tes
ted
o
n
th
e
T
e
m
i
V
3
ro
b
o
t
in
c
o
n
tro
ll
e
d
e
n
v
iro
n
m
e
n
ts.
T
ra
in
e
d
u
sin
g
F
ER+
a
n
d
CK+
d
a
tas
e
ts
a
n
d
o
p
ti
m
ize
d
w
it
h
T
e
n
so
rF
lo
w
L
it
e
(T
F
L
it
e
)
a
n
d
M
o
b
il
e
Ne
tV2
,
t
h
e
m
o
d
e
l
a
c
h
iev
e
d
a
v
a
li
d
a
ti
o
n
a
c
c
u
ra
c
y
o
f
9
2
.
3
2
%
.
I
ts
p
e
rf
o
rm
a
n
c
e
wa
s
a
ss
e
ss
e
d
o
n
th
e
T
e
m
i
V
3
r
o
b
o
t
a
n
d
a
S
a
m
su
n
g
A
5
2
sm
a
rtp
h
o
n
e
,
f
o
c
u
sin
g
o
n
C
P
U
u
sa
g
e
,
m
e
m
o
r
y
,
a
n
d
p
o
w
e
r
c
o
n
su
m
p
ti
o
n
.
Cro
ss
-
d
e
v
ice
c
o
m
p
a
ti
b
il
it
y
a
n
d
re
a
l
-
ti
m
e
p
e
rf
o
r
m
a
n
c
e
c
h
a
ll
e
n
g
e
s
w
e
re
a
d
d
re
ss
e
d
th
r
o
u
g
h
m
o
d
e
l
q
u
a
n
ti
z
a
ti
o
n
a
n
d
th
re
a
d
o
p
ti
m
iza
ti
o
n
.
Re
a
l
-
ti
m
e
t
e
stin
g
o
n
th
e
T
e
m
i
V
3
sh
o
w
e
d
a
n
o
v
e
ra
ll
a
c
c
u
ra
c
y
o
f
8
2
.
2
8
%
,
w
it
h
e
m
o
ti
o
n
-
sp
e
c
if
ic
a
c
c
u
ra
c
i
e
s
ra
n
g
in
g
f
ro
m
4
6
.
1
9
%
to
9
2
.
2
8
%
.
T
h
is
stu
d
y
o
ff
e
r
s
p
ra
c
ti
c
a
l
in
sig
h
ts
f
o
r
o
p
ti
m
izin
g
F
E
R
s
y
ste
m
s
a
c
ro
ss
a
n
d
ro
i
d
p
latf
o
rm
s,
w
it
h
p
o
ten
ti
a
l
a
p
p
l
ica
ti
o
n
s
in
e
d
u
c
a
ti
o
n
,
h
e
a
lt
h
c
a
re
,
a
n
d
c
u
sto
m
e
r
se
rv
ic
e
.
T
h
e
re
su
lt
s
su
p
p
o
rt
th
e
f
e
a
sib
il
it
y
o
f
im
p
le
m
e
n
ti
n
g
F
E
R
m
o
d
e
ls
a
s
b
a
c
k
e
n
d
s
in
a
n
d
r
o
i
d
a
p
p
li
c
a
ti
o
n
s,
e
n
a
b
li
n
g
m
o
re
in
tu
it
iv
e
a
n
d
re
sp
o
n
siv
e
HRI.
F
u
t
u
re
w
o
rk
w
il
l
f
o
c
u
s
o
n
i
m
p
ro
v
in
g
m
o
d
e
l
e
ff
icie
n
c
y
f
o
r
lo
w
e
r
-
e
n
d
d
e
v
ice
s
a
n
d
e
x
p
lo
ri
n
g
o
n
-
d
e
v
ice
lea
rn
in
g
tec
h
n
iq
u
e
s to
b
o
o
st ac
c
u
ra
c
y
in
d
iv
e
rse
re
a
l
-
w
o
rld
e
n
v
iro
n
m
e
n
ts.
K
ey
w
o
r
d
s
:
Facial
ex
p
r
ess
io
n
r
ec
o
g
n
itio
n
I
m
ag
e
p
r
o
ce
s
s
i
n
g
L
i
g
h
t
w
ei
g
h
t
m
o
d
el
R
o
b
o
t in
teg
r
atio
n
T
em
i V
3
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
:
R
o
zita
J
ailan
i
Facu
lt
y
o
f
E
lectr
ical
E
n
g
in
ee
r
in
g
,
U
n
i
v
er
s
iti T
ek
n
o
lo
g
i M
AR
A
Sh
a
h
A
la
m
,
Selan
g
o
r
,
Ma
la
y
s
i
a
E
m
ail:
r
o
zitaj
@
u
i
t
m
.
ed
u
.
m
y
1.
I
NT
RO
D
UCT
I
O
N
Hu
m
an
-
r
o
b
o
t
in
ter
ac
tio
n
(
H
R
I
)
p
lay
s
a
s
i
g
n
i
f
ica
n
t
r
o
le
in
v
ar
io
u
s
s
ec
to
r
s
,
i
n
clu
d
i
n
g
h
ea
l
t
h
ca
r
e
an
d
HR
I
.
O
n
e
n
o
tab
le
ap
p
licatio
n
is
a
s
s
i
s
ti
v
e
r
o
b
o
ts
th
a
t
u
tili
ze
ar
tif
icia
l
i
n
telli
g
e
n
ce
(
A
I
)
to
p
r
o
v
id
e
e
m
o
tio
n
al
s
u
p
p
o
r
t
in
eld
er
l
y
ca
r
e
s
ett
i
n
g
s
[
1
]
,
[
2
]
.
HR
I
h
as
ad
v
a
n
ce
d
co
n
s
id
er
ab
l
y
w
it
h
t
h
e
in
te
g
r
atio
n
of
A
I
tech
n
o
lo
g
ies.
Fo
r
in
s
ta
n
ce
,
n
atu
r
al
lan
g
u
a
g
e
p
r
o
ce
s
s
in
g
(
NL
P
)
tech
n
o
lo
g
ie
s
h
a
v
e
en
ab
led
s
ea
m
le
s
s
co
m
m
u
n
icatio
n
w
it
h
u
s
er
s
ac
r
o
s
s
d
iv
er
s
e
ac
ce
n
ts
,
lan
g
u
ag
e
s
,
a
n
d
s
p
ee
ch
p
atter
n
s
[
3
]
.
A
d
d
itio
n
all
y
,
co
m
p
u
ter
v
is
io
n
tech
n
o
lo
g
ie
s
f
ac
il
itate
th
e
r
ec
o
g
n
i
tio
n
o
f
h
u
m
a
n
g
e
s
tu
r
es
an
d
ex
p
r
es
s
io
n
s
,
allo
w
i
n
g
f
o
r
m
o
r
e
n
atu
r
al
in
ter
ac
tio
n
s
[
4
]
,
[
5
]
.
Fu
tu
r
e
d
ev
elo
p
m
en
t
s
i
n
HR
I
ar
e
ex
p
ec
ted
to
in
co
r
p
o
r
ate
m
u
lti
-
m
o
d
al
i
n
ter
ac
tio
n
ca
p
ab
ilit
ies,
co
m
b
in
i
n
g
s
p
ee
c
h
,
g
e
s
tu
r
e
s
,
f
ac
ial
e
x
p
r
ess
io
n
s
,
an
d
to
u
c
h
,
to
en
ab
le
ev
e
n
m
o
r
e
n
at
u
r
al
an
d
in
t
u
iti
v
e
in
ter
ac
tio
n
s
.
Facial
ex
p
r
es
s
io
n
r
ec
o
g
n
itio
n
(
FER)
li
g
h
t
w
ei
g
h
t
m
o
d
els
ca
n
b
e
ca
teg
o
r
ized
in
to
t
wo
class
es
:
r
eg
r
ess
io
n
m
o
d
els
a
n
d
clas
s
if
icatio
n
m
o
d
els.
T
h
is
s
t
u
d
y
f
o
cu
s
es
o
n
c
lass
if
ica
tio
n
m
o
d
el
s
,
w
h
ich
ar
e
co
n
s
id
er
ed
to
b
e
m
o
r
e
r
eliab
le
.
A
co
m
p
ar
is
o
n
o
f
d
if
f
er
en
t
m
o
d
els
co
n
d
u
cted
b
y
o
th
er
r
esear
ch
er
s
is
s
h
o
w
n
i
n
T
ab
le
1
.
B
ased
o
n
th
is
tab
le,
th
e
s
u
cc
es
s
r
ate
o
f
FER
d
etec
tio
n
a
m
o
n
g
v
ar
io
u
s
lig
h
t
w
e
ig
h
t
m
o
d
els
r
an
g
e
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
I
SS
N:
2089
-
4864
Dep
lo
yme
n
t a
n
d
ev
a
lu
a
tio
n
o
f
fa
cia
l e
xp
r
ess
io
n
r
ec
o
g
n
itio
n
o
n
A
n
d
r
o
id
…
(
Mo
h
a
ma
d
Ha
r
iz
N
a
z
a
mid
)
43
f
r
o
m
ap
p
r
o
x
i
m
a
tel
y
5
5
%
to
5
7
%
f
o
r
T
en
s
o
r
Flo
w
L
ite
(
T
FL
ite)
m
o
d
el
s
.
T
h
e
h
i
g
h
e
s
t
s
u
cc
ess
r
ate
o
b
s
er
v
ed
w
a
s
5
6
.
8
9
%,
ac
h
iev
ed
u
s
i
n
g
a
T
FL
ite
m
o
d
el
w
ith
t
h
e
M
n
as
Net
n
et
w
o
r
k
[
6
]
.
T
ab
le
1
.
T
h
e
co
m
p
ar
is
o
n
o
f
d
i
f
f
er
en
t li
g
h
t
w
ei
g
h
t
m
o
d
els
[
6
]
M
o
d
e
l
B
a
t
c
h
D
r
o
p
o
u
t
O
p
t
i
mi
z
e
r
LR
S
u
c
c
e
ss ra
t
e
(
%)
TF
TFL
i
t
e
M
n
a
sN
e
t
A
=
1
.
5
D
=
3
8
N
o
n
e
A
d
a
m
0
.
0
0
0
1
5
6
.
8
6
4
5
6
.
8
8
9
Ef
f
i
c
i
e
n
t
N
e
t
B
0
8
N
o
n
e
S
G
D
0
.
0
1
5
6
.
1
6
4
5
6
.
1
6
4
M
n
a
sN
e
t
A
=
1
.
0
D
=
7
8
N
o
n
e
A
d
a
m
0
.
0
0
0
1
5
5
.
9
8
9
5
6
.
0
1
4
Ef
f
i
c
i
e
n
t
N
e
t
B
0
8
N
o
n
e
A
d
a
m
0
.
0
0
1
5
5
.
9
1
4
5
5
.
9
3
9
Ef
f
i
c
i
e
n
t
N
e
t
B
0
8
N
o
n
e
A
d
a
m
0
.
0
0
1
5
5
.
6
8
9
5
5
.
6
8
9
D
e
n
se
N
e
t
1
2
1
8
N
o
n
e
A
d
a
m
0
.
0
0
0
1
5
5
.
6
3
9
5
5
.
6
3
9
Ef
f
i
c
i
e
n
t
N
e
t
B
1
8
N
o
n
e
S
G
D
0
.
0
1
5
5
.
4
8
9
5
5
.
4
8
9
Ef
f
i
c
i
e
n
t
N
e
t
B
0
4
N
o
n
e
A
d
a
m
0
.
0
0
1
5
5
.
2
6
4
5
5
.
2
6
4
M
n
a
sN
e
t
A
=
1
.
0
D
=
3
8
N
o
n
e
A
d
a
m
0
.
0
0
0
1
5
5
.
1
1
4
5
5
.
1
1
4
Ef
f
i
c
i
e
n
t
N
e
t
B
0
4
N
o
n
e
S
G
D
0
.
0
1
5
5
.
0
8
9
5
5
.
0
8
9
M
o
b
i
l
e
N
e
t
V
2
8
0
.
2
A
d
a
m
0
.
