I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
pu
t
er
E
ng
ineering
(
I
J
E
CE
)
Vo
l.
1
6
,
No
.
4
,
A
u
g
u
s
t
20
2
6
,
p
p
.
1
8
9
9
~
1
9
1
2
I
SS
N:
2088
-
8
7
0
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijece.
v
1
6
i
4
.
pp
1
8
9
9
-
1
9
1
2
1899
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
Rea
l
-
time
f
a
cia
l a
nd bo
dy
pos
e emo
tion recog
nition
f
o
r children
with
a
ut
ism
bas
e
d on Y
O
LOv8
an
d LST
M
Siti
Nuro
hm
a
hwa
t
i
1
,
Ana
nd
a
P
utr
a
K
a
nieza
2
,
Ade
Rif
k
y
Set
ia
wa
n
3
,
Ahm
a
d F
a
dla
n
2
1
D
e
p
a
r
t
me
n
t
o
f
P
h
y
si
c
s E
d
u
c
a
t
i
o
n
,
F
a
c
u
l
t
y
o
f
Te
a
c
h
e
r
Tr
a
i
n
i
n
g
a
n
d
Ed
u
c
a
t
i
o
n
,
S
e
b
e
l
a
s
M
a
r
e
t
U
n
i
v
e
r
si
t
y
,
S
u
r
a
k
a
r
t
a
,
I
n
d
o
n
e
s
i
a
2
D
e
p
a
r
t
m
e
n
t
o
f
E
l
e
c
t
r
i
c
a
l
E
n
g
i
n
e
e
r
i
n
g
,
F
a
c
u
l
t
y
o
f
En
g
i
n
e
e
r
i
n
g
,
S
e
b
e
l
a
s
M
a
r
e
t
U
n
i
v
e
r
s
i
t
y
,
S
u
r
a
k
a
r
t
a
,
I
n
d
o
n
e
si
a
3
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
mat
i
c
s
E
n
g
i
n
e
e
r
i
n
g
,
V
o
c
a
t
i
o
n
a
l
S
c
h
o
o
l
s,
S
e
b
e
l
a
s
M
a
r
e
t
U
n
i
v
e
r
si
t
y
,
S
u
r
a
k
a
r
t
a
,
I
n
d
o
n
e
s
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Dec
1
6
,
2
0
2
5
R
ev
is
ed
Ma
y
1
1
,
2
0
2
6
Acc
ep
ted
J
u
l 2
2
,
2
0
2
6
Ch
il
d
re
n
wit
h
a
u
ti
sm
sp
e
c
tru
m
d
iso
r
d
e
r
(ASD)
o
fte
n
fa
c
e
c
h
a
ll
e
n
g
e
s
i
n
re
c
o
g
n
izin
g
a
n
d
e
x
p
re
ss
in
g
e
m
o
ti
o
n
s,
wh
ic
h
c
a
n
a
ffe
c
t
th
e
ir
b
e
h
a
v
io
r
a
n
d
p
a
rti
c
ip
a
ti
o
n
in
i
n
c
lu
si
v
e
c
las
sro
o
m
e
n
v
iro
n
m
e
n
ts.
Th
is
stu
d
y
p
r
o
p
o
se
s
a
re
a
l
-
ti
m
e
m
u
lt
imo
d
a
l
e
m
o
ti
o
n
re
c
o
g
n
i
ti
o
n
s
y
ste
m
in
teg
ra
ti
n
g
d
e
e
p
lea
rn
i
n
g
a
n
d
In
tern
e
t
o
f
Th
i
n
g
s
(Io
T)
tec
h
n
o
l
o
g
ies
to
su
p
p
o
rt
e
a
rly
e
m
o
ti
o
n
a
l
m
o
n
it
o
r
in
g
i
n
c
h
il
d
re
n
wi
th
AS
D.
Th
e
fra
m
e
wo
rk
c
o
m
b
in
e
s
Y
OLOv
8
f
o
r
fa
c
ial
e
x
p
re
ss
io
n
d
e
tec
ti
o
n
a
n
d
YO
LOv
8
-
b
a
se
d
p
o
se
e
stim
a
ti
o
n
fo
r
b
o
d
y
m
o
v
e
m
e
n
t
a
n
a
ly
si
s,
a
l
o
n
g
with
a
lo
n
g
sh
o
rt
-
term
m
e
m
o
ry
(LS
T
M
)
n
e
two
rk
fo
r
tem
p
o
ra
l
e
m
o
ti
o
n
c
las
sifica
ti
o
n
.
At
t
h
e
fa
c
ial
lev
e
l,
th
e
sy
ste
m
re
c
o
g
n
ize
s
fiv
e
e
m
o
ti
o
n
a
l
sta
tes
:
sa
d
,
h
a
p
p
y
,
n
e
u
tral,
b
o
re
d
o
m
,
a
n
d
tan
tru
m
.
At
th
e
tem
p
o
ra
l
lev
e
l,
t
h
e
LS
T
M
m
o
d
e
l
c
las
sifies
b
e
h
a
v
i
o
ra
l
se
q
u
e
n
c
e
s
in
to
th
re
e
c
a
teg
o
ries
:
n
e
u
tral/
b
o
re
d
,
h
a
p
p
y
,
a
n
d
tan
tr
u
m
,
e
n
a
b
li
n
g
h
i
e
ra
rc
h
ica
l
e
m
o
ti
o
n
i
n
terp
re
tati
o
n
fr
o
m
in
sta
n
tan
e
o
u
s
e
x
p
re
ss
io
n
s
to
tem
p
o
ra
l
p
a
tt
e
rn
s.
Ex
p
e
rime
n
tal
re
su
lt
s
sh
o
w
t
h
a
t
t
h
e
fa
c
ial
e
x
p
re
ss
io
n
m
o
d
e
l
a
c
h
i
e
v
e
s
9
2
%
p
re
c
isio
n
,
w
h
i
le
t
h
e
LS
T
M
-
b
a
se
d
c
las
sifier
re
a
c
h
e
s
9
5
%
p
e
a
k
v
a
li
d
a
ti
o
n
a
c
c
u
ra
c
y
a
n
d
9
3
.
3
3
%
fi
n
a
l
tes
t
a
c
c
u
ra
c
y
.
Th
e
sy
ste
m
is
d
e
p
l
o
y
e
d
o
n
a
we
b
-
b
a
se
d
m
o
n
i
to
ri
n
g
p
latf
o
rm
wit
h
re
a
l
-
ti
m
e
n
o
t
ifi
c
a
ti
o
n
s
fo
r
e
d
u
c
a
to
rs
a
n
d
p
a
re
n
ts.
T
h
e
p
r
o
p
o
se
d
a
p
p
ro
a
c
h
d
e
m
o
n
stra
tes
e
ffe
c
ti
v
e
n
e
ss
in
p
ro
v
id
i
n
g
ti
m
e
ly
e
m
o
ti
o
n
a
l
i
n
sig
h
ts
to
su
p
p
o
rt
e
a
rly
in
ter
v
e
n
ti
o
n
a
n
d
imp
ro
v
e
in
c
lu
siv
e
e
d
u
c
a
ti
o
n
f
o
r
c
h
il
d
re
n
with
ASD.
K
ey
w
o
r
d
s
:
Au
tis
m
s
p
ec
tr
u
m
d
is
o
r
d
er
E
m
o
tio
n
d
etec
tio
n
I
n
clu
s
iv
e
ed
u
ca
tio
n
L
o
n
g
s
h
o
r
t
-
ter
m
m
e
m
o
r
y
R
ea
l
-
tim
e
m
o
n
ito
r
in
g
YOL
Ov
8
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
:
Sit
i N
u
r
o
h
m
ah
wati
Dep
ar
tm
en
t o
f
Ph
y
s
ics E
d
u
ca
t
io
n
,
Facu
lty
o
f
T
ea
ch
er
T
r
ain
i
n
g
an
d
E
d
u
ca
tio
n
,
Seb
elas M
a
r
et
Un
iv
er
s
ity
Ken
tin
g
an
,
3
6
I
r
.
Su
tam
i Ro
ad
,
J
eb
r
es Dist
r
ict,
Su
r
ak
ar
ta
5
7
1
2
6
,
I
n
d
o
n
esia
E
m
ail: r
ah
m
awa
ttii
4
5
@
s
tu
d
en
t.u
n
s
.
ac
.
id
1.
I
NT
RO
D
UCT
I
O
N
Au
tis
m
s
p
ec
tr
u
m
d
is
o
r
d
er
(
ASD)
is
a
n
e
u
r
o
d
e
v
elo
p
m
en
t
al
co
n
d
itio
n
c
h
ar
ac
ter
ized
b
y
en
d
u
r
in
g
ch
allen
g
es
in
s
o
cial
co
m
m
u
n
i
ca
tio
n
an
d
i
n
ter
ac
tio
n
,
alo
n
g
with
th
e
p
r
esen
ce
o
f
r
estricte
d
,
r
ep
etitiv
e
p
atter
n
s
o
f
b
eh
av
io
r
,
i
n
ter
ests
,
o
r
a
ctiv
ities
[
1
]
.
ASD
is
a
d
e
v
elo
p
m
en
tal
d
is
o
r
d
e
r
th
at
s
ig
n
if
ican
tly
a
f
f
ec
ts
co
m
m
u
n
icatio
n
,
b
eh
av
io
r
,
a
n
d
s
o
cial
in
ter
ac
tio
n
[
2
]
.
R
ec
en
t
g
lo
b
al
esti
m
ates
o
f
au
tis
m
s
p
ec
tr
u
m
d
is
o
r
d
e
r
p
r
ev
alen
ce
s
h
o
w
v
ar
iab
ilit
y
a
cr
o
s
s
s
tu
d
ies,
with
r
ep
o
r
ted
r
ates
r
an
g
in
g
ap
p
r
o
x
im
ately
f
r
o
m
6
0
to
1
0
0
p
er
1
0
,
0
0
0
ch
ild
r
en
b
ased
o
n
lar
g
e
-
s
ca
le
r
ev
iews
an
d
m
eta
-
an
al
y
s
es
co
n
d
u
cted
ac
r
o
s
s
m
u
ltip
le
co
u
n
tr
ies
b
etwe
en
2
0
0
8
an
d
2
0
2
1
[
3
]
.
C
h
ild
r
en
with
ASD
co
m
m
o
n
ly
p
r
esen
t
d
if
f
er
en
ce
s
in
la
n
g
u
a
g
e
c
o
m
p
r
eh
en
s
io
n
a
n
d
s
o
cial
co
m
m
u
n
icatio
n
,
alo
n
g
s
id
e
at
y
p
ical
s
o
cial
in
ter
ac
tio
n
p
atter
n
s
,
in
clu
d
in
g
r
e
d
u
ce
d
u
s
e
o
f
g
estu
r
es
an
d
n
o
n
v
e
r
b
al
b
e
h
av
io
r
s
,
as
well
as
d
if
f
icu
lties
in
m
o
to
r
c
o
o
r
d
in
atio
n
[
4
]
.
T
h
ese
ch
all
en
g
es
ar
e
f
u
r
th
e
r
co
m
p
o
u
n
d
e
d
b
y
th
e
ea
r
ly
em
er
g
en
ce
o
f
c
h
ar
ac
ter
is
tic
au
tis
tic
f
ea
tu
r
es
in
ch
ild
h
o
o
d
,
wh
i
ch
p
er
s
is
t
o
v
er
tim
e
an
d
co
n
tr
ib
u
te
to
s
ig
n
if
ican
t i
m
p
air
m
en
ts
in
d
aily
f
u
n
cti
o
n
i
n
g
[
5
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
1
6
,
No
.
4
,
Au
g
u
s
t
20
2
6
:
1
8
9
9
-
1
912
1900
I
n
in
clu
s
iv
e
ed
u
ca
tio
n
al
s
ettin
g
s
,
id
en
tify
in
g
an
d
i
n
ter
p
r
etin
g
th
e
em
o
tio
n
al
s
tates
o
f
ch
ild
r
en
wit
h
ASD
r
em
ain
s
a
s
ig
n
if
ican
t
ch
allen
g
e
f
o
r
ed
u
ca
to
r
s
[
6
]
.
Data
f
r
o
m
t
h
e
I
n
d
o
n
esian
Min
is
tr
y
o
f
E
d
u
ca
tio
n
an
d
C
u
ltu
r
e
(
2
0
2
1
)
r
e
p
o
r
ts
2
,
2
7
0
s
p
ec
ial
s
ch
o
o
ls
,
2
7
,
2
6
5
te
ac
h
er
s
,
an
d
1
3
3
,
8
2
6
s
tu
d
e
n
t
s
with
d
is
ab
ilit
ie
s
,
r
esu
ltin
g
in
an
im
b
alan
ce
b
etwe
en
av
ailab
le
s
u
p
p
o
r
t
a
n
d
s
tu
d
en
t
n
ee
d
s
[
7
]
.
T
h
is
lim
itatio
n
r
ed
u
ce
s
th
e
ab
ilit
y
o
f
teac
h
er
s
to
p
r
o
v
id
e
i
n
d
iv
i
d
u
alize
d
atten
tio
n
an
d
r
ea
l
-
ti
m
e
em
o
tio
n
al
m
o
n
ito
r
in
g
.
C
o
n
s
eq
u
en
tly
,
b
o
t
h
ed
u
ca
to
r
s
an
d
p
a
r
en
ts
f
ac
e
d
if
f
icu
lties
in
r
esp
o
n
d
in
g
ap
p
r
o
p
r
iately
to
ch
ild
r
en
'
s
em
o
tio
n
al
co
n
d
itio
n
s
[
8
]
,
[
9
]
,
[
1
0
]
.
T
h
e
p
r
e
v
alen
ce
o
f
th
ese
b
eh
av
io
r
s
is
s
u
b
s
tan
tially
h
ig
h
er
in
ch
ild
r
e
n
with
ASD
wit
h
ap
p
r
o
x
im
ately
o
f
35
%
–
5
0
%
co
m
p
a
r
ed
to
th
o
s
e
with
in
tellectu
al
d
is
ab
ili
ties
a
b
o
u
t
5
%
–
1
7
%,
m
ak
i
n
g
it
two
to
th
r
ee
tim
es
m
o
r
e
co
m
m
o
n
i
n
th
e
ASD
p
o
p
u
lati
o
n
[
1
1
]
.
T
h
is
h
ig
h
lig
h
ts
th
e
u
r
g
en
t
n
ee
d
f
o
r
ef
f
ec
tiv
e
to
o
ls
to
s
u
p
p
o
r
t
em
o
tio
n
al
m
o
n
ito
r
in
g
an
d
ea
r
ly
in
ter
v
en
t
io
n
.
R
ec
en
t
ad
v
an
ce
s
in
tech
n
o
lo
g
y
,
p
ar
tic
u
lar
ly
in
ass
is
tiv
e
an
d
in
tellig
en
t
s
y
s
tem
s
,
h
a
v
e
s
h
o
wn
s
ig
n
if
ican
t
p
o
te
n
tial
in
im
p
r
o
v
in
g
ac
ce
s
s
ib
ilit
y
an
d
en
g
a
g
em
en
t
f
o
r
ch
ild
r
e
n
with
d
is
ab
ilit
ies
[
1
2
]
,
[
1
3
]
.
Sev
er
al
s
tu
d
ies
h
av
e
e
x
p
lo
r
e
d
em
o
tio
n
r
ec
o
g
n
itio
n
u
s
in
g
im
ag
e
p
r
o
ce
s
s
in
g
an
d
d
ee
p
lear
n
in
g
a
p
p
r
o
ac
h
es
.
Pre
v
io
u
s
s
tu
d
ies
h
av
e
in
tr
o
d
u
ce
d
r
ea
l
-
tim
e
e
m
o
tio
n
d
etec
tio
n
s
y
s
tem
s
u
s
in
g
s
tatic
im
ag
es
o
r
v
id
e
o
-
b
ased
ap
p
r
o
ac
h
es;
h
o
wev
e
r
,
t
h
eir
p
er
f
o
r
m
a
n
ce
r
e
m
ain
s
lim
ited
i
n
d
y
n
am
ic
class
r
o
o
m
en
v
ir
o
n
m
en
ts
[
1
4
]
.