0
0
1
5
4
.
8
3
9
5
4
.
8
3
9
G
h
o
st
N
e
t
8
N
o
n
e
A
d
a
m
0
.
0
0
0
1
5
4
.
7
1
4
5
4
.
7
1
4
M
n
a
sN
e
t
A
=
1
.
0
D
=
1
8
N
o
n
e
A
d
a
m
0
.
0
0
0
1
5
4
.
5
8
9
5
4
.
5
6
4
M
o
b
i
l
e
N
e
t
V
3
L
a
r
g
e
A
=
1
.
2
5
16
0
.
5
A
d
a
m
0
.
0
0
1
5
4
.
4
8
9
5
4
.
3
8
9
Ef
f
i
c
i
e
n
t
N
e
t
V
2
B
1
16
N
o
n
e
S
G
D
0
.
0
1
5
4
.
4
6
4
5
4
.
4
3
9
M
o
b
i
l
e
N
e
t
V
3
L
a
r
g
e
A
=
1
.
2
5
16
0
.
2
A
d
a
m
0
.
0
0
1
5
4
.
3
6
4
5
4
.
3
3
9
Ef
f
i
c
i
e
n
t
N
e
t
V
2
B
0
8
N
o
n
e
S
G
D
0
.
0
1
5
4
.
2
6
4
5
4
.
2
3
9
S
q
u
e
e
z
e
N
e
t
C
=
1
.
0
8
0
.
2
A
d
a
m
0
.
0
0
0
1
5
4
.
1
3
9
5
4
.
1
1
4
Ef
f
i
c
i
e
n
t
N
e
t
V
2
B
1
8
N
o
n
e
S
G
D
0
.
0
1
5
4
.
0
6
4
5
4
.
0
3
9
M
o
b
i
l
e
N
e
t
V
3
S
mal
l
A
=
1
.
2
5
16
0
.
5
A
d
a
m
0
.
0
0
1
5
3
.
9
6
3
5
3
.
9
3
8
Me
an
w
h
ile,
r
o
b
o
t
ap
p
licatio
n
s
h
a
v
e
b
ec
o
m
e
i
n
d
is
p
e
n
s
ab
le
in
v
ar
io
u
s
s
ec
to
r
s
,
p
ar
ticu
la
r
l
y
w
it
h
in
m
en
tal
h
ea
lt
h
,
ed
u
ca
tio
n
,
an
d
s
er
v
ice
s
[
7
]
-
[
1
1
]
,
d
u
e
to
th
ei
r
d
ed
icate
d
f
u
n
c
tio
n
ali
t
y
a
n
d
r
eso
u
r
ce
ef
f
icie
n
c
y
.
T
h
e
d
em
a
n
d
f
o
r
lig
h
ter
,
m
o
r
e
ef
f
icien
t
m
o
d
el
s
s
u
itab
le
f
o
r
r
o
b
o
tic
ap
p
licatio
n
s
h
as
led
to
i
n
n
o
v
ati
v
e
s
o
l
u
tio
n
s
s
u
c
h
as
t
h
e
lo
ca
l
s
lid
in
g
w
i
n
d
o
w
atte
n
tio
n
n
et
w
o
r
k
(
SW
A
-
Net)
,
w
h
ich
ac
h
ie
v
es
an
ac
c
u
r
ac
y
o
f
9
0
.
0
3
%
o
n
R
A
F
-
DB
d
atasets
[
1
2
]
,
an
d
r
ec
en
t
tr
an
s
f
o
r
m
er
-
b
ased
FER
ap
p
r
o
ac
h
es,
s
u
c
h
as
th
e
atte
n
ti
v
e
p
o
o
lin
g
v
i
s
io
n
tr
an
s
f
o
r
m
er
b
y
Xu
e
et
a
l
.
[
1
3
]
,
r
ep
o
r
t
an
a
cc
u
r
ac
y
o
f
8
9
.
9
4
%
o
n
th
e
R
AF
-
DB
d
atas
et,
s
h
o
w
i
n
g
s
tr
o
n
g
g
en
er
aliza
tio
n
in
r
ea
l
-
w
o
r
ld
s
ce
n
ar
io
s
.
T
h
ese
ad
v
an
ce
m
e
n
ts
d
eliv
er
im
p
r
es
s
iv
e
r
es
u
lts
d
es
p
ite
r
ed
u
ce
d
m
o
d
el
s
izes a
n
d
r
eso
u
r
ce
d
e
m
an
d
s
.
An
o
th
er
i
m
p
o
r
tan
t
co
m
p
o
n
e
n
t
o
f
th
e
FE
R
m
o
d
el
is
t
h
e
d
atasets
u
s
e
d
f
o
r
tr
ain
i
n
g
.
B
ased
o
n
th
e
s
tu
d
y
,
t
h
er
e
ar
e
m
u
ltip
le
t
y
p
e
s
o
f
d
atasets
av
ailab
le
f
o
r
tr
ain
i
n
g
FE
R
m
o
d
els.
On
e
o
f
t
h
e
m
o
s
t
co
m
m
o
n
l
y
u
s
ed
b
en
c
h
m
ar
k
s
i
s
FE
R
2
0
1
3
an
d
its
ex
te
n
d
ed
v
er
s
io
n
,
F
E
R
+
[
1
4
]
.
FER+
d
ataset
s
ar
e
co
n
s
id
er
ed
clea
n
er
an
d
ca
n
p
r
o
v
id
e
b
etter
ac
cu
r
ac
y
.
A
d
d
itio
n
all
y
,
th
e
C
K+
d
at
aset
[
1
5
]
,
o
r
ig
in
all
y
co
m
p
r
i
s
in
g
v
id
eo
s
eq
u
e
n
ce
s
,
h
as
b
ee
n
ad
ap
ted
b
y
s
o
m
e
r
e
s
ea
r
ch
er
s
i
n
to
4
8
×4
8
g
r
a
y
s
ca
le
i
m
a
g
es
f
o
r
u
s
e
in
i
m
ag
e
clas
s
if
ica
tio
n
tr
ain
i
n
g
.
FER+
h
as
d
e
m
o
n
s
tr
ated
g
o
o
d
ac
cu
r
ac
y
i
n
p
o
o
lin
g
v
i
s
io
n
tr
an
s
f
o
r
m
er
r
esear
ch
,
ac
h
i
ev
in
g
a
v
alid
atio
n
ac
cu
r
ac
y
o
f
8
9
.
9
4
%
[
1
3
]
.
Me
a
n
w
h
ile,
t
h
e
C
K+
d
atase
t
is
o
n
e
o
f
t
h
e
b
est,
o
f
f
er
in
g
a
n
ac
c
u
r
ac
y
o
f
1
0
0
%
[
1
6
]
.
T
h
e
h
ig
h
ac
cu
r
ac
y
o
f
t
h
e
C
K+
d
ataset
ca
n
b
e
attr
ib
u
ted
to
th
e
co
n
s
is
ten
c
y
o
f
t
h
e
d
ata
i
n
i
m
ag
e
clas
s
if
i
ca
tio
n
.
FER+
al
s
o
b
en
ef
its
f
r
o
m
b
ein
g
a
clea
n
v
er
s
io
n
o
f
t
h
e
FE
R
2
0
1
3
d
ataset
[
1
7
]
,
w
ith
s
o
m
e
n
o
is
e
r
e
m
o
v
ed
.
T
h
e
n
o
is
e
i
n
th
e
d
ata
ca
n
b
e
s
ee
n
i
n
Fi
g
u
r
e
1
.
Fig
u
r
e
1
.
T
h
e
n
o
is
e
in
th
e
FER 2
0
1
3
d
atasets
T
h
e
i
m
p
le
m
e
n
tat
io
n
o
f
li
g
h
t
w
ei
g
h
t
FE
R
m
o
d
els
o
n
d
i
f
f
er
en
t
m
o
b
ile
d
ev
ice
s
h
as
b
e
en
w
id
el
y
s
tu
d
ied
,
w
it
h
m
u
l
tip
le
m
o
d
els
b
ein
g
d
ep
lo
y
ed
ac
r
o
s
s
v
ar
io
u
s
m
o
b
ile
p
latf
o
r
m
s
,
as
il
lu
s
tr
ate
d
in
T
ab
le
2
in
th
e
p
r
ac
tical
r
ev
ie
w
[
6
]
o
f
t
h
e
FE
R
s
y
s
te
m
in
th
e
F
i
g
u
r
e
1
.
W
h
i
le
in
t
h
e
m
icr
o
co
n
tr
o
ller
,
it
h
a
s
h
i
g
h
ca
p
ab
ilit
y
i
n
r
ea
l
ti
m
e
ap
p
licatio
n
e
m
b
ed
d
ed
in
th
er
ap
e
u
tic
s
y
s
te
m
s
[
1
8
]
.
Ho
w
e
v
er
,
th
e
i
n
te
g
r
atio
n
o
f
th
e
s
e
m
o
d
els
i
n
to
r
ea
l
r
o
b
o
ts
s
till
r
eq
u
ir
es
te
s
ti
n
g
to
ev
al
u
ate
th
eir
p
er
f
o
r
m
an
c
e
in
r
ea
l
-
ti
m
e
d
etec
tio
n
a
n
d
t
h
e
ac
cu
r
ac
y
o
f
FE
R
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
9
-
4864
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
,
Vo
l.
1
5
,
No
.
1
,
Ma
r
c
h
202
6
:
42
-
53
44
in
a
r
ea
l
-
ti
m
e
e
n
v
ir
o
n
m
en
t.
C
u
r
r
en
tl
y
,
th
e
ap
p
licatio
n
o
f
FE
R
o
n
t
h
e
T
em
i
V3
r
o
b
o
t
is
s
til
l
u
n
d
er
ex
p
lo
r
atio
n
,
as
th
er
e
i
s
li
m
ited
r
esear
ch
o
n
its
u
s
e
in
r
ea
l
-
li
f
e
i
m
p
le
m
e
n
tatio
n
s
.
I
n
FER
s
y
s
te
m
s
f
o
r
m
o
b
ile
d
ev
ices
o
r
s
m
ar
tp
h
o
n
es,
e
v
al
u
atio
n
m
etr
i
cs
in
cl
u
d
e
late
n
c
y
in
m
illi
s
ec
o
n
d
s
.
A
d
d
itio
n
all
y
,
ass
e
s
s
i
n
g
th
e
ac
c
u
r
ac
y
o
f
th
e
m
o
d
el
s
d
u
r
i
n
g
i
m
p
le
m
e
n
tatio
n
in
v
o
l
v
es
co
m
p
ar
i
n
g
th
e
m
o
d
el'
s
o
u
tp
u
t
w
i
th
ac
tu
al
h
u
m
a
n
e
m
o
tio
n
s
.
T
h
ese
ass
es
s
m
en
t
m
etr
ics
d
eter
m
in
e
th
e
ef
f
icie
n
c
y
o
f
th
e
m
o
d
el
's
p
r
o
ce
s
s
in
g
d
u
r
i
n
g
in
ter
ac
tio
n
s
w
ith
h
u
m
a
n
s
a
n
d
ev
alu
a
te
th
e
m
o
d
el's r
eliab
ilit
y
in
r
ea
l
-
l
i
f
e
i
m
p
le
m
en
ta
tio
n
s
.
T
ab
le
2
.
T
esti
n
g
o
f
d
i
f
f
er
e
n
t
m
o
d
el
o
n
d
i
f
f
er
en
t a
n
d
r
o
id
d
e
v
ice
[
6
]
M
o
d
e
l
S
i
z
e
(
MB
)
D
e
v
i
c
e
l
a
t
e
n
c
y
(
ms
)
Z
e
n
F
o
n
e
5
P
i
x
e
l
6
G
a
l
a
x
y
A
4
0
R
a
s
p
b
e
r
r
y
P
i
4
D
e
n
se
N
e
t
1
2
1
1
5
.
3
9
6
6
3
1
7
1
7
5
1
3
2
9
8
Ef
f
i
c
i
e
n
t
N
e
t
B
0
2
.
4
87
29
79
3
3
3
M
n
a
sN
e
t
A
=
1
.
0
D
=
1
5
.
7
7
1
0
4
39
86
4
6
1
M
n
a
sN
e
t
A
=
1
.
5
D
=
3
1
2
.
2
4
2
4
9
66
2
2
7
5
6
2
M
o
b
i
l
e
N
e
t
V
2
6
.
7
6
94
37
1
0
6
4
7
2
M
o
b
i
l
e
N
e
t
V
3
L
a
r
g
e
A
=
1
.
0
M
I
N
I
4
.
6
74
28
78
4
0
2
M
o
b
i
l
e
N
e
t
V
3
L
a
r
g
e
A
=
1
.
0
7
.
6
83
32
91
7
4
6
S
h
u
f
f
l
e
N
e
t
C
H
=
6
4
0
.
7
8
76
36
84
74
S
h
u
f
f
l
e
N
e
t
C
H
=
1
2
8
1
.
8
2
1
7
7
70
1
7
0
1
8
6
S
h
u
f
f
l
e
N
e
t
C
H
=
2
0
0
3
.
4
1
3
5
1
1
5
2
3
4
1
3
3
8
S
h
u
f
f
l
e
N
e
t
V
2
S
F
=
0
.
5
4
.
4
7
57
25
63
1
2
5
7
S
h
u
f
f
l
e
N
e
t
V
2
S
F
=
1
.
0
9
.
6
6
1
6
4
60
1
4
2
1
2
7
0
S
h
u
f
f
l
e
N
e
t
V
2
S
F
=
1
.
5
1
7
.
6
6
2
6
8
74
2
5
3
1
3
2
1
S
q
u
e
e
z
e
N
e
t
C
=
1
.
0
2
.
4
1
92
38
94
3
9
7
T
h
e
T
em
i
V3
,
ill
u
s
tr
ated
i
n
Fig
u
r
e
2
,
is
a
n
ad
v
a
n
ce
d
s
er
v
ice
r
o
b
o
t
f
ea
tu
r
i
n
g
s
m
ar
t
p
ath
p
lan
n
in
g
,
f
ac
e
tr
ac
k
i
n
g
,
N
L
P
,
tex
t
-
to
-
s
p
ee
ch
(
T
T
S)
,
s
p
ee
ch
-
to
-
te
x
t
(
STT
)
,
an
d
ca
m
er
a
ca
p
ab
ilit
ie
s
.
W
h
ile
th
e
r
o
b
o
t
s
u
p
p
o
r
ts
v
ar
io
u
s
c
u
s
to
m
izat
i
o
n
ap
p
r
o
ac
h
es,
th
is
s
t
u
d
y
f
o
cu
s
es
o
n
a
n
d
r
o
id
ap
p
licatio
n
d
ev
elo
p
m
e
n
t
b
y
in
te
g
r
atin
g
th
e
ex
i
s
ti
n
g
s
o
f
t
war
e
d
ev
elo
p
m
en
t
k
it
(
SDK)
p
r
o
v
id
ed
b
y
T
em
i
.
T
h
e
an
d
r
o
i
d
ap
p
licatio
n
w
as
in
s
ta
lled
o
n
th
e
T
em
i
V3
u
s
i
n
g
t
h
e
a
n
d
r
o
id
d
eb
u
g
b
r
id
g
e
(
A
DB
)
[
1
9
]
,
en
ab
lin
g
s
ea
m
le
s
s
d
ep
lo
y
m
e
n
t
an
d
o
p
tim
izatio
n
o
f
th
e
r
o
b
o
t’
s
ad
v
an
ce
d
f
u
n
ctio
n
alitie
s
to
alig
n
w
it
h
t
h
e
o
b
j
ec
tiv
es o
f
th
e
r
ese
ar
ch
.
Fig
u
r
e
2
.
R
o
b
o
t
T
em
i
V3
I
n
th
is
p
ap
er
,
th
e
au
th
o
r
w
il
l
co
n
tr
ib
u
te
to
a
FER
m
o
d
el
u
s
in
g
a
co
m
b
in
a
tio
n
o
f
th
e
FE
R
+
d
ataset
an
d
C
K+
4
8
d
ataset,
e
m
p
lo
y
i
n
g
th
e
T
en
s
o
r
Flo
w
f
r
a
m
e
w
o
r
k
an
d
th
e
p
r
e
-
tr
ai
n
ed
Mo
b
ileNetV2
[
2
0
]
n
et
w
o
r
k
.
T
h
e
p
ap
e
r
'
s
co
n
tr
ib
u
tio
n
s
i
n
cl
u
d
e
d
ev
elo
p
in
g
a
n
d
o
p
tim
izin
g
an
an
d
r
o
id
ap
p
licatio
n
f
o
r
im
p
le
m
e
n
tat
io
n
o
n
an
d
r
o
id
d
ev
ices
an
d
th
e
T
e
m
i
V3
r
o
b
o
t.
T
h
e
an
d
r
o
id
ap
p
licatio
n
w
ill
b
e
d
ev
elo
p
ed
u
s
i
n
g
t
h
e
R
o
b
o
t
T
em
i
SDK
[
1
9
]
an
d
Op
en
C
V
A
P
I
[
2
1
]
f
o
r
r
ea
l
-
ti
m
e
d
etec
tio
n
.
Op
ti
m
iza
tio
n
e
f
f
o
r
ts
w
i
ll
f
o
cu
s
o
n
es
tab
lis
h
in
g
g
u
id
eli
n
e
s
f
o
r
d
ev
elo
p
in
g
a
n
o
p
tim
ized
an
d
r
o
id
ap
p
licatio
n
.
I
t
w
i
ll
b
eg
i
n
w
it
h
b
ac
k
g
r
o
u
n
d
in
f
o
r
m
atio
n
an
d
r
esear
ch
o
n
t
h
e
i
m
p
le
m
e
n
tati
o
n
o
f
li
g
h
t
w
eig
h
t
m
o
d
els
i
n
an
d
r
o
id
s
y
s
te
m
s
,
alo
n
g
w
ith
th
e
d
ev
elo
p
m
e
n
t
o
f
m
o
d
el
s
u
s
i
n
g
t
h
e
T
FL
ite
f
r
a
m
e
w
o
r
k
[
2
2
]
.
Fo
llo
w
i
n
g
th
i
s
,
th
e
p
ap
er
w
il
l
d
etail
th
e
m
et
h
o
d
o
lo
g
y
an
d
ap
p
r
o
ac
h
u
s
ed
to
i
m
p
le
m
en
t
th
e
FE
R
m
o
d
el
i
n
to
a
cu
s
to
m
an
d
r
o
id
ap
p
licatio
n
,
as
w
el
l
as
th
e
ap
p
r
o
ac
h
em
p
lo
y
ed
to
co
llect
an
d
an
al
y
ze
r
ea
l
-
ti
m
e
d
etec
tio
n
d
ata
u
s
in
g
t
h
e
FE
R
s
y
s
te
m
o
n
t
h
e
T
e
m
i
V3
.
T
h
e
s
u
b
s
eq
u
en
t
s
ec
tio
n
w
il
l
p
r
esen
t
th
e
r
esu
lts
o
f
th
e
p
er
f
o
r
m
a
n
ce
an
al
y
s
is
an
d
o
f
f
er
o
p
tim
izatio
n
s
u
g
g
e
s
tio
n
s
f
o
r
ap
p
d
ev
elo
p
m
e
n
t
to
en
h
a
n
ce
t
h
e
b
ac
k
e
n
d
.
F
o
llo
w
i
n
g
t
h
is
,
t
h
e
p
ap
er
w
i
ll
d
is
c
u
s
s
t
h
e
tes
tin
g
o
f
th
e
FER
s
y
s
te
m
o
n
t
h
e
T
e
m
i
V3
,
f
o
cu
s
in
g
o
n
co
n
tr
o
lled
s
u
b
j
ec
ts
an
d
en
v
ir
o
n
m
e
n
ts
.
Fi
n
all
y
,
th
e
au
t
h
o
r
w
ill
co
n
cl
u
d
e
th
e
f
i
n
d
in
g
s
an
d
s
u
g
g
es
t
i
m
p
r
o
v
e
m
en
t
s
an
d
d
ir
ec
tio
n
s
f
o
r
f
u
t
u
r
e
d
ev
elo
p
m
en
t.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
I
SS
N:
2089
-
4864
Dep
lo
yme
n
t a
n
d
ev
a
lu
a
tio
n
o
f
fa
cia
l e
xp
r
ess
io
n
r
ec
o
g
n
itio
n
o
n
A
n
d
r
o
id
…
(
Mo
h
a
ma
d
Ha
r
iz
N
a
z
a
mid
)
45
2.
RE
S
E
ARCH
M
E
T
H
O
D
T
h
is
s
ec
tio
n
ex
p
lai
n
s
t
h
e
m
et
h
o
d
o
lo
g
y
ad
o
p
ted
in
th
is
s
t
u
d
y
in
d
etail.
I
t
b
eg
in
s
w
it
h
an
i
n
tr
o
d
u
ctio
n
to
th
e
cu
s
to
m
FER
T
FLite
m
o
d
el,
th
e
an
d
r
o
id
ap
p
licatio
n
in
teg
r
atio
n
w
it
h
t
h
e
cu
s
to
m
T
FL
ite
m
o
d
el,
a
n
d
th
e
ap
p
r
o
ac
h
u
s
ed
to
co
llect
d
ata
f
o
r
co
n
tr
o
lle
d
s
u
b
j
ec
ts
f
o
r
th
e
r
ea
l
-
ti
m
e
d
etec
tio
n
te
s
t.
2
.
1
.
F
a
cia
l e
x
press
io
n r
ec
o
g
nitio
n
T
en
s
o
rF
lo
w
L
it
e
m
o
d
el
T
h
e
d
atasets
u
s
ed
to
tr
ain
t
h
e
FER
m
o
d
el
w
er
e
a
co
m
b
in
at
i
o
n
o
f
t
h
e
FE
R
+
[
1
4
]
an
d
m
o
d
if
ied
C
K+
[
1
5
]
w
it
h
t
h
e
s
p
ec
i
f
icatio
n
o
f
4
8
×4
8
in
g
r
ay
s
ca
les
d
ata
s
ets
.
T
o
g
eth
er
,
th
e
y
co
n
tai
n
8
.
1
5
9
b
lack
-
a
n
d
-
w
h
ite
(
g
r
a
y
s
ca
le)
p
h
o
to
s
w
it
h
a
r
es
o
lu
tio
n
o
f
4
8
b
y
4
8
p
ix
el
s
.
F
o
r
tr
ain
in
g
an
d
te
s
ti
n
g
,
t
h
e
d
a
taset
is
d
i
v
id
ed
in
to
t
w
o
s
e
ts
:
7
.
3
4
3
p
h
o
to
s
f
o
r
tr
ain
in
g
a
n
d
8
1
6
p
h
o
to
s
f
o
r
test
i
n
g
.
T
h
e
i
m
a
g
es
r
ep
r
esen
t
s
e
v
en
f
u
n
d
a
m
en
ta
l
f
ac
ia
l
e
m
o
tio
n
s
:
an
g
er
,
f
ea
r
,
s
ad
n
es
s
,
n
eu
tr
alit
y
,
h
ap
p
in
e
s
s
,
s
u
r
p
r
is
e,
an
d
d
is
g
u
s
t.
T
h
e
n
u
m
b
er
o
f
im
a
g
es
v
ar
ies
ac
r
o
s
s
th
e
s
e
s
ev
e
n
d
is
tin
c
t
e
m
o
tio
n
ca
teg
o
r
ies.
Fi
g
u
r
e
3
i
l
lu
s
tr
ate
s
t
h
e
p
r
o
p
o
s
ed
m
o
d
el
ar
ch
itect
u
r
e
f
o
r
t
h
e
FER
s
y
s
te
m
.