Oth
er
wo
r
k
s
em
p
lo
y
e
d
im
ag
e
p
r
o
ce
s
s
in
g
tech
n
iq
u
es
f
o
r
f
ac
ial
ex
p
r
ess
io
n
r
ec
o
g
n
itio
n
[
1
5
]
,
[
1
6
]
;
h
o
wev
e
r
,
th
ese
m
eth
o
d
s
r
e
q
u
ir
e
co
n
tr
o
lled
c
o
n
d
itio
n
s
,
s
u
c
h
as
d
ir
ec
t
f
ac
e
o
r
ien
tatio
n
to
war
d
th
e
ca
m
er
a,
r
e
d
u
cin
g
th
ei
r
r
o
b
u
s
tn
ess
in
r
ea
l
-
wo
r
ld
s
ce
n
ar
io
s
.
I
n
ad
d
itio
n
,
m
o
s
t
e
x
is
tin
g
a
p
p
r
o
ac
h
es
r
el
y
o
n
s
in
g
le
-
m
o
d
ality
d
ata
an
d
o
f
ten
s
tr
u
g
g
le
to
ac
cu
r
ately
d
etec
t
s
u
b
tle
o
r
m
ix
ed
e
m
o
tio
n
al
ex
p
r
ess
io
n
s
.
Fu
r
th
er
m
o
r
e,
t
h
e
lack
o
f
r
ea
l
-
tim
e
m
o
n
ito
r
in
g
an
d
in
teg
r
atio
n
wi
th
co
m
m
u
n
icatio
n
p
latf
o
r
m
s
l
im
its
th
eir
p
r
ac
tical
ap
p
licab
il
ity
f
o
r
co
n
tin
u
o
u
s
em
o
tio
n
al
tr
ac
k
i
n
g
in
class
r
o
o
m
en
v
ir
o
n
m
en
ts
.
T
o
ad
d
r
ess
th
ese
lim
itatio
n
s
,
t
h
is
s
tu
d
y
p
r
o
p
o
s
es th
e
d
ev
elo
p
m
en
t o
f
a
m
u
ltimo
d
al
r
ea
l
-
ti
m
e
em
o
tio
n
d
etec
tio
n
s
y
s
tem
b
ased
o
n
d
ee
p
lear
n
in
g
an
d
I
n
ter
n
et
o
f
T
h
i
n
g
s
(
I
o
T
)
tech
n
o
l
o
g
ies.
T
h
e
s
y
s
tem
in
teg
r
ates y
o
u
o
n
ly
lo
o
k
o
n
ce
(
YOL
Ov
8
)
f
o
r
f
ac
ial
ex
p
r
ess
io
n
d
etec
tio
n
,
YOL
Ov
8
-
b
ased
k
ey
p
o
in
t
d
etec
tio
n
f
o
r
b
o
d
y
m
o
v
em
en
t
an
aly
s
is
,
an
d
lo
n
g
s
h
o
r
t
-
ter
m
m
e
m
o
r
y
(
L
STM
)
f
o
r
tem
p
o
r
al
em
o
tio
n
class
if
icatio
n
.
T
h
e
s
y
s
tem
is
im
p
lem
en
ted
o
n
a
R
asp
b
er
r
y
Pi
p
latf
o
r
m
eq
u
ip
p
ed
with
a
ca
m
er
a
f
o
r
r
ea
l
-
tim
e
d
ata
ac
q
u
is
itio
n
[
1
7
]
.
T
h
e
p
r
o
ce
s
s
ed
d
ata
ar
e
tr
an
s
m
itted
th
r
o
u
g
h
a
n
I
o
T
-
b
ased
i
n
f
r
astr
u
ctu
r
e
an
d
v
is
u
alize
d
o
n
a
we
b
-
b
ased
d
ash
b
o
ar
d
,
with
ad
d
itio
n
al
n
o
tific
atio
n
s
u
p
p
o
r
t to
p
r
o
v
id
e
r
ea
l
-
tim
e
u
p
d
ates to
ed
u
ca
to
r
s
an
d
p
ar
e
n
ts
.
T
h
e
m
ain
co
n
tr
ib
u
tio
n
s
o
f
th
i
s
s
tu
d
y
ar
e
as
f
o
llo
ws:
i
)
t
h
e
i
n
teg
r
atio
n
o
f
m
u
ltimo
d
al
v
is
u
al
f
ea
tu
r
es
co
m
b
in
in
g
f
ac
ial
an
d
b
o
d
y
m
o
v
em
en
t
d
ata
f
o
r
im
p
r
o
v
e
d
e
m
o
tio
n
r
ec
o
g
n
itio
n
ac
cu
r
ac
y
,
i
i
)
th
e
a
p
p
licatio
n
o
f
L
STM
to
ca
p
tu
r
e
tem
p
o
r
al
e
m
o
tio
n
al
p
atter
n
s
in
ch
ild
r
en
with
ASD,
an
d
iii
)
th
e
d
e
v
elo
p
m
en
t
o
f
a
r
ea
l
-
tim
e
m
o
n
ito
r
in
g
s
y
s
tem
t
h
at
s
u
p
p
o
r
ts
ea
r
ly
i
n
ter
v
en
tio
n
in
in
clu
s
iv
e
ed
u
ca
tio
n
al
en
v
ir
o
n
m
e
n
t
s
.
Un
lik
e
p
r
ev
io
u
s
s
tu
d
ies
th
at
p
r
im
ar
ily
r
ely
o
n
s
in
g
le
-
m
o
d
ality
f
ac
ial
an
aly
s
is
,
th
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
co
m
b
in
es
f
ac
ial
ex
p
r
ess
io
n
r
ec
o
g
n
itio
n
,
b
o
d
y
k
ey
p
o
i
n
t
d
etec
tio
n
,
tem
p
o
r
al
m
o
d
elin
g
,
an
d
I
o
T
-
b
ased
m
o
n
ito
r
in
g
in
a
u
n
if
ied
r
ea
l
-
tim
e
s
y
s
tem
.
T
h
is
s
tu
d
y
a
im
s
to
en
h
an
ce
em
o
tio
n
al
awa
r
en
ess
,
r
ed
u
ce
th
e
r
is
k
o
f
s
elf
-
in
ju
r
io
u
s
b
eh
av
io
r
,
an
d
s
u
p
p
o
r
t m
o
r
e
ef
f
ec
tiv
e
lear
n
in
g
e
x
p
er
ien
ce
s
f
o
r
ch
ild
r
en
with
ASD
[
1
8
]
.
2.
M
E
T
H
O
D
2
.
1
.
Resea
rc
h desi
g
n a
n
d da
t
a
s
et
a
cquis
it
io
n
T
h
is
s
tu
d
y
ad
o
p
ts
a
r
esear
ch
an
d
d
ev
elo
p
m
en
t
(
R
&
D)
ap
p
r
o
ac
h
to
d
esig
n
an
d
im
p
lem
en
t
a
r
ea
l
-
tim
e
em
o
tio
n
d
etec
tio
n
s
y
s
tem
f
o
r
ch
ild
r
en
with
ASD.
T
h
e
r
e
s
ea
r
ch
is
co
n
d
u
cted
i
n
a
ch
r
o
n
o
lo
g
ical
p
r
o
ce
s
s
co
n
s
is
tin
g
o
f
d
ata
ac
q
u
is
itio
n
,
p
r
ep
r
o
ce
s
s
in
g
,
m
o
d
el
d
ev
elo
p
m
en
t,
tr
ain
in
g
,
an
d
s
y
s
tem
d
ep
lo
y
m
e
n
t.
T
h
e
d
ataset
u
tili
ze
d
in
th
is
r
esear
c
h
was
o
b
tain
ed
f
r
o
m
two
p
r
im
ar
y
s
o
u
r
ce
s
.
T
h
e
f
ir
s
t
s
o
u
r
ce
i
n
v
o
lv
e
d
d
ir
ec
t
d
ata
co
llectio
n
f
r
o
m
au
tis
tic
ch
ild
r
en
at
Har
m
o
n
y
Su
r
ak
ar
ta
Sp
ec
ial
E
d
u
ca
tio
n
Sch
o
o
l,
wh
ile
th
e
s
ec
o
n
d
co
m
p
r
is
ed
p
u
b
licly
av
ailab
le
d
atasets
r
etr
iev
ed
f
r
o
m
t
h
e
K
ag
g
le
p
latf
o
r
m
.
A
to
tal
o
f
6
.
8
3
9
f
ac
ial
ex
p
r
ess
io
n
im
ag
es
wer
e
co
m
p
iled
,
r
ep
r
e
s
en
tin
g
v
ar
i
o
u
s
em
o
tio
n
al
s
ta
tes
in
clu
d
in
g
h
ap
p
in
ess
,
s
ad
n
ess
,
n
eu
tr
ality
,
a
n
d
an
g
er
.
T
h
ese
d
atasets
wer
e
p
r
ep
r
o
ce
s
s
ed
an
d
an
n
o
tated
to
e
n
s
u
r
e
co
n
s
is
ten
cy
an
d
q
u
ality
b
ef
o
r
e
b
ei
n
g
u
s
ed
f
o
r
m
o
d
el
tr
ain
in
g
.
T
h
e
f
ac
ial
ex
p
r
ess
io
n
im
ag
es
wer
e
s
u
b
s
eq
u
en
tly
u
s
ed
as
in
p
u
t
f
o
r
t
h
e
YOL
Ov
8
o
b
ject
d
etec
tio
n
m
o
d
el
to
p
e
r
f
o
r
m
e
m
o
tio
n
class
if
icatio
n
.
T
h
e
d
ataset
d
is
tr
ib
u
tio
n
is
illu
s
tr
ated
in
Fig
u
r
e
1
.
Fo
r
b
o
d
y
m
o
v
em
en
t
an
aly
s
is
,
th
is
s
tu
d
y
em
p
l
o
y
s
th
e
YO
L
Ov
8
p
o
s
e
esti
m
atio
n
m
o
d
el
to
ex
tr
ac
t
s
p
atial
k
ey
p
o
in
ts
f
r
o
m
v
id
e
o
d
ata.
T
h
e
m
o
d
el
is
tr
ain
ed
u
s
in
g
th
e
co
m
m
o
n
o
b
jects
in
co
n
t
ex
t
(
C
OC
O)
d
ataset
f
o
r
m
at,
wh
ich
d
e
f
in
es
1
7
h
u
m
an
b
o
d
y
k
ey
p
o
in
ts
,
in
clu
d
in
g
th
e
n
o
s
e,
ey
es,
ea
r
s
,
s
h
o
u
ld
er
s
,
elb
o
ws,
wr
is
ts
,
h
ip
s
,
k
n
ee
s
,
a
n
d
a
n
k
les.
T
h
e
YOL
Ov
8
p
o
s
e
m
o
d
el
is
u
ti
lized
to
au
to
m
atica
lly
d
etec
t
an
d
e
x
tr
ac
t
b
o
d
y
lan
d
m
ar
k
s
f
r
o
m
v
id
e
o
r
ec
o
r
d
i
n
g
s
o
f
ch
ild
r
e
n
with
ASD.
T
h
ese
k
ey
p
o
in
ts
r
ep
r
esen
t
s
p
atial
f
ea
tu
r
es
o
f
b
o
d
y
m
o
v
em
en
ts
an
d
ar
e
u
s
ed
as
in
p
u
t
f
o
r
s
u
b
s
eq
u
e
n
t
tem
p
o
r
al
a
n
aly
s
is
.
T
h
e
s
tr
u
ctu
r
e
o
f
th
e
k
ey
p
o
in
t
a
n
n
o
tatio
n
s
is
illu
s
tr
ated
in
Fig
u
r
e
2
.
T
o
class
if
y
em
o
tio
n
-
r
elate
d
m
o
v
em
en
t
p
atter
n
s
,
v
id
eo
d
a
tasets
wer
e
co
llected
th
r
o
u
g
h
two
m
ain
m
eth
o
d
s
:
d
ir
ec
t
r
ec
o
r
d
in
g
s
at
Har
m
o
n
y
Su
r
a
k
ar
ta
Sp
ec
ial
E
d
u
ca
tio
n
Sch
o
o
l
an
d
c
u
r
at
ed
v
id
e
o
m
ater
ials
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
R
ea
l
-
time
fa
cia
l a
n
d
b
o
d
y
p
o
s
e
emo
tio
n
r
ec
o
g
n
itio
n
f
o
r
ch
ild
r
en
…
(
S
iti N
u
r
o
h
ma
h
w
a
ti
)
1901
o
b
tain
ed
f
r
o
m
o
n
lin
e
s
o
u
r
ce
s
.
T
h
e
v
id
eo
s
ca
p
tu
r
e
d
v
ar
io
u
s
ac
tiv
ities
an
d
em
o
tio
n
al
b
eh
av
io
r
s
ex
h
ib
ited
b
y
ch
ild
r
en
with
ASD.
T
h
e
co
lle
cted
v
id
eo
s
wer
e
s
eg
m
en
ted
i
n
to
s
eq
u
en
ce
s
an
d
m
an
u
ally
an
n
o
tated
b
ased
o
n
f
ac
ial
ex
p
r
ess
io
n
s
,
b
o
d
y
m
o
v
e
m
en
ts
,
an
d
co
n
tex
t
u
al
b
e
h
av
io
r
al
cu
es.
T
h
is
a
n
n
o
tatio
n
p
r
o
c
ess
en
s
u
r
es
th
at
th
e
s
eq
u
en
tial
d
ata
ac
c
u
r
ately
r
ep
r
esen
ts
d
iv
er
s
e
em
o
tio
n
al
s
tates.
T
h
e
p
r
ep
ar
e
d
s
eq
u
e
n
ce
s
ar
e
th
en
s
tr
u
ctu
r
ed
as
tim
e
-
s
er
ies d
ata
to
b
e
u
s
ed
as i
n
p
u
t f
o
r
th
e
L
STM
m
o
d
el.
Fig
u
r
e
1
.
Facial
ex
p
r
ess
io
n
d
a
taset c
o
n
s
is
tin
g
o
f
m
u
ltip
le
ca
t
eg
o
r
ies
Fig
u
r
e
2
.
Key
p
o
in
ts
r
ep
r
esen
tatio
n
b
ased
o
n
th
e
C
OC
O
d
ataset
co
v
er
in
g
1
7
p
o
in
ts
o
f
h
u
m
an
b
o
d
y
2
.
2
.
YO
L
O
v
8
-
ba
s
ed
f
a
cia
l e
x
press
io
n det
ec
t
io
n
Ob
ject
d
etec
tio
n
is
a
f
u
n
d
am
en
tal
task
in
th
e
f
ield
o
f
d
ee
p
lear
n
in
g
t
h
at
in
v
o
l
v
es
id
en
t
if
y
in
g
an
d
lo
ca
lizin
g
o
b
jects
with
in
a
d
ig
ital
im
ag
e.
I
n
th
is
s
tu
d
y
,
th
e
YOL
Ov
8
m
o
d
el
is
em
p
lo
y
e
d
d
u
e
to
its
en
h
an
ce
d
ac
cu
r
ac
y
,
f
aster
d
etec
tio
n
s
p
e
ed
,
an
d
im
p
r
o
v
e
d
a
b
ilit
y
to
r
e
co
g
n
ize
s
m
all
a
n
d
c
o
m
p
lex
o
b
jects
co
m
p
ar
e
d
to
its
p
r
ed
ec
ess
o
r
s
[
1
9
]
,
[
2
0
]
.
Y
OL
Ov
8
in
tr
o
d
u
ce
s
ar
ch
itectu
r
al
im
p
r
o
v
em
en
t,
in
clu
d
in
g
an
en
h
an
ce
d
b
ac
k
b
o
n
e
n
etwo
r
k
,
a
n
an
ch
o
r
-
f
r
ee
d
et
ec
tio
n
h
ea
d
,
a
n
d
m
u
lti
-
s
ca
le
f
ea
tu
r
e
f
u
s
io
n
,
wh
ich
b
o
o
s
t
class
if
icatio
n
an
d
lo
ca
lizatio
n
p
er
f
o
r
m
an
ce
.