T
h
e
i
n
p
u
t
la
y
er
r
ec
eiv
es
p
r
e
-
p
r
o
ce
s
s
ed
g
r
a
y
s
c
ale
f
ac
ial
i
m
ag
e
s
o
f
s
ize
4
8
×4
8
p
ix
els,
w
h
ic
h
ar
e
th
en
r
esized
to
9
6
×9
6
an
d
c
o
n
v
er
ted
to
R
GB
f
o
r
m
at
to
m
atc
h
th
e
in
p
u
t
r
eq
u
ir
e
m
e
n
ts
o
f
th
e
p
r
e
-
tr
ain
ed
Mo
b
ileNetV2
[
2
0
]
m
o
d
el.
T
h
i
s
m
o
d
el,
w
ith
i
n
cl
u
d
e
to
p
=False,
s
er
v
es
a
s
a
p
o
w
er
f
u
l
f
e
atu
r
e
ex
tr
ac
to
r
.
An
atten
tio
n
la
y
er
is
ap
p
lied
to
en
h
a
n
ce
t
h
e
m
o
d
el
’
s
f
o
cu
s
o
n
s
alien
t
f
ac
ial
r
e
g
io
n
s
.
T
h
e
ex
t
r
ac
ted
f
ea
tu
r
es
ar
e
th
en
p
as
s
ed
th
r
o
u
g
h
a
g
lo
b
al
av
er
ag
e
p
o
o
lin
g
(
G
A
P
)
lay
er
to
r
ed
u
ce
d
im
e
n
s
io
n
alit
y
.
T
o
m
iti
g
ate
o
v
er
f
itti
n
g
,
t
w
o
Dr
o
p
o
u
t
la
y
er
s
w
it
h
a
d
r
o
p
o
u
t
r
ate
o
f
0
.
5
ar
e
u
s
ed
.
A
f
u
ll
y
co
n
n
ec
ted
d
en
s
e
la
y
er
w
it
h
1
2
8
u
n
it
s
a
n
d
R
eL
U
ac
ti
v
atio
n
i
n
tr
o
d
u
ce
s
n
o
n
-
li
n
ea
r
it
y
,
f
o
llo
w
ed
b
y
a
f
in
al
d
en
s
e
o
u
tp
u
t la
y
er
w
it
h
a
s
i
g
m
o
id
ac
ti
v
atio
n
f
o
r
bi
n
ar
y
class
if
ica
tio
n
.
T
h
is
ar
ch
itect
u
r
e
en
ab
les
th
e
m
o
d
el
to
p
r
ed
ict
o
n
e
o
f
t
w
o
f
ac
ial
ex
p
r
ess
io
n
clas
s
es.
T
h
e
tr
ain
ed
m
o
d
el
is
later
co
n
v
er
te
d
to
T
FL
ite
[
2
2
]
f
o
r
m
at
to
en
a
b
le
ef
f
icie
n
t d
ep
lo
y
m
en
t o
n
ed
g
e
d
ev
ices.
Fig
u
r
e
3
.
T
h
e
FER
m
o
d
el
n
et
w
o
r
k
s
2
.
2
.
Andro
id a
pp
lica
t
io
n f
o
r
f
a
cia
l e
x
press
io
n r
ec
o
g
nitio
n
T
h
e
FER
s
y
s
te
m
d
es
ig
n
i
n
an
d
r
o
id
Stu
d
io
,
as
illu
s
tr
ated
in
Fig
u
r
e
4
,
b
eg
in
s
w
i
th
i
n
itia
liz
atio
n
u
s
in
g
th
e
Haa
r
C
ascad
e
m
o
d
el
an
d
Op
en
C
V
[
2
0
]
to
d
etec
t
th
e
p
r
esen
ce
o
f
a
f
ac
e
o
n
th
e
s
cr
ee
n
b
ef
o
r
e
d
eter
m
i
n
i
n
g
th
e
ex
p
r
ess
io
n
.
T
h
e
d
ev
elo
p
e
d
m
o
d
el
is
a
s
in
g
le
-
o
u
tp
u
t
s
y
s
te
m
,
p
r
o
d
u
cin
g
o
u
tp
u
ts
in
t
h
e
r
an
g
e
o
f
0
t
o
6
.
5
,
w
it
h
ea
c
h
r
an
g
e
co
r
r
esp
o
n
d
in
g
to
a
s
p
ec
if
ic
e
m
o
tio
n
clas
s
i
f
icatio
n
.
T
h
is
s
y
s
te
m
w
a
s
d
ev
elo
p
ed
in
an
d
r
o
id
s
tu
d
io
to
f
ac
ilit
ate
i
m
p
le
m
e
n
t
atio
n
o
n
m
u
ltip
le
d
ev
ices.
T
h
e
ap
p
licatio
n
co
n
s
i
s
ts
o
f
t
h
r
ee
m
a
i
n
ac
ti
v
itie
s
:
Ma
in
A
cti
v
it
y
,
C
a
m
er
a
A
cti
v
it
y
,
an
d
FacialE
x
p
r
ess
io
n
A
cti
v
i
t
y
.
T
h
e
Ma
in
A
ct
iv
it
y
s
er
v
e
s
as
th
e
lan
d
in
g
p
ag
e,
f
ea
t
u
r
in
g
a
b
u
tto
n
th
at
n
a
v
ig
ates
to
th
e
C
a
m
er
a
A
cti
v
it
y
a
n
d
FacialE
x
p
r
es
s
io
n
A
c
tiv
i
t
y
f
o
r
r
ea
l
-
ti
m
e
FE
R
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
9
-
4864
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
,
Vo
l.
1
5
,
No
.
1
,
Ma
r
c
h
202
6
:
42
-
53
46
d
etec
tio
n
.
T
h
e
in
ter
f
ac
e
f
o
r
th
e
Ma
in
Acti
v
it
y
i
s
s
h
o
wn
in
F
ig
u
r
e
4
(
a)
o
n
th
e
lef
t.
Me
an
w
h
ile,
t
h
e
C
a
m
er
a
A
ct
iv
i
t
y
an
d
FacialE
x
p
r
ess
io
n
Acti
v
it
y
h
an
d
le
r
ea
l
-
ti
m
e
f
ac
e
d
etec
tio
n
an
d
F
E
R
p
r
o
ce
s
s
in
g
,
as
d
ep
icted
in
th
e
f
lo
w
c
h
ar
t
in
Fig
u
r
e
5
,
w
it
h
th
eir
in
ter
f
a
ce
d
is
p
lay
ed
in
Fi
g
u
r
e
4
(
b
)
o
n
th
e
r
ig
h
t.
T
h
e
d
ev
elo
p
m
en
t
o
f
th
e
ap
p
licatio
n
w
as
g
u
id
ed
b
y
th
e
R
o
b
o
T
e
m
i
d
o
cu
m
en
ta
tio
n
[
1
9
]
,
Op
en
C
V
d
o
cu
m
e
n
tatio
n
[
2
1
]
,
an
d
an
d
r
o
id
d
ev
elo
p
er
d
o
cu
m
e
n
tatio
n
[
2
3
]
,
[
2
4
]
.
(
a)
(
b
)
Fig
u
r
e
4
.
T
h
e
ap
p
licatio
n
v
ie
w
f
o
r
;
(
a)
Ma
in
A
cti
v
it
y
a
n
d
(
b
)
C
a
m
er
a
A
cti
v
it
y
an
d
FacialE
x
p
r
ess
io
n
Acti
v
it
y
Fig
u
r
e
5
.
T
h
e
f
lo
w
c
h
ar
t f
o
r
an
d
r
o
id
a
p
p
licatio
n
d
ev
elo
p
ed
o
n
an
d
r
o
id
s
tu
d
io
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
I
SS
N:
2089
-
4864
Dep
lo
yme
n
t a
n
d
ev
a
lu
a
tio
n
o
f
fa
cia
l e
xp
r
ess
io
n
r
ec
o
g
n
itio
n
o
n
A
n
d
r
o
id
…
(
Mo
h
a
ma
d
Ha
r
iz
N
a
z
a
mid
)
47
2
.
3
.
Da
t
a
c
o
llect
io
n o
n
perf
o
r
m
a
nce
a
na
ly
s
i
s
T
h
e
alg
o
r
ith
m
u
n
d
er
w
e
n
t
s
y
s
t
e
m
atic
e
v
al
u
atio
n
ac
r
o
s
s
d
i
v
e
r
s
e
d
ev
ices,
i
n
clu
d
i
n
g
th
e
T
em
i
V3
an
d
th
e
Sa
m
s
u
n
g
Gala
x
y
A
5
2
2
0
2
1
.
T
h
e
T
em
i
V3
is
eq
u
ip
p
ed
w
it
h
a
n
A
R
M
He
x
a
-
C
o
r
e
p
r
o
ce
s
s
o
r
an
d
4
GB
o
f
R
A
M
[
2
5
]
,
[
2
6
]
,
w
h
ile
th
e
S
a
m
s
u
n
g
Gala
x
y
A
5
2
2
0
2
1
f
ea
tu
r
es
an
Octa
-
C
o
r
e
C
P
U
(
2
x
2
.
3
GHz
Kr
y
o
4
6
5
Go
ld
&
6
x
1
.
8
GHz
Kr
y
o
4
6
5
Sil
v
er
)
,
8
GB
o
f
R
A
M,
an
d
a
n
A
d
r
en
o
6
1
8
GP
U
[
2
7
]
,
[
2
8
]
.
T
h
ese
ev
alu
at
io
n
s
ai
m
ed
to
ass
es
s
p
er
f
o
r
m
a
n
ce
m
etr
ics
s
u
c
h
as
late
n
c
y
an
d
c
o
m
p
u
tatio
n
al
ef
f
icie
n
c
y
,
en
s
u
r
in
g
t
h
e
al
g
o
r
ith
m
's
co
m
p
atib
il
it
y
a
n
d
ef
f
ec
ti
v
e
n
es
s
ac
r
o
s
s
d
if
f
er
en
t
h
ar
d
w
ar
e
p
la
tf
o
r
m
s
.
T
h
e
test
in
g
p
r
o
ce
d
u
r
e
f
o
r
th
e
alg
o
r
ith
m
i
n
v
o
l
v
ed
in
itia
lizi
n
g
th
e
p
r
o
g
r
a
m
o
n
t
h
e
s
p
ec
i
f
ie
d
d
ev
ices,
f
o
llo
w
ed
b
y
a
co
n
tin
u
o
u
s
r
u
n
d
u
r
atio
n
o
f
ap
p
r
o
x
i
m
a
tel
y
t
wen
t
y
m
i
n
u
te
s
.
P
ar
ticip
an
ts
p
o
s
itio
n
ed
th
e
m
s
el
v
es
in
f
r
o
n
t
o
f
th
e
ca
m
er
a
an
d
in
t
er
ac
ted
w
it
h
th
e
d
ev
ice
s
to
te
s
t
th
e
FER
s
y
s
te
m
.
T
h
ey
n
a
v
ig
ated
to
th
e
m
ai
n
p
ag
e
an
d
ac
ce
s
s
ed
th
e
FER
s
y
s
te
m
b
y
ac
ti
v
ati
n
g
t
h
e
d
esig
n
a
ted
b
u
tto
n
w
it
h
i
n
th
e
ap
p
licati
o
n
.
A
d
d
itio
n
all
y
,
a
n
o
n
-
d
etec
tio
n
s
ce
n
ar
io
w
a
s
tes
ted
b
y
h
a
v
i
n
g
p
ar
ticip
an
ts
ex
i
t
th
e
ca
m
er
a
f
r
a
m
e,
en
s
u
r
i
n
g
n
o
s
u
b
j
ec
t
r
em
ai
n
ed
w
it
h
i
n
it
s
b
o
u
n
d
ar
ies
w
h
ile
t
h
e
FER s
y
s
te
m
co
n
ti
n
u
ed
to
o
p
er
ate.
T
h
is
ev
alu
at
io
n
ai
m
s
to
ass
ess
p
o
w
er
,
C
P
U,
an
d
R
AM
co
n
s
u
m
p
tio
n
to
g
u
id
e
ap
p
licatio
n
o
p
tim
izatio
n
f
o
r
b
ac
k
en
d
d
ep
lo
y
m
e
n
t.
R
A
M
u
s
ag
e
is
m
ea
s
u
r
ed
u
n
d
er
tw
o
co
n
d
itio
n
s
:
n
o
n
-
d
etec
tio
n
(
n
o
u
s
er
in
f
r
a
m
e)
a
n
d
d
etec
tio
n
(
u
s
er
p
r
esen
t)
,
h
elp
in
g
id
e
n
ti
f
y
o
p
p
o
r
tu
n
ities
to
r
elea
s
e
u
n
n
ec
es
s
ar
y
ac
ti
v
it
ies
a
n
d
r
ed
u
ce
m
e
m
o
r
y
lo
ad
.
C
P
U
co
n
s
u
m
p
tio
n
is
an
al
y
ze
d
to
u
n
d
er
s
ta
n
d
its
i
m
p
ac
t
o
n
s
y
s
te
m
late
n
c
y
,
w
it
h
co
m
p
ar
is
o
n
s
b
et
w
ee
n
t
w
o
d
ev
ices
o
f
f
er
in
g
in
s
i
g
h
ts
b
ased
o
n
h
ar
d
w
ar
e
ca
p
ab
ilit
ies.
P
o
w
er
co
n
s
u
m
p
t
io
n
is
also
ev
al
u
ated
to
m
i
n
i
m
ize
en
er
g
y
u
s
a
g
e
d
u
r
i
n
g
co
n
ti
n
u
o
u
s
o
p
er
atio
n
,
en
s
u
r
in
g
t
h
e
ap
p
licatio
n
r
e
m
ai
n
s
ef
f
icien
t
w
h
e
n
r
u
n
n
in
g
i
n
th
e
b
ac
k
g
r
o
u
n
d
o
f
an
d
r
o
id
s
y
s
te
m
s
.
2.
4
.
Da
t
a
c
o
llect
io
n o
n c
o
ntr
o
lled
s
et
t
ing
s
I
n
th
is
s
tu
d
y
,
th
e
d
ata
co
llectio
n
w
a
s
d
o
n
e
f
o
r
co
n
tr
o
lled
s
ettin
g
s
(
w
ell
li
g
h
ted
r
o
o
m
)
as
ill
u
s
tr
ated
in
Fig
u
r
e
6
.
T
h
e
lin
e
o
f
s
ig
h
t
o
f
th
e
s
u
b
j
ec
t
w
ill
f
ac
e
th
e
s
cr
e
en
o
f
T
em
i
V3
as
in
Fig
u
r
e
6
(
a)
,
w
h
ile
in
s
o
m
e
ca
s
es,
th
er
e
ar
e
d
if
f
er
e
n
ce
s
i
n
th
e
h
ei
g
h
t,
th
e
T
em
i
V3
s
cr
e
en
w
ill
tilt
to
m
a
k
e
s
u
r
e
th
e
l
in
e
o
f
s
i
g
h
t
o
f
th
e
s
u
b
j
ec
t f
ac
es t
h
e
s
cr
ee
n
a
s
s
h
o
w
n
i
n
Fi
g
u
r
e
6
(
b
)
.
(
a)
(
b
)
Fig
u
r
e
6
.
Data
co
llectio
n
s
et
u
p
; (
a)
th
e
co
n
tr
o
lled
s
ettin
g
s
f
o
r
d
ata
co
llectio
n
an
d
(
b
)
T
em
i V
3
ca
p
ab
ilit
ies f
o
r
f
ac
e
tr
ac
k
i
n
g
b
ased
o
n
d
if
f
er
en
t h
u
m
an
h
ei
g
h
t
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
is
s
ec
tio
n
w
il
l
d
etail
t
h
e
r
e
s
u
lt
s
o
b
tain
ed
f
r
o
m
t
h
e
s
tu
d
y
,
d
iv
id
ed
in
to
s
y
s
te
m
o
p
ti
m
iz
atio
n
a
n
d
an
al
y
s
is
,
FER
T
FL
ite
m
o
d
el
r
esu
lt,
a
n
d
FER r
ea
l ti
m
e
d
etec
tio
n
test
i
n
g
o
n
co
n
tr
o
lled
s
u
b
jects.
3
.
1
.
L
ig
ht
w
eig
ht
f
a
ci
a
l e
x
press
io
n r
ec
o
g
nitio
n
m
o
del
T
h
e
lig
h
t
w
ei
g
h
t
m
o
d
el
d
ev
el
o
p
ed
u
s
in
g
t
h
e
n
et
w
o
r
k
d
es
cr
ib
ed
in
th
e
m
eth
o
d
o
lo
g
y
a
ch
iev
ed
a
v
alid
atio
n
ac
c
u
r
ac
y
o
f
9
2
.
8
6
%.
T
h
is
ac
cu
r
ac
y
i
s
co
n
s
id
er
ed
f
air
w
h
e
n
co
m
p
ar
ed
to
p
r
ev
io
u
s
s
t
u
d
ies,
s
u
c
h
as
th
e
v
alid
atio
n
ac
cu
r
ac
y
o
f
8
9
.
9
4
%
w
a
s
r
ep
o
r
ted
b
y
Xu
e
et
a
l
.
[
1
3
]
u
s
in
g
a
tr
an
s
f
o
r
m
er
-
b
ased
m
o
d
el,
an
d
th
e
C
K+
d
ataset,
w
h
ich
r
ep
o
r
ted
a
v
alid
atio
n
ac
c
u
r
ac
y
o
f
1
0
0
%
[
1
6
]
,
[
2
9
]
,
[
3
0
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
9
-
4864
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
,
Vo
l.
1
5
,
No
.
1
,
Ma
r
c
h
202
6
:
42
-
53
48
3
.
2
.
Andro
id a
pp
lica
t
io
n per
f
o
r
m
a
nce
a
na
ly
s
is
a
nd
o
pti
m
i
za
t
io
n (
T
e
m
i
V3
v
s
Sa
m
s
u
ng
A5
2
2
0
2
1
)
3
.
2
.
1
.
CP
U
p
er
f
o
r
m
a
nce
T
h
e
T
em
i
V3
C
P
U
p
er
f
o
r
m
a
n
ce
p
ea
k
s
at
3
0
%
u
s
a
g
e,
d
e
m
o
n
s
tr
ati
n
g
ef
f
icie
n
t
o
p
er
atio
n
c
o
m
p
ar
ab
le
to
s
tan
d
ar
d
ap
p
licatio
n
s
.
Fi
g
u
r
e
7
illu
s
tr
ates
in
ter
v
als
o
f
id
l
e
C
P
U
u
s
a
g
e,
co
in
cid
i
n
g
w
it
h
p
er
io
d
s
w
h
e
n
FE
R
ac
tiv
it
y
ce
a
s
es
a
f
ter
p
ar
ticip
a
n
ts
d
is
co
n
ti
n
u
e
d
etec
tio
n
.
T
h
ese
f
in
d
i
n
g
s
s
u
g
g
est
th
a
t
t
h
e
FER
s
y
s
te
m
ca
n
ef
f
ec
tiv
e
l
y
o
p
er
ate
co
n
cu
r
r
en
t
l
y
w
i
th
o
th
er
ap
p
lic
atio
n
s
,
d
el
iv
er
in
g
f
ee
d
b
ac
k
b
ased
o
n
FE
R
m
o
d
e
d
ec
i
s
io
n
s
.
T
h
e
C
P
U
is
in
v
er
s
el
y
p
r
o
p
o
r
ti
o
n
al
to
th
e
laten
c
y
o
f
t
h
e
ap
p
licatio
n
.
Me
an
w
h
ile,
t
h
e
Sa
m
s
u
n
g
A
5
2
ex
h
ib
its
a
p
ea
k
C
P
U
co
n
s
u
m
p
tio
n
o
f
3
1
%,
as
d
ep
icted
in
Fig
u
r
e
7
(
b
)
,
w
h
ich
is
co
m
p
ar
ab
le
to
th
e
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
T
em
i
V3
.
No
tab
l
y
,
C
P
U
u
s
ag
e
d
r
o
p
s
al
m
o
s
t
to
ze
r
o
wh
en
th
e
ac
ti
v
it
y
i
s
ter
m
in
ate
d
,
s
u
c
h
as
w
h
e
n
a
p
ar
ticip
an
t
r
etu
r
n
s
to
th
e
m
ai
n
p
ag
e.
No
tab
ilit
y
h
ap
p
en
s
i
n
th
e
T
em
i
V3
as
in
Fig
u
r
e
7
(
a)
.
T
h
e
C
P
U
p
o
w
er
is
in
v
er
s
el
y
p
r
o
p
o
r
tio
n
al
to
th
e
l
aten
c
y
o
f
t
h
e
ap
p
licatio
n
.
B
y
k
n
o
w
i
n
g
th
i
s
,
p
o
w
er
o
p
ti
m
iza
tio
n
ca
n
b
e
d
o
n
e
b
y
d
estro
y
i
n
g
th
e
ac
t
iv
i
t
y
a
n
d
n
a
v
ig
a
tin
g
it to
Ma
i
n
A
cti
v
it
y
wh
en
t
h
er
e
is
n
o
FER d
etec
tio
n
h
ap
p
en
in
g
.
(
a)
(
b
)
Fig
u
r
e
7
.
C
P
U
p
er
f
o
r
m
a
n
ce
o
f
;
(
a)
T
em
i
V3
an
d
(
b
)
Sa
m
s
u
n
g
A
5
2
3
.
2
.
2
.
M
em
o
ry
c
o
ns
u
m
ptio
n
Fig
u
r
e
8
s
h
o
w
s
t
h
e
R
A
M
co
n
s
u
m
p
t
io
n
f
o
r
T
em
i
V3
an
d
Sa
m
s
u
n
g
A
5
2
.
T
h
e
T
em
i
V3
ex
h
ib
its
a
p
ea
k
R
A
M
u
s
ag
e
o
f
7
2
0
.
4
MB
o
u
t
o
f
it
s
av
a
ilab
le
4
GB
,
as
d
ep
icted
in
Fi
g
u
r
e
8
(
a)
.
T
h
is
p
ea
k
o
cc
u
r
s
n
o
tab
l
y
w
h
e
n
p
ar
ticip
an
ts
m
o
v
e
o
u
t
o
f
f
r
a
m
e
o
r
d
is
tan
ce
th
e
m
s
e
l
v
es
f
r
o
m
t
h
e
d
ev
ice,
u
n
d
er
s
c
o
r
in
g
th
e
n
ee
d
to
o
p
tim
ize
t
h
e
ap
p
licatio
n
to
tr
ig
g
er
FER
s
elec
ti
v
el
y
,
s
u
c
h
as
b
y
u
s
i
n
g
a
b
u
tto
n
o
r
d
etec
tin
g
a
f
ac
e
i
n
f
r
o
n
t
o
f
th
e
ca
m
er
a.
Me
an
w
h
i
le,
Fig
u
r
e
8
(
b
)
d
is
p
lay
s
t
h
e
m
e
m
o
r
y
co
n
s
u
m
p
tio
n
o
f
t
h
e
Sa
m
s
u
n
g
A
5
2
d
u
r
in
g
FE
R
s
y
s
te
m
o
p
er
atio
n
.
P
ea
k
s
in
th
e
g
r
ap
h
o
cc
u
r
w
h
e
n
p
ar
ticip
an
t
s
m
o
v
e
o
u
t
o
f
f
r
a
m
e
o
r
d
is
tan
ce
th
e
m
s
elv
e
s
f
r
o
m
th
e
d
ev
ice,
r
eq
u
ir
in
g
i
n
cr
ea
s
ed
p
r
o
ce
s
s
in
g
m
e
m
o
r
y
to
d
etec
t
f
ac
ial
ex
p
r
ess
io
n
s
.
T
h
ese
is
s
u
e
s
ca
n
b
e
o
v
er
co
m
e
b
y
d
estro
y
in
g
th
e
ac
tiv
it
y
an
d
n
a
v
i
g
ati
n
g
it
to
Ma
in
A
cti
v
it
y
w
h
e
n
th
er
e
is
n
o
FER
d
etec
tio
n
h
ap
p
en
in
g
.
(
a)
(
b
)
Fig
u
r
e
8
.