T
h
is
m
ak
es
it
p
ar
ticu
lar
ly
s
u
itab
le
f
o
r
d
etec
tin
g
s
u
b
tle
an
d
d
y
n
a
m
ic
f
ac
ial
ex
p
r
ess
io
n
s
in
ch
ild
r
e
n
with
a
u
tis
m
[
2
1
]
.
T
h
e
f
ac
ial
em
o
tio
n
d
etec
tio
n
p
r
o
ce
s
s
co
n
s
is
ts
o
f
th
r
ee
s
tag
es:
f
ac
ial
ex
p
r
ess
io
n
d
ataset
p
r
ep
ar
atio
n
,
tr
ain
in
g
o
f
th
e
d
ee
p
lear
n
in
g
m
o
d
el
th
r
o
u
g
h
p
h
ases
o
f
tr
ai
n
in
g
,
v
alid
atio
n
,
an
d
test
in
g
,
an
d
d
ep
l
o
y
m
en
t
f
o
r
r
ea
l
-
tim
e
in
f
er
e
n
ce
.
T
h
e
tr
ain
i
n
g
p
h
ase
u
tili
ze
s
a
p
r
e
-
tr
ai
n
e
d
YOL
Ov
8
m
o
d
el
f
i
n
e
-
tu
n
e
d
with
a
to
tal
o
f
1
0
0
ep
o
ch
s
to
ad
ap
t
to
d
o
m
ai
n
-
s
p
ec
if
ic
f
ac
ial
f
ea
tu
r
es.
T
h
e
tr
ain
ed
m
o
d
el
is
ca
p
ab
le
o
f
d
etec
tin
g
an
d
class
if
y
in
g
v
ar
io
u
s
em
o
tio
n
al
ex
p
r
ess
io
n
s
ex
h
ib
ited
b
y
a
u
tis
tic
ch
ild
r
en
in
r
ea
l
tim
e,
e
n
ab
lin
g
c
o
n
tin
u
o
u
s
em
o
tio
n
m
o
n
ito
r
in
g
in
in
clu
s
iv
e
class
r
o
o
m
en
v
ir
o
n
m
en
t.
2
.
3
.
YO
L
O
v
8
-
ba
s
ed
bo
dy
key
po
int
det
ec
t
io
n
Key
p
o
in
t
-
b
ased
o
b
ject
d
etec
ti
o
n
is
an
ad
v
an
ce
d
c
o
m
p
u
ter
v
is
io
n
ap
p
r
o
ac
h
th
at
id
e
n
tifie
s
o
b
jects
b
y
lo
ca
tin
g
cr
itical
b
o
d
y
lan
d
m
a
r
k
s
,
s
u
ch
as
jo
in
ts
an
d
s
k
eleta
l
p
o
in
ts
.
T
h
is
m
eth
o
d
is
p
ar
ticu
lar
ly
ef
f
ec
tiv
e
f
o
r
an
aly
zin
g
d
y
n
am
ic
b
o
d
y
m
o
v
em
en
ts
,
in
clu
d
in
g
ar
m
an
d
u
p
p
er
-
b
o
d
y
g
estu
r
es
in
ch
ild
r
e
n
ca
p
tu
r
ed
th
r
o
u
g
h
v
id
eo
s
eq
u
e
n
ce
s
[
2
2
]
.
I
n
th
is
s
tu
d
y
,
th
e
YOL
Ov
8
k
ey
p
o
i
n
t
d
etec
tio
n
m
o
d
el
is
em
p
lo
y
ed
to
ex
tr
ac
t
b
o
d
y
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
1
6
,
No
.
4
,
Au
g
u
s
t
20
2
6
:
1
8
9
9
-
1
912
1902
lan
d
m
ar
k
s
f
r
o
m
web
ca
m
r
ec
o
r
d
in
g
s
o
f
ch
ild
r
e
n
with
ASD.
T
h
e
m
o
d
el
is
in
itialized
with
C
OC
O
-
2
0
1
7
p
r
etr
ain
ed
weig
h
ts
,
wh
ich
p
r
o
v
id
e
an
n
o
tatio
n
s
f
o
r
1
7
s
tan
d
a
r
d
h
u
m
an
k
e
y
p
o
i
n
ts
[
2
3
]
.
T
h
e
tr
ain
in
g
p
r
o
ce
s
s
is
co
n
d
u
cte
d
o
v
er
1
.
0
0
0
e
p
o
ch
s
t
o
en
s
u
r
e
s
tab
le
co
n
v
er
g
e
n
ce
a
n
d
r
o
b
u
s
t
f
ea
t
u
r
e
e
x
tr
ac
tio
n
.
Y
OL
Ov
8
is
s
elec
ted
d
u
e
to
its
h
ig
h
in
f
er
en
ce
s
p
e
ed
an
d
s
u
p
er
io
r
ac
cu
r
ac
y
in
b
o
th
o
b
ject
d
etec
tio
n
a
n
d
p
o
s
e
esti
m
atio
n
ta
s
k
s
,
m
ak
in
g
it
s
u
itab
le
f
o
r
r
ea
l
-
tim
e
ap
p
licatio
n
s
in
d
y
n
am
ic
clas
s
r
o
o
m
en
v
ir
o
n
m
en
ts
.
T
h
e
ex
tr
ac
ted
k
ey
p
o
in
ts
ar
e
u
s
ed
as sp
atial
f
ea
tu
r
e
r
ep
r
ese
n
tatio
n
s
f
o
r
s
u
b
s
eq
u
en
t e
m
o
ti
o
n
class
if
icatio
n
.
T
h
ese
f
ea
tu
r
es e
n
ab
le
th
e
s
y
s
tem
to
an
aly
ze
b
o
d
y
m
o
v
em
en
t
p
atter
n
s
ass
o
ciate
d
with
d
if
f
er
en
t
em
o
tio
n
al
s
tates.
Fu
r
th
er
m
o
r
e,
th
e
r
ea
l
-
tim
e
k
ey
p
o
i
n
t
tr
ac
k
in
g
m
ec
h
an
is
m
s
u
p
p
o
r
ts
p
o
ten
tial
ap
p
lic
atio
n
s
in
b
eh
av
io
r
al
m
o
n
ito
r
in
g
,
em
o
tio
n
-
awa
r
e
ass
is
tiv
e
s
y
s
tem
s
,
an
d
ea
r
ly
in
ter
v
en
tio
n
s
tr
ateg
ies f
o
r
ch
ild
r
en
with
ASD.
2
.
4
.
L
ST
M
-
ba
s
ed
m
o
v
em
ent
pa
t
t
er
n det
ec
t
io
n
T
h
e
L
STM
n
etwo
r
k
is
im
p
le
m
en
ted
in
th
is
s
tu
d
y
as
th
e
te
m
p
o
r
al
class
if
icatio
n
m
o
d
el
t
o
r
ec
o
g
n
ize
em
o
tio
n
al
s
tates
b
ased
o
n
s
eq
u
en
tial
m
o
v
em
en
t
d
ata.
L
ST
M
is
p
ar
ticu
lar
ly
s
u
itab
le
f
o
r
tim
e
-
s
er
ies
an
aly
s
is
d
u
e
to
its
g
ated
m
em
o
r
y
a
r
c
h
itectu
r
e,
wh
ich
en
a
b
les
th
e
n
etwo
r
k
to
r
etai
n
lo
n
g
-
ter
m
d
ep
en
d
e
n
cies
an
d
ca
p
tu
r
e
tem
p
o
r
al
p
atter
n
s
ef
f
ec
tiv
ely
[
2
4
]
,
[
2
5
]
.
I
n
th
is
s
tu
d
y
,
th
e
L
STM
m
o
d
el
is
em
p
lo
y
ed
to
class
if
y
em
o
tio
n
al
s
tates
b
ased
o
n
s
eq
u
en
ce
s
o
f
b
o
d
y
k
ey
p
o
in
ts
ex
tr
ac
ted
f
r
o
m
v
id
eo
f
r
am
es.
T
h
e
p
r
o
p
o
s
ed
ar
ch
itectu
r
e
co
n
s
is
ts
o
f
m
u
ltip
le
h
id
d
en
lay
e
r
s
f
o
llo
wed
b
y
a
So
f
tMa
x
o
u
tp
u
t
lay
er
to
class
if
y
em
o
tio
n
al
ca
teg
o
r
ies,
in
clu
d
i
n
g
h
a
p
p
in
e
s
s
,
s
ad
n
ess
,
n
eu
tr
ality
,
b
o
r
e
d
o
m
,
an
d
tan
tr
u
m
.
T
h
e
m
o
d
e
l
is
tr
ain
ed
u
s
in
g
b
ac
k
p
r
o
p
ag
atio
n
th
r
o
u
g
h
tim
e
with
ca
teg
o
r
ical
cr
o
s
s
-
en
tr
o
p
y
lo
s
s
an
d
t
h
e
Ad
am
o
p
t
im
izer
to
im
p
r
o
v
e
lear
n
in
g
ef
f
icien
c
y
an
d
g
e
n
e
r
aliza
tio
n
ca
p
ab
ilit
y
.
Mo
d
el
p
er
f
o
r
m
an
ce
is
ev
alu
ated
u
s
in
g
a
s
ep
ar
ate
test
d
ataset
an
d
m
ea
s
u
r
ed
th
r
o
u
g
h
s
tan
d
ar
d
ev
al
u
atio
n
m
etr
ics,
in
clu
d
in
g
ac
cu
r
ac
y
,
p
r
e
cisi
o
n
,
r
ec
all,
an
d
F1
-
s
co
r
e.
T
h
e
d
etailed
p
er
f
o
r
m
an
ce
ev
alu
atio
n
m
etr
ics
o
f
th
e
m
o
d
el,
in
clu
d
in
g
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
s
co
r
e
ac
r
o
s
s
th
e
test
ed
class
es,
ar
e
p
r
esen
ted
in
T
ab
le
1
.
Pre
v
io
u
s
s
tu
d
ies
h
a
v
e
d
em
o
n
s
tr
ated
th
e
ef
f
ec
tiv
e
n
ess
o
f
L
STM
in
em
o
tio
n
r
ec
o
g
n
itio
n
task
s
.
Hu
an
g
et
a
l.
[
2
6
]
a
p
p
lied
a
la
y
er
ed
L
STM
ar
ch
itectu
r
e,
f
o
r
s
eq
u
en
tial
d
ata
m
o
d
elin
g
a
n
d
ac
h
iev
ed
im
p
r
o
v
ed
class
if
icatio
n
p
er
f
o
r
m
an
ce
t
h
r
o
u
g
h
p
ar
am
ete
r
o
p
tim
izatio
n
.
Ok
a
et
a
l.
[
2
7
]
p
r
o
p
o
s
ed
an
atten
tio
n
-
b
ased
L
STM
m
o
d
el
f
o
r
E
E
G
-
b
ased
em
o
tio
n
r
ec
o
g
n
itio
n
,
h
ig
h
lig
h
tin
g
th
e
a
b
ilit
y
o
f
tem
p
o
r
al
n
e
two
r
k
s
to
im
p
r
o
v
e
em
o
tio
n
al
s
tate
esti
m
atio
n
.
I
n
ad
d
itio
n
,
Nan
d
in
i
et
a
l.
[
2
8
]
e
m
p
h
asize
th
at
L
STM
’
s
g
ated
m
em
o
r
y
s
tr
u
ct
u
r
e
is
ef
f
ec
tiv
e
in
r
etain
i
n
g
lo
n
g
-
r
an
g
e
d
ep
e
n
d
en
cies
f
o
r
em
o
tio
n
p
r
ed
ictio
n
.
I
n
th
e
co
n
te
x
t
o
f
t
h
is
s
tu
d
y
,
th
e
u
s
e
o
f
L
STM
en
h
an
ce
s
th
e
s
y
s
tem
’
s
ab
ilit
y
to
lear
n
tem
p
o
r
al
p
atter
n
s
in
b
o
d
y
m
o
v
em
en
t
d
ata,
th
er
eb
y
im
p
r
o
v
in
g
th
e
r
o
b
u
s
tn
ess
o
f
em
o
tio
n
d
ete
ctio
n
in
ch
ild
r
e
n
with
ASD.
T
ab
le
1
.
Per
f
o
r
m
an
ce
m
etr
ics
M
e
t
r
i
c
C
l
a
s
s
0
C
l
a
s
s
1
W
e
i
g
h
t
e
d
a
v
e
r
a
g
e
P
r
e
c
i
s
i
o
n
0
.
9
9
0
.
5
5
0
.
9
7
R
e
c
a
l
l
0
.
9
8
0
.
6
7
0
.
9
6
F1
-
sc
o
r
e
0
.
9
8
0
.
6
0
0
.
9
6
S
u
p
p
o
r
t
2
0
6
9
2
1
5
2
.
5
.
Sy
s
t
e
m
im
plem
ent
a
t
io
n
a
nd
t
esting
T
h
e
p
r
o
p
o
s
ed
em
o
tio
n
d
etec
t
io
n
s
y
s
tem
was
im
p
lem
en
ted
u
s
in
g
a
R
asp
b
er
r
y
Pi
in
teg
r
a
ted
with
a
web
ca
m
f
o
r
r
ea
l
-
tim
e
d
ata
a
cq
u
is
itio
n
in
class
r
o
o
m
en
v
ir
o
n
m
en
ts
.
T
h
e
s
y
s
tem
co
m
b
in
es
YOL
Ov
8
-
b
ased
f
ac
ial
ex
p
r
ess
io
n
d
etec
tio
n
an
d
b
o
d
y
k
e
y
p
o
in
t
d
etec
tio
n
to
ca
p
tu
r
e
s
p
atial
f
ea
tu
r
es
f
r
o
m
ch
ild
r
en
with
ASD.
T
h
e
ex
tr
ac
ted
b
o
d
y
k
ey
p
o
in
t
s
ar
e
s
u
b
s
eq
u
en
tly
p
r
o
ce
s
s
ed
u
s
in
g
a
L
STM
n
etwo
r
k
t
o
an
aly
ze
tem
p
o
r
al
m
o
v
em
en
t
p
atter
n
s
an
d
class
if
y
em
o
tio
n
al
s
tates,
in
clu
d
in
g
h
ap
p
in
ess
,
s
ad
n
ess
,
n
eu
tr
al
ity
,
b
o
r
e
d
o
m
,
an
d
tan
tr
u
m
.
T
h
e
o
v
e
r
all
im
p
lem
e
n
tatio
n
ar
ch
itectu
r
e
o
f
t
h
e
p
r
o
p
o
s
ed
s
y
s
tem
is
illu
s
tr
ated
in
Fig
u
r
e
3
.
Fig
u
r
e
3
ill
u
s
t
r
a
tes
t
h
e
o
v
e
r
a
ll
im
p
l
em
en
tat
io
n
ar
ch
ite
ct
u
r
e
o
f
t
h
e
p
r
o
p
o
s
e
d
m
u
l
tim
o
d
a
l
em
o
ti
o
n
r
e
co
g
n
it
io
n
s
y
s
te
m
f
o
r
c
h
il
d
r
e
n
wit
h
ASD
in
i
n
cl
u
s
i
v
e
e
d
u
c
a
tio
n
a
l
e
n
v
ir
o
n
m
e
n
ts
.
T
h
e
i
m
p
l
em
e
n
t
ati
o
n
p
r
o
c
ess
co
n
s
is
ts
o
f
s
e
v
e
r
a
l
s
ta
g
es
,
i
n
c
l
u
d
i
n
g
r
e
al
-
ti
m
e
v
i
d
e
o
c
a
p
t
u
r
e,
p
r
e
p
r
o
c
ess
i
n
g
,
f
a
cia
l
e
x
p
r
ess
i
o
n
d
ete
cti
o
n
,
b
o
d
y
k
e
y
p
o
in
t
e
x
t
r
a
cti
o
n
,
te
m
p
o
r
al
s
eq
u
e
n
ce
p
r
o
ce
s
s
i
n
g
,
a
n
d
e
m
o
tio
n
class
if
ica
ti
o
n
.