Me
m
o
r
y
co
n
s
u
m
p
tio
n
;
(
a)
T
em
i
V3
an
d
(
b
)
Sa
m
s
u
n
g
A
5
2
3
.
2
.
3
.
E
nerg
y
c
o
ns
u
m
ptio
n
E
n
er
g
y
co
n
s
u
m
p
tio
n
d
ata,
illu
s
tr
ated
in
Fi
g
u
r
e
9
,
is
cr
u
cial
f
o
r
o
p
tim
izi
n
g
b
atter
y
u
s
a
g
e
d
u
r
in
g
ap
p
licatio
n
d
ev
elo
p
m
e
n
t.
I
n
te
r
v
als
o
f
n
o
b
atter
y
co
n
s
u
m
p
ti
o
n
co
in
cid
e
w
it
h
t
h
e
ce
s
s
atio
n
o
f
FE
R
ac
ti
v
it
y
,
s
u
g
g
e
s
ti
n
g
th
at
ef
f
ec
ti
v
e
m
a
n
ag
e
m
e
n
t
o
f
FE
R
ac
tiv
it
y
ca
n
s
ig
n
i
f
ica
n
tl
y
e
x
te
n
d
th
e
T
em
i
V3
'
s
b
atter
y
li
f
e
as
s
h
o
w
n
in
Fi
g
u
r
e
9
(
a)
.
I
n
co
n
tr
ast,
Fig
u
r
e
9
(
b
)
d
ep
icts
th
e
en
er
g
y
co
n
s
u
m
p
tio
n
o
f
th
e
F
E
R
s
y
s
te
m
o
n
t
h
e
Sa
m
s
u
n
g
A
5
2
,
s
h
o
w
i
n
g
m
o
d
e
r
ate
p
o
w
er
co
n
s
u
m
p
tio
n
at
its
p
ea
k
.
Si
m
ilar
to
C
P
U
u
s
a
g
e,
en
er
g
y
co
n
s
u
m
p
tio
n
d
ec
r
ea
s
es si
g
n
i
f
ica
n
tl
y
w
h
e
n
t
h
e
FER ac
ti
v
it
y
s
to
p
s
,
as sh
o
wn
in
Fig
u
r
e
9
(
b
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
I
SS
N:
2089
-
4864
Dep
lo
yme
n
t a
n
d
ev
a
lu
a
tio
n
o
f
fa
cia
l e
xp
r
ess
io
n
r
ec
o
g
n
itio
n
o
n
A
n
d
r
o
id
…
(
Mo
h
a
ma
d
Ha
r
iz
N
a
z
a
mid
)
49
(
a)
(
b
)
Fig
u
r
e
9
.
E
n
er
g
y
co
n
s
u
m
p
tio
n
; (
a)
T
em
i V
3
an
d
(
b
)
Sa
m
s
u
n
g
A
5
2
3
.
2
.
4
.
O
pti
m
iza
t
io
n
Me
an
w
h
ile,
th
e
laten
c
y
o
f
ea
ch
d
ev
ice
is
as
f
o
llo
w
s
:
T
em
i
V3
w
it
h
1
2
0
m
s
an
d
Sa
m
s
u
n
g
Gala
x
y
A
5
2
2
0
2
1
w
it
h
9
2
m
s
.
B
ase
d
o
n
th
e
d
ata
p
r
esen
ted
,
o
p
ti
m
izi
n
g
t
h
e
FER
s
y
s
te
m
i
m
p
l
e
m
en
tatio
n
ca
n
b
e
ac
h
iev
ed
th
r
o
u
g
h
s
e
v
er
al
k
e
y
ch
ar
ac
ter
is
tics
.
First,
co
n
d
iti
o
n
al
a
ctiv
atio
n
en
s
u
r
es
th
e
F
E
R
s
y
s
te
m
i
s
o
n
l
y
ac
tiv
e
w
h
e
n
a
f
ac
e
is
d
etec
ted
,
u
tili
zin
g
m
o
d
els
lik
e
Op
en
C
V
o
r
a
b
u
tto
n
-
tr
ig
g
er
ed
m
ec
h
an
is
m
to
in
itiate
it.
Seco
n
d
,
R
eso
u
r
ce
Ma
n
ag
e
m
e
n
t
is
cr
itical,
r
eq
u
ir
i
n
g
p
r
o
p
e
r
ter
m
i
n
atio
n
in
t
h
e
ap
p
licati
o
n
co
d
e
to
p
r
ev
en
t
u
n
n
ec
es
s
ar
y
p
o
w
er
,
R
A
M,
o
r
C
P
U
co
n
s
u
m
p
t
io
n
d
u
r
in
g
id
l
e
p
er
io
d
s
.
L
astl
y
,
d
ev
ice
p
er
f
o
r
m
a
n
ce
h
i
g
h
li
g
h
t
s
ef
f
icien
t
laten
c
y
a
n
d
o
p
er
atio
n
al
ef
f
ec
tiv
e
n
e
s
s
o
n
b
o
th
th
e
T
em
i
V3
r
o
b
o
t
an
d
Sa
m
s
u
n
g
Gala
x
y
A
5
2
,
ef
f
ec
tiv
e
l
y
m
an
a
g
i
n
g
C
P
U,
R
A
M,
an
d
en
er
g
y
u
s
a
g
e.
Fu
r
t
h
er
co
n
s
id
er
atio
n
s
f
o
r
d
ev
elo
p
in
g
t
h
e
FER
s
y
s
te
m
in
an
a
n
d
r
o
id
ap
p
licatio
n
in
cl
u
d
e
ad
d
r
ess
in
g
n
o
is
e
i
s
s
u
es
t
h
at
ar
is
e
w
h
en
u
s
er
s
ar
e
d
is
tan
t
f
r
o
m
th
e
d
e
v
ice
's
ca
m
er
a,
p
o
s
itio
n
ed
o
u
ts
id
e
t
h
e
f
r
a
m
e,
o
r
w
h
en
n
o
f
ac
e
i
s
d
etec
ted
.
T
h
ese
ch
allen
g
es
ca
n
b
e
ef
f
ec
tiv
e
l
y
m
an
a
g
ed
b
y
in
teg
r
ati
n
g
a
f
ac
e
d
etec
tio
n
m
o
d
el
in
to
t
h
e
b
ac
k
en
d
o
f
t
h
e
a
n
d
r
o
id
ap
p
licatio
n
,
en
s
u
r
i
n
g
t
h
at
t
h
e
FER s
y
s
te
m
ac
ti
v
ate
s
o
n
l
y
u
p
o
n
d
etec
tin
g
a
f
ac
e
[
3
1
]
,
[
3
2
]
.
T
o
co
n
clu
d
e
th
is
p
ar
t
o
f
th
e
r
esu
lt
s
,
p
er
f
o
r
m
a
n
ce
an
al
y
s
i
s
w
a
s
co
n
d
u
cted
to
o
p
tim
ize
th
e
ac
tiv
it
y
lif
ec
y
cle
m
a
n
ag
e
m
e
n
t
(
s
u
c
h
as
ac
ti
v
it
y
cr
ea
tio
n
an
d
d
estru
ctio
n
)
i
n
J
av
a
-
b
ased
an
d
r
o
id
ap
p
licatio
n
d
ev
elo
p
m
en
t.
T
h
e
an
al
y
s
i
s
r
ev
ea
led
an
in
v
er
s
e
r
elatio
n
s
h
i
p
b
etw
ee
n
C
P
U
q
u
alit
y
an
d
ap
p
licatio
n
laten
c
y
:
b
etter
C
P
Us
r
esu
lt
in
lo
w
er
laten
c
y
.
Des
p
ite
th
e
v
ar
y
i
n
g
C
P
U
p
er
f
o
r
m
a
n
ce
,
o
v
er
all
C
P
U
c
o
n
s
u
m
p
tio
n
r
e
m
ain
s
co
n
s
tan
t,
as
t
h
e
ap
p
licatio
n
is
t
h
e
s
a
m
e.
T
h
e
co
m
p
ar
is
o
n
b
et
w
ee
n
t
h
e
Sa
m
s
u
n
g
A
5
2
an
d
T
e
m
i
V3
d
ev
ices
w
as
u
n
d
er
tak
e
n
to
o
b
s
er
v
e
th
e
s
u
cc
e
s
s
o
f
o
p
ti
m
i
za
tio
n
ef
f
o
r
ts
.
F
u
r
th
er
te
s
ti
n
g
o
f
r
ea
l
-
ti
m
e
FE
R
d
etec
tio
n
o
n
th
e
T
e
m
i V
3
is
p
l
an
n
ed
to
ass
e
s
s
t
h
e
m
o
d
el
'
s
ac
cu
r
ac
y
i
n
r
ea
l
-
t
i
m
e
s
ce
n
ar
io
s
[
3
3
]
,
[
3
4
]
.
3
.
3
.
Co
ntr
o
lled su
bje
ct
da
t
a
co
llect
io
n
T
h
e
FER
s
y
s
te
m
d
e
m
o
n
s
tr
ate
d
v
ar
iab
le
p
er
f
o
r
m
a
n
ce
ac
r
o
s
s
3
0
s
u
b
j
ec
ts
i
n
a
co
n
tr
o
lled
en
v
ir
o
n
m
en
t,
as
ill
u
s
tr
ated
i
n
Fi
g
u
r
e
1
0
.
T
h
e
s
y
s
te
m
ac
h
ie
v
ed
an
av
er
a
g
e
ac
cu
r
ac
y
o
f
8
2
.
2
8
%,
w
it
h
in
d
iv
id
u
al
s
u
b
j
ec
t
ac
cu
r
ac
ies
r
an
g
in
g
f
r
o
m
3
9
.
1
3
%
to
1
0
0
%.
No
ta
b
ly
,
f
iv
e
s
u
b
j
ec
ts
(
1
8
,
2
1
,
2
2
,
2
3
,
an
d
3
0
)
ex
h
ib
ited
p
er
f
ec
t
r
ec
o
g
n
itio
n
r
ates,
w
h
i
le
th
e
m
aj
o
r
ity
(
ap
p
r
o
x
i
m
atel
y
2
0
o
u
t
o
f
3
0
)
s
h
o
w
ed
ac
cu
r
ac
ies
ab
o
v
e
8
0
%.
Ho
w
e
v
er
,
th
e
s
y
s
te
m
s
tr
u
g
g
led
w
i
th
a
f
e
w
s
u
b
j
ec
ts
,
p
ar
ticu
lar
l
y
s
u
b
j
ec
ts
4
,
5
,
an
d
1
4
,
w
h
er
e
ac
c
u
r
ac
y
f
ell
b
elo
w
7
0
%.
T
h
is
w
id
e
p
er
f
o
r
m
a
n
ce
r
an
g
e
s
u
g
g
e
s
ts
t
h
at
w
h
ile
th
e
FE
R
s
y
s
te
m
is
h
i
g
h
l
y
ef
f
ec
ti
v
e
f
o
r
m
o
s
t
s
u
b
j
ec
ts
,
it
m
a
y
b
e
s
en
s
it
iv
e
to
i
n
d
iv
id
u
al
f
ac
ia
l c
h
ar
ac
ter
is
tic
s
o
r
ex
p
r
ess
io
n
n
u
a
n
ce
s
.
T
h
e
p
r
esen
ce
o
f
b
o
t
h
p
er
f
ec
t sco
r
es a
n
d
s
ig
n
i
f
ica
n
tl
y
lo
w
er
ac
cu
r
ac
y
h
ig
h
li
g
h
ts
t
h
e
n
ee
d
f
o
r
f
u
r
th
er
in
v
esti
g
atio
n
in
to
th
e
f
ac
to
r
s
in
f
l
u
e
n
ci
n
g
r
ec
o
g
n
itio
n
ac
cu
r
ac
y
.
T
h
es
e
f
i
n
d
in
g
s
u
n
d
er
s
co
r
e
th
e
i
m
p
o
r
tan
ce
o
f
r
ef
i
n
i
n
g
th
e
alg
o
r
ith
m
to
en
h
an
c
e
co
n
s
is
ten
c
y
ac
r
o
s
s
d
iv
er
s
e
s
u
b
j
ec
ts
,
p
o
ten
tiall
y
t
h
r
o
u
g
h
i
m
p
r
o
v
ed
f
ea
tu
r
e
ex
tr
ac
tio
n
o
r
th
e
in
co
r
p
o
r
atio
n
o
f
ad
ap
tiv
e
lear
n
i
n
g
tec
h
n
iq
u
es t
o
ac
co
u
n
t f
o
r
in
d
i
v
id
u
al
v
ar
iat
io
n
s
i
n
f
ac
i
al
e
x
p
r
ess
io
n
s
.
T
h
e
FER
s
y
s
te
m
d
e
m
o
n
s
tr
ate
d
v
ar
y
i
n
g
lev
el
s
o
f
ac
cu
r
ac
y
ac
r
o
s
s
d
if
f
er
en
t
e
x
p
r
ess
io
n
s
w
h
e
n
test
ed
o
n
3
0
s
u
b
j
ec
ts
in
a
co
n
tr
o
lled
en
v
ir
o
n
m
e
n
t.
As
ill
u
s
tr
ated
i
n
Fi
g
u
r
e
1
1
,
th
e
s
y
s
te
m
ex
ce
l
led
in
r
ec
o
g
n
iz
in
g
h
ap
p
y
(
9
2
.
2
8
%)
an
d
n
eu
tr
al
(
9
1
.
8
2
%)
ex
p
r
ess
io
n
s
,
s
h
o
w
e
d
m
o
d
er
ate
p
er
f
o
r
m
a
n
ce
f
o
r
s
ad
(
7
5
.
8
9
%),
an
d
s
tr
u
g
g
led
w
i
th
a
n
g
r
y
(
6
1
.
4
2
%)
an
d
s
u
r
p
r
is
e
(
4
6
.
1
8
%)
ex
p
r
ess
io
n
s
.
T
h
e
o
v
er
all
av
er
a
g
e
ac
c
u
r
ac
y
w
a
s
7
3
.
5
2
%.
T
h
ese
r
esu
lts
h
i
g
h
li
g
h
t
a
s
ig
n
i
f
ica
n
t
d
is
p
ar
it
y
in
th
e
s
y
s
te
m
's
ab
ilit
y
to
d
etec
t
d
if
f
e
r
en
t
e
m
o
tio
n
al
s
tates,
w
it
h
p
ar
ticu
lar
ch
alle
n
g
e
s
in
d
is
ti
n
g
u
i
s
h
in
g
b
et
w
ee
n
s
u
r
p
r
is
e
an
d
an
g
er
.
T
h
e
h
ig
h
ac
c
u
r
ac
y
f
o
r
h
ap
p
y
an
d
n
eu
tr
a
l
ex
p
r
ess
io
n
s
s
u
g
g
est
s
r
o
b
u
s
t
p
er
f
o
r
m
an
ce
f
o
r
th
e
s
e
co
m
m
o
n
s
tates,
w
h
ic
h
is
p
r
o
m
i
s
i
n
g
f
o
r
m
an
y
r
ea
l
-
w
o
r
ld
ap
p
licatio
n
s
.
Ho
w
ev
er
,
t
h
e
lo
w
er
ac
c
u
r
ac
y
f
o
r
an
g
r
y
an
d
s
u
r
p
r
is
e
ex
p
r
es
s
io
n
s
,
a
lo
n
g
w
it
h
r
ep
o
r
ted
in
s
ta
n
ce
s
o
f
m
is
cla
s
s
i
f
icatio
n
b
et
w
ee
n
s
ad
/an
g
r
y
an
d
n
eu
tr
al
ex
p
r
es
s
io
n
s
,
i
n
d
icate
s
ar
ea
s
r
eq
u
ir
in
g
i
m
p
r
o
v
e
m
e
n
t.
T
h
es
e
f
in
d
in
g
s
u
n
d
e
r
s
c
o
r
e
th
e
n
e
e
d
f
o
r
f
u
r
th
e
r
r
ef
in
em
en
t
o
f
th
e
F
E
R
a
lg
o
r
i
th
m
s
,
es
p
e
c
i
a
lly
in
d
if
f
e
r
e
n
ti
a
ti
n
g
b
e
t
w
e
en
ea
s
il
y
co
n
f
u
s
ed
e
x
p
r
ess
io
n
s
,
t
o
en
h
an
ce
t
h
e
s
y
s
te
m
's r
eliab
il
it
y
ac
r
o
s
s
a
b
r
o
ad
er
r
an
g
e
o
f
em
o
tio
n
al
s
tates.
Ho
w
e
v
er
,
th
er
e
w
er
e
ch
alle
n
g
es
n
o
ted
in
t
h
e
r
ec
o
g
n
it
io
n
o
f
s
u
r
p
r
is
e
an
d
an
g
er
ca
u
s
i
n
g
t
h
e
ac
cu
r
ac
y
to
b
e
lo
w
,
w
h
ich
s
o
m
eti
m
e
s
co
n
f
u
s
ed
w
it
h
ea
c
h
o
th
er
.
Sp
ec
if
icall
y
,
a
n
g
r
y
ex
p
r
es
s
io
n
s
w
er
e
o
cc
asio
n
all
y
m
is
tak
e
n
f
o
r
s
ad
o
r
s
u
r
p
r
is
ed
,
an
d
v
ice
v
er
s
a.
A
d
d
itio
n
all
y
,
m
is
cla
s
s
i
f
icatio
n
s
o
f
s
ad
an
d
an
g
r
y
e
x
p
r
ess
io
n
s
s
o
m
eti
m
es r
es
u
lted
i
n
a
n
e
u
tr
a
l f
a
cial
e
x
p
r
ess
io
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
9
-
4864
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
,
Vo
l.
1
5
,
No
.
1
,
Ma
r
c
h
202
6
:
42
-
53
50
Fig
u
r
e
1
0
.
FER
d
etec
tio
n
ac
cu
r
ac
y
f
o
r
3
0
s
u
b
j
ec
ts
Fig
u
r
e
1
1
.
FER d
etec
tio
n
ac
cu
r
ac
y
f
o
r
5
d
if
f
er
en
t e
x
p
r
ess
io
n
s
4.
CO
NCLU
SI
O
N
T
h
e
s
tu
d
y
a
i
m
ed
to
e
v
al
u
ate
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
a
FE
R
s
y
s
t
e
m
o
n
m
o
b
ile
d
ev
ices
a
n
d
th
e
T
em
i
V
3
r
o
b
o
t,
f
o
cu
s
in
g
o
n
ac
cu
r
ac
y
,
laten
c
y
,
a
n
d
r
ea
l
-
ti
m
e
ap
p
licatio
n
s
.
T
h
e
FER
s
y
s
te
m
ac
h
iev
ed
a
v
a
lid
atio
n
ac
cu
r
ac
y
o
f
9
2
.
8
6
%,
s
u
r
p
ass
in
g
t
h
e
Vis
io
n
T
r
an
s
f
o
r
m
er
m
o
d
el
’
s
8
9
.
9
4
%
ac
cu
r
ac
y
.
P
er
f
o
r
m
an
ce
an
a
l
y
s
is
r
ev
ea
led
th
at
C
P
U
p
o
w
er
i
s
in
v
er
s
el
y
p
r
o
p
o
r
tio
n
al
to
ap
p
licatio
n
laten
c
y
,
h
ig
h
li
g
h
tin
g
t
h
e
n
ee
d
f
o
r
o
p
tim
izatio
n
tec
h
n
iq
u
e
s
to
en
h
an
ce
p
o
w
er
ef
f
icie
n
c
y
a
n
d
R
A
M
m
an
a
g
e
m
e
n
t
i
n
an
d
r
o
i
d
ap
p
licatio
n
s
.
T
h
e
o
p
tim
izatio
n
s
u
g
g
est
io
n
h
ad
b
ee
n
p
r
o
p
o
s
ed
in
th
e
r
e
s
u
lt
to
f
u
r
th
er
o
p
ti
m
ize
t
h
e
a
n
d
r
o
id
ap
p
licatio
n
d
ev
elo
p
ed
.
T
h
is
p
ap
er
h
as
p
r
esen
ted
t
h
e
p
r
eli
m
i
n
ar
y
r
e
s
u
l
t
o
n
th
e
te
s
t
o
f
t
h
e
a
n
d
r
o
id
ap
p
licatio
n
d
ev
elo
p
ed
in
to
T
em
i
V3
o
n
t
h
e
co
n
tr
o
lle
d
s
u
b
j
ec
t.
T
h
is
p
ar
t
o
f
th
e
r
esu
lt
w
i
ll
b
e
f
u
r
th
er
d
is
c
u
s
s
ed
o
n
th
e
co
n
ti
n
u
atio
n
o
f
th
is
p
ap
er
.
Fo
r
f
u
tu
r
e
w
o
r
k
,
ex
p
an
d
i
n
g
th
e
test
in
g
o
f
th
e
FER
s
y
s
te
m
ac
r
o
s
s
a
b
r
o
ad
er
r
an
g
e
o
f
m
o
d
el
ar
ch
itect
u
r
es,
in
cl
u
d
in
g
Mo
b
ileNetV2
,
Mo
b
ileNetV3
,
an
d
E
f
f
icien
tNet,
is
r
ec
o
m
m
e
n
d
ed
to
ev
alu
ate
tr
ad
e
-
of
f
s
b
et
w
ee
n
ac
cu
r
ac
y
a
n
d
r
eso
u
r
ce
ef
f
icie
n
c
y
.
I
n
co
r
p
o
r
atin
g
lar
g
er
an
d
m
o
r
e
d
iv
er
s
e
d
atasets
,
s
u
c
h
as
Af
f
ec
tNet
w
ill
h
elp
i
m
p
r
o
v
e
th
e
g
en
er
aliza
tio
n
o
f
th
e
m
o
d
el
ac
r
o
s
s
ag
e
g
r
o
u
p
s
,
eth
n
icities
,
an
d
lig
h
ti
n
g
co
n
d
itio
n
s
.
F
u
r
th
er
,
th
e
ad
o
p
tio
n
o
f
T
en
s
o
r
Flo
w
L
i
te
an
d
o
th
er
lig
h
t
w
ei
g
h
t
d
ep
lo
y
m
e
n
t
f
r
a
m
e
w
o
r
k
s
s
h
o
u
ld
co
n
tin
u
e
to
b
e
ex
p
lo
r
ed
f
o
r
o
p
ti
m
izi
n
g
i
n
f
er
e
n
ce
o
n
ed
g
e
an
d
e
m
b
ed
d
ed
s
y
s
te
m
s
.
F
u
t
u
r
e
ef
f
o
r
ts
m
a
y
al
s
o
in
v
o
l
v
e
b
en
c
h
m
ar
k
in
g
t
h
e
FE
R
s
y
s
te
m
o
n
ad
d
itio
n
al
e
m
b
e
d
d
ed
p
latf
o
r
m
s
s
u
ch
as
R
asp
b
er
r
y
P
i,
E
d
g
e
T
P
U
,
o
r
NVI
DI
A
J
etso
n
Na
n
o
to
ass
es
s
cr
o
s
s
-
p
lat
f
o
r
m
p
er
f
o
r
m
an
ce
u
n
d
er
r
ea
l
-
w
o
r
ld
co
n
s
tr
ain
ts
.
T
h
is
w
o
u
ld
en
ab
le
th
e
d
ev
elo
p
m
e
n
t
o
f
m
o
r
e
r
o
b
u
s
t,
ad
ap
tiv
e
FER
s
y
s
t
e
m
s
ca
p
ab
le
o
f
h
an
d
li
n
g
v
ar
i
ab
le
en
v
ir
o
n
m
en
ts
.
T
h
ese
i
m
p
r
o
v
e
m
e
n
t
s
w
ill
e
n
h
an
ce
th
e
s
y
s
te
m
's
r
eliab
ilit
y
a
n
d
u
s
ab
ili
t
y
ac
r
o
s
s
ap
p
licatio
n
d
o
m
ai
n
s
s
u
c
h
as
ass
is
tiv
e
r
o
b
o
tics
,
e
m
o
tio
n
-
a
w
ar
e
lear
n
in
g
e
n
v
ir
o
n
m
en
ts
,
an
d
m
o
b
ile
s
ec
u
r
it
y
s
y
s
te
m
s
.