Y
OL
O
v
8
i
s
u
t
iliz
e
d
f
o
r
f
a
cia
l
ex
p
r
ess
i
o
n
a
n
d
b
o
d
y
k
ey
p
o
i
n
t
d
et
ec
ti
o
n
,
w
h
il
e
t
h
e
L
ST
M
n
e
tw
o
r
k
is
em
p
l
o
y
e
d
t
o
an
al
y
ze
s
e
q
u
e
n
tia
l
m
o
v
em
en
t
p
att
er
n
s
.
T
h
e
clas
s
if
i
ed
em
o
t
io
n
al
s
ta
tes
a
r
e
tr
an
s
m
it
te
d
t
h
r
o
u
g
h
an
I
o
T
-
b
a
s
ed
co
m
m
u
n
i
ca
t
io
n
f
r
am
ew
o
r
k
a
n
d
d
is
p
la
y
ed
o
n
a
w
eb
-
b
as
ed
m
o
n
it
o
r
i
n
g
d
as
h
b
o
a
r
d
.
I
n
ad
d
i
ti
o
n
,
t
h
e
s
y
s
t
em
is
i
n
t
e
g
r
ate
d
wit
h
a
W
h
atsA
p
p
n
o
t
if
i
ca
ti
o
n
s
e
r
v
ic
e
t
o
p
r
o
v
i
d
e
r
e
al
-
ti
m
e
e
m
o
ti
o
n
a
l st
at
u
s
u
p
d
ates
t
o
e
d
u
ca
to
r
s
a
n
d
p
ar
en
ts
.
Sy
s
tem
test
in
g
was
co
n
d
u
cted
u
n
d
e
r
r
ea
l
-
wo
r
ld
class
r
o
o
m
co
n
d
itio
n
s
to
e
v
alu
ate
th
e
p
er
f
o
r
m
a
n
ce
an
d
p
r
ac
tical
ap
p
licab
ilit
y
o
f
th
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
.
T
h
e
p
er
f
o
r
m
an
ce
e
v
alu
atio
n
f
o
cu
s
ed
o
n
b
o
th
f
ac
ial
-
ex
p
r
ess
io
n
-
b
ased
an
d
m
o
v
em
en
t
-
b
ased
em
o
tio
n
r
ec
o
g
n
itio
n
m
o
d
els
u
s
in
g
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
a
n
d
F1
-
s
co
r
e
m
etr
ics.
T
h
e
ex
p
er
i
m
en
tal
r
esu
lts
in
d
icate
th
at
th
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
ac
h
iev
ed
s
tab
le
an
d
r
eliab
le
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
R
ea
l
-
time
fa
cia
l a
n
d
b
o
d
y
p
o
s
e
emo
tio
n
r
ec
o
g
n
itio
n
f
o
r
ch
ild
r
en
…
(
S
iti N
u
r
o
h
ma
h
w
a
ti
)
1903
p
er
f
o
r
m
an
ce
u
n
d
er
d
y
n
am
ic
class
r
o
o
m
co
n
d
itio
n
s
.
T
h
e
i
n
teg
r
atio
n
o
f
m
u
ltimo
d
al
em
o
tio
n
r
ec
o
g
n
itio
n
an
d
I
o
T
-
b
ased
m
o
n
ito
r
in
g
p
r
o
v
i
d
es
a
p
r
ac
tical
s
o
lu
tio
n
f
o
r
co
n
tin
u
o
u
s
em
o
tio
n
al
o
b
s
e
r
v
atio
n
a
n
d
ea
r
l
y
in
ter
v
en
tio
n
s
u
p
p
o
r
t in
in
clu
s
i
v
e
ed
u
ca
tio
n
al
en
v
ir
o
n
m
e
n
ts
.
Fig
u
r
e
3
.
Ar
c
h
itectu
r
e
o
f
th
e
p
r
o
p
o
s
ed
m
u
ltimo
d
al
em
o
tio
n
r
ec
o
g
n
itio
n
a
n
d
m
o
n
ito
r
in
g
s
y
s
tem
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
3
.
1
.
F
a
cia
l e
x
press
io
n det
ec
t
io
n per
f
o
rm
a
nce
T
h
e
f
ac
ial
e
x
p
r
ess
io
n
r
ec
o
g
n
itio
n
m
o
d
el
b
ased
o
n
YOL
O
v
8
was
tr
ain
e
d
th
r
o
u
g
h
s
ev
e
r
al
s
tag
es,
in
clu
d
in
g
d
ataset
ac
q
u
is
itio
n
,
p
r
ep
r
o
ce
s
s
in
g
,
tr
ain
i
n
g
,
v
alid
atio
n
,
a
n
d
test
in
g
.
T
h
e
t
r
ain
in
g
p
r
o
ce
s
s
was
co
n
d
u
cte
d
iter
ativ
ely
f
o
r
1
0
0
ep
o
ch
s
to
o
p
tim
ize
t
h
e
m
o
d
el’
s
lear
n
in
g
ca
p
ab
ilit
y
an
d
im
p
r
o
v
e
its
g
en
er
aliza
tio
n
p
e
r
f
o
r
m
an
ce
i
n
r
ec
o
g
n
izin
g
em
o
tio
n
al
ex
p
r
ess
io
n
s
o
f
ch
ild
r
en
with
A
SD.
T
h
e
ev
alu
ated
em
o
tio
n
ca
teg
o
r
ies
co
n
s
is
ted
o
f
n
e
u
tr
al,
h
a
p
p
y
,
s
ad
,
a
n
d
an
g
r
y
e
x
p
r
ess
io
n
s
,
wh
ich
we
r
e
co
n
s
is
ten
tly
ap
p
lied
th
r
o
u
g
h
o
u
t
th
e
e
x
p
er
im
e
n
tal
an
aly
s
is
.
T
o
ev
alu
ate
th
e
class
if
icatio
n
p
er
f
o
r
m
a
n
ce
ac
r
o
s
s
d
if
f
er
en
t
em
o
tio
n
al
ca
teg
o
r
ies,
th
e
p
r
ec
is
io
n
,
r
ec
al
l,
an
d
F1
-
s
co
r
e
m
etr
ics we
r
e
c
alcu
lated
,
as su
m
m
ar
ized
in
T
ab
le
1
.
Fig
u
r
e
4
illu
s
tr
ates
th
e
tr
ain
in
g
an
d
v
alid
atio
n
p
er
f
o
r
m
a
n
ce
m
etr
ics
o
b
tain
ed
d
u
r
in
g
th
e
f
ac
ial
ex
p
r
ess
io
n
d
etec
tio
n
p
r
o
ce
s
s
.
T
h
e
Av
er
ag
e
Pre
cisi
o
n
at
I
o
U
th
r
esh
o
ld
0
.
5
(
AP5
0
)
an
d
AP5
0
–
9
5
v
al
u
es
in
cr
ea
s
ed
s
ig
n
if
ican
tly
d
u
r
in
g
th
e
ea
r
ly
tr
ain
in
g
ep
o
c
h
s
an
d
g
r
ad
u
ally
co
n
v
er
g
ed
to
war
d
s
tab
le
v
alu
es
in
th
e
later
ep
o
ch
s
.
T
h
is
tr
en
d
in
d
i
ca
tes
th
at
th
e
p
r
o
p
o
s
ed
YOL
Ov
8
-
b
ased
m
o
d
el
was
ab
le
to
ef
f
ec
tiv
ely
lear
n
r
ep
r
esen
tativ
e
f
ac
ial
f
ea
tu
r
es
wh
ile
m
ain
tain
in
g
s
tab
le
d
etec
tio
n
p
er
f
o
r
m
an
ce
ac
r
o
s
s
d
if
f
er
en
t
I
n
ter
s
ec
tio
n
o
v
er
Un
io
n
(
I
o
U)
th
r
esh
o
ld
s
r
an
g
in
g
f
r
o
m
0
.
5
t
o
0
.
9
5
.