T
h
e
f
in
d
i
n
g
s
o
f
th
is
s
tu
d
y
la
y
a
s
o
lid
f
o
u
n
d
atio
n
f
o
r
s
u
c
h
ad
v
a
n
ce
m
en
t
s
,
u
n
d
e
r
s
co
r
in
g
t
h
e
tr
an
s
f
o
r
m
ati
v
e
p
o
ten
tial
o
f
FER
i
n
en
ab
lin
g
m
o
r
e
r
esp
o
n
s
iv
e
a
n
d
in
t
u
iti
v
e
HR
I
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
R
ec
o
n
f
i
g
u
r
ab
le
&
E
m
b
ed
d
ed
Sy
s
t
I
SS
N:
2089
-
4864
Dep
lo
yme
n
t a
n
d
ev
a
lu
a
tio
n
o
f
fa
cia
l e
xp
r
ess
io
n
r
ec
o
g
n
itio
n
o
n
A
n
d
r
o
id
…
(
Mo
h
a
ma
d
Ha
r
iz
N
a
z
a
mid
)
51
ACK
NO
WL
E
D
G
M
E
NT
S
T
h
e
au
th
o
r
s
w
o
u
ld
lik
e
to
ex
p
r
ess
th
eir
s
i
n
ce
r
e
g
r
atit
u
d
e
to
Un
i
v
er
s
iti
T
ek
n
o
lo
g
i
M
AR
A
(
UiT
M)
f
o
r
p
r
o
v
id
in
g
r
esear
ch
s
u
p
p
o
r
t
an
d
g
r
an
t
f
ac
ilit
ie
s
;
Kizz
u
Kid
s
R
eh
ab
ili
tatio
n
C
e
n
tr
e
f
o
r
g
e
n
er
o
u
s
l
y
s
p
o
n
s
o
r
in
g
th
e
T
em
i
V3
r
o
b
o
t,
w
h
ic
h
p
lay
ed
a
p
iv
o
tal
r
o
le
in
test
in
g
an
d
s
o
f
t
w
ar
e
d
ev
elo
p
m
e
n
t;
an
d
E
UR
E
K
A
R
o
b
o
tics
C
en
tr
e,
C
ar
d
i
f
f
Me
tr
o
p
o
litan
Un
i
v
er
s
it
y
,
UK,
f
o
r
th
e
ir
v
al
u
ab
le
r
esear
ch
co
llab
o
r
atio
n
.
F
UNDIN
G
I
NF
O
RM
AT
I
O
N
T
h
is
w
o
r
k
w
as
f
u
n
d
ed
b
y
th
e
Stra
teg
ic
R
esear
c
h
P
ar
tn
er
s
h
i
p
(
SR
P
)
g
r
an
t
(
1
0
0
-
R
MC
5
/3
/SR
P
I
NT
(
0
4
7
/2
0
2
3
)
an
d
1
0
0
-
R
MC
5
/3
/SR
P
I
NT
(
0
4
8
/2
0
2
3
)
)
f
r
o
m
t
h
e
Un
iv
er
s
iti
T
ek
n
o
lo
g
i
MA
R
A
(
UiT
M)
,
Ma
la
y
s
ia
.
AUTHO
R
CO
NT
RIB
UT
I
O
NS ST
A
T
E
M
E
NT
T
h
is
jo
u
r
n
al
u
s
e
s
th
e
C
o
n
tr
ib
u
to
r
R
o
les
T
ax
o
n
o
m
y
(
C
R
ed
iT
)
t
o
r
ec
o
g
n
ize
in
d
i
v
id
u
al
au
t
h
o
r
co
n
tr
ib
u
tio
n
s
,
r
ed
u
ce
au
t
h
o
r
s
h
ip
d
is
p
u
tes,
an
d
f
ac
ilit
ate
co
lla
b
o
r
atio
n
.
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
Mo
h
a
m
ad
Har
iz
B
in
Naz
a
m
id
✓
✓
✓
✓
✓
✓
✓
✓
✓
R
o
zita
J
ailan
i
✓
✓
✓
✓
✓
✓
✓
✓
✓
Nu
r
Kh
al
id
ah
Z
ak
ar
ia
✓
✓
✓
✓
✓
An
w
ar
P
.
P
.
A
b
d
u
l
Ma
j
ee
d
✓
✓
✓
✓
✓
C
:
C
o
n
c
e
p
t
u
a
l
i
z
a
t
i
o
n
M
:
M
e
t
h
o
d
o
l
o
g
y
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
d
a
t
i
o
n
Fo
:
Fo
r
mal
a
n
a
l
y
si
s
I
:
I
n
v
e
st
i
g
a
t
i
o
n
R
:
R
e
so
u
r
c
e
s
D
:
D
a
t
a
C
u
r
a
t
i
o
n
O
:
W
r
i
t
i
n
g
-
O
r
i
g
i
n
a
l
D
r
a
f
t
E
:
W
r
i
t
i
n
g
-
R
e
v
i
e
w
&
E
d
i
t
i
n
g
Vi
:
Vi
su
a
l
i
z
a
t
i
o
n
Su
:
Su
p
e
r
v
i
si
o
n
P
:
P
r
o
j
e
c
t
a
d
mi
n
i
st
r
a
t
i
o
n
Fu
:
Fu
n
d
i
n
g
a
c
q
u
i
si
t
i
o
n
CO
NF
L
I
C
T
O
F
I
N
T
E
R
E
S
T
ST
A
T
E
M
E
NT
Au
t
h
o
r
s
s
tate
n
o
co
n
f
lic
t o
f
i
n
t
er
est.
E
T
H
I
CAL AP
P
RO
V
AL
T
h
e
r
esear
ch
r
elate
d
to
h
u
m
a
n
u
s
e
h
as
b
ee
n
co
m
p
l
ied
w
it
h
all
th
e
r
elev
a
n
t
n
atio
n
al
r
eg
u
l
atio
n
s
an
d
in
s
t
itu
tio
n
al
p
o
licies
i
n
ac
co
r
d
an
ce
w
ith
t
h
e
ten
e
ts
o
f
th
e
He
ls
in
k
i
Dec
lar
atio
n
an
d
h
as
b
ee
n
ap
p
r
o
v
ed
b
y
th
e
Un
i
v
er
s
iti
T
ek
n
o
lo
g
i
M
AR
A
(
UiT
M)
E
th
ics
C
o
m
m
it
t
ee
u
n
d
er
ap
p
r
o
v
al
n
u
m
b
er
o
f
R
E
C
/0
6
/2
0
2
4
(
P
G/FB
/2
1
)
.
DATA AV
AI
L
AB
I
L
I
T
Y
Der
iv
ed
d
ata
s
u
p
p
o
r
tin
g
th
e
f
i
n
d
in
g
s
o
f
t
h
is
s
t
u
d
y
ar
e
a
v
ailab
le
f
r
o
m
th
e
co
r
r
esp
o
n
d
in
g
au
t
h
o
r
[
in
itial
s
,
R
J
]
o
n
r
eq
u
est.
RE
F
E
R
E
NC
E
S
[
1
]
A
.
O
b
a
i
g
b
e
n
a
e
t
a
l
.
,
“
A
I
a
n
d
h
u
m
a
n
-
r
o
b
o
t
i
n
t
e
r
a
c
t
i
o
n
:
A
r
e
v
i
e
w
o
f
r
e
c
e
n
t
a
d
v
a
n
c
e
s a
n
d
c
h
a
l
l
e
n
g
e
s,”
G
S
C
A
d
v
a
n
c
e
d
Re
se
a
rc
h
a
n
d
Re
v
i
e
w
s
,
v
o
l
.
1
8
,
n
o
.
2
,
p
p
.
3
2
1
–
3
3
0
,
2
0
2
4
,
d
o
i
:
1
0
.
3
0
5
7
4
/
g
s
c
a
r
r
.
2
0
2
4
.
1
8
.
2
.
0
0
7
0
.
[
2
]
B
.
S
a
w
i
k
e
t
a
l
.
,
“
R
o
b
o
t
s
f
o
r
El
d
e
r
l
y
C
a
r
e
:
R
e
v
i
e
w
,
M
u
l
t
i
-
C
r
i
t
e
r
i
a
O
p
t
i
mi
z
a
t
i
o
n
M
o
d
e
l
a
n
d
Q
u
a
l
i
t
a
t
i
v
e
C
a
se
S
t
u
d
y
,
”
H
e
a
l
t
h
c
a
r
e
,
v
o
l
.
1
1
,
n
o
.
9
,
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
h
e
a
l
t
h
c
a
r
e
1
1
0
9
1
2
8
6
.
[
3
]
Y
.
D
o
n
g
a
n
d
X
.
Z
h
o
u
,
“
A
d
v
a
n
c
e
m
e
n
t
s
i
n
A
I
-
d
r
i
v
e
n
mu
l
t
i
l
i
n
g
u
a
l
c
o
m
p
r
e
h
e
n
si
o
n
f
o
r
so
c
i
a
l
r
o
b
o
t
i
n
t
e
r
a
c
t
i
o
n
s:
A
n
e
x
t
e
n
si
v
e
r
e
v
i
e
w
,
”
El
e
c
t
ro
n
i
c
Re
se
a
r
c
h
Arc
h
i
v
e
,
v
o
l
.
3
1
,
n
o
.
1
1
,
p
p
.
6
6
0
0
–
6
6
3
3
,
2
0
2
3
,
d
o
i
:
1
0
.
3
9
3
4
/
e
r
a
.
2
0
2
3
3
3
4
.
[
4
]
M
.
S
p
e
z
i
a
l
e
t
t
i
,
G
.
P
l
a
c
i
d
i
,
a
n
d
S
.
R
o
ssi
,
“
Emo
t
i
o
n
R
e
c
o
g
n
i
t
i
o
n
f
o
r
H
u
man
-
R
o
b
o
t
I
n
t
e
r
a
c
t
i
o
n
:
R
e
c
e
n
t
A
d
v
a
n
c
e
s
a
n
d
F
u
t
u
r
e
P
e
r
sp
e
c
t
i
v
e
s,”
Fr
o
n
t
i
e
rs
i
n
R
o
b
o
t
i
c
s
a
n
d
AI
,
v
o
l
.
7
,
2
0
2
0
,
d
o
i
:
1
0
.
3
3
8
9
/
f
r
o
b
t
.
2
0
2
0
.
5
3
2
2
7
9
.
[
5
]
H
.
D
o
n
g
,
“
R
e
se
a
r
c
h
P
r
o
g
r
e
ss
a
n
d
R
e
v
i
e
w
o
n
S
e
r
v
i
c
e
I
n
t
e
r
a
c
t
i
o
n
b
e
t
w
e
e
n
I
n
t
e
l
l
i
g
e
n
t
S
e
r
v
i
c
e
R
o
b
o
t
s
a
n
d
C
u
s
t
o
me
r
s,”
i
n
Pro
c
e
e
d
i
n
g
s
-
2
0
2
3
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
S
e
r
v
i
c
e
R
o
b
o
t
i
c
s,
I
C
o
S
R
2
0
2
3
,
2
0
2
3
,
p
p
.
7
8
–
85
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
o
S
R
5
9
9
8
0
.
2
0
2
3
.
0
0
0
2
5
.
[
6
]
M
.
K
r
u
mn
i
k
l
a
n
d
V
.
M
a
i
w
a
l
d
,
“
F
a
c
i
a
l
Emo
t
i
o
n
R
e
c
o
g
n
i
t
i
o
n
f
o
r
M
o
b
i
l
e
D
e
v
i
c
e
s:
A
P
r
a
c
t
i
c
a
l
R
e
v
i
e
w
,
”
I
EEE
Ac
c
e
ss
,
v
o
l
.
1
2
,
p
p
.
1
5
7
3
5
–
1
5
7
4
7
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
4
.
3
3
5
8
4
5
5
.
Evaluation Warning : The document was created with Spire.PDF for Python.