T
h
e
p
r
ec
is
io
n
an
d
r
ec
all
m
etr
i
cs
also
d
em
o
n
s
tr
ated
co
n
s
is
te
n
t
im
p
r
o
v
em
e
n
t
th
r
o
u
g
h
o
u
t
th
e
tr
ain
in
g
p
h
ase,
with
b
o
th
m
etr
ics
ap
p
r
o
ac
h
in
g
v
alu
es
clo
s
e
to
0
.
9
0
o
n
th
e
v
alid
atio
n
d
ataset.
Af
ter
1
0
0
tr
ain
in
g
ep
o
ch
s
,
th
e
p
r
o
p
o
s
ed
m
o
d
el
a
ch
iev
ed
a
v
alid
atio
n
p
r
ec
is
io
n
o
f
9
2
%,
ac
co
m
p
an
ied
b
y
s
tab
le
r
ec
all
an
d
m
ea
n
Av
er
ag
e
Pre
cisi
o
n
(
m
AP
)
v
a
lu
es.
T
h
e
co
n
v
e
r
g
en
ce
o
f
th
ese
ev
alu
atio
n
m
etr
ics
in
d
ica
tes
th
at
th
e
m
o
d
el
ac
h
iev
ed
b
alan
ce
d
p
e
r
f
o
r
m
an
ce
b
etwe
en
p
r
ed
ictio
n
ac
cu
r
ac
y
an
d
d
etec
tio
n
ca
p
ab
ilit
y
wh
ile
m
in
im
izin
g
f
alse
-
p
o
s
itiv
e
class
if
icatio
n
s
.
Fu
r
th
er
m
o
r
e
,
th
e
s
tab
le
co
n
v
e
r
g
en
ce
p
atter
n
s
u
g
g
ests
th
at
th
e
m
o
d
el
was
ab
le
to
m
ain
tain
g
o
o
d
g
e
n
er
aliza
tio
n
p
er
f
o
r
m
an
ce
with
o
u
t sig
n
if
ica
n
t o
v
er
f
itti
n
g
d
u
r
in
g
th
e
tr
ain
i
n
g
p
r
o
ce
s
s
.
Fig
u
r
e
5
p
r
esen
ts
s
ev
er
al
ex
am
p
les
o
f
f
ac
ial
em
o
tio
n
d
ete
ctio
n
r
esu
lts
p
r
o
d
u
ce
d
b
y
th
e
p
r
o
p
o
s
ed
s
y
s
tem
.
T
h
e
m
o
d
el
s
u
cc
ess
f
u
lly
id
en
tifie
d
th
e
p
r
ed
ef
in
e
d
e
m
o
tio
n
ca
teg
o
r
ies,
in
clu
d
in
g
n
eu
tr
al,
h
a
p
p
y
,
s
ad
,
an
d
tan
tr
u
m
ex
p
r
ess
io
n
s
,
u
n
d
er
v
ar
y
in
g
f
ac
ial
o
r
ie
n
tatio
n
s
an
d
class
r
o
o
m
co
n
d
iti
o
n
s
.
T
h
ese
r
esu
lts
d
em
o
n
s
tr
ate
th
at
th
e
p
r
o
p
o
s
e
d
f
ac
ial
ex
p
r
ess
io
n
r
ec
o
g
n
itio
n
f
r
am
ewo
r
k
is
ca
p
a
b
le
o
f
p
er
f
o
r
m
in
g
r
ea
l
-
tim
e
em
o
tio
n
d
etec
tio
n
in
d
y
n
am
ic
ed
u
ca
tio
n
al
en
v
ir
o
n
m
en
ts
in
v
o
lv
in
g
c
h
ild
r
en
with
ASD.
C
o
m
p
ar
ed
with
co
n
v
en
tio
n
al
im
ag
e
-
b
ased
f
ac
ial
em
o
tio
n
r
ec
o
g
n
itio
n
a
p
p
r
o
ac
h
es,
th
e
p
r
o
p
o
s
ed
YOL
Ov
8
-
b
ased
f
r
a
m
ewo
r
k
p
r
o
v
id
es
im
p
r
o
v
ed
d
etec
tio
n
s
ta
b
ilit
y
an
d
r
ea
l
-
tim
e
p
r
o
ce
s
s
in
g
ca
p
ab
ilit
y
th
r
o
u
g
h
ef
f
icien
t
f
ea
tu
r
e
ex
tr
ac
tio
n
.
T
h
e
in
teg
r
atio
n
o
f
f
ac
ial
ex
p
r
ess
io
n
an
aly
s
is
with
in
th
e
m
u
ltimo
d
al
em
o
tio
n
r
ec
o
g
n
itio
n
s
y
s
tem
also
co
n
tr
ib
u
tes
to
m
o
r
e
r
eliab
le
em
o
tio
n
al
m
o
n
ito
r
i
n
g
in
i
n
clu
s
iv
e
class
r
o
o
m
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
1
6
,
No
.
4
,
Au
g
u
s
t
20
2
6
:
1
8
9
9
-
1
912
1904
en
v
ir
o
n
m
en
ts
.
T
h
ese
f
in
d
in
g
s
in
d
icate
th
at
th
e
p
r
o
p
o
s
ed
m
eth
o
d
h
as
s
tr
o
n
g
p
o
ten
tial
to
s
u
p
p
o
r
t
co
n
tin
u
o
u
s
em
o
tio
n
al
o
b
s
er
v
atio
n
a
n
d
ea
r
ly
b
eh
av
i
o
r
al
in
ter
v
e
n
tio
n
f
o
r
ch
ild
r
en
with
ASD.
Fig
u
r
e
4
.
T
r
ain
in
g
m
etr
ic
cu
r
v
es f
o
r
m
o
d
el
ev
alu
atio
n
o
f
f
ac
ial
ex
p
r
ess
io
n
d
etec
tio
n
Fig
u
r
e
5
.
E
m
o
tio
n
d
etec
tio
n
r
e
s
u
lts
o
f
n
eu
tr
al,
h
a
p
p
y
,
s
ad
,
an
d
tan
tr
u
m
e
x
p
r
ess
io
n
s
3
.
2
.
B
o
dy
k
ey
po
ints det
ec
t
io
n a
nd
po
s
t
ura
l a
na
ly
s
is
Fig
u
r
e
6
p
r
esen
ts
th
e
ev
al
u
atio
n
m
etr
ics
o
b
tain
ed
d
u
r
in
g
th
e
tr
ain
in
g
p
r
o
ce
s
s
o
f
th
e
YOL
Ov
8
-
b
ased
b
o
d
y
k
ey
p
o
in
t
d
etec
tio
n
m
o
d
el
u
s
in
g
C
OC
O
-
2
0
1
7
d
ataset.
T
h
e
ev
alu
atio
n
was
co
n
d
u
cte
d
o
n
th
e
v
alid
atio
n
d
ataset
to
ass
ess
th
e
m
o
d
el’
s
p
er
f
o
r
m
an
ce
i
n
d
etec
tin
g
b
o
d
y
lan
d
m
ar
k
s
an
d
esti
m
atin
g
h
u
m
an
p
o
s
tu
r
e
u
n
d
e
r
d
if
f
er
en
t
m
o
v
em
en
t
co
n
d
itio
n
s
.
Sev
er
al
p
er
f
o
r
m
a
n
ce
in
d
icato
r
s
wer
e
an
aly
ze
d
,
in
clu
d
in
g
m
AP
at
an
I
n
ter
s
ec
tio
n
o
v
er
Un
i
o
n
th
r
esh
o
ld
o
f
0
.
5
,
m
AP
ac
r
o
s
s
I
o
U
th
r
esh
o
ld
s
f
r
o
m
0
.
5
to
0
.
9
5
,
a
s
well
as
p
r
ec
is
io
n
an
d
r
ec
all
f
o
r
b
o
th
b
o
u
n
d
in
g
b
o
x
(
B
)
an
d
k
ey
p
o
in
t (
P)
p
r
ed
ic
tio
n
s
.
T
h
e
ex
p
er
im
en
tal
r
esu
lts
d
e
m
o
n
s
tr
ate
th
at
t
h
e
p
r
o
p
o
s
ed
m
o
d
el
ac
h
iev
e
d
s
tab
le
c
o
n
v
er
g
e
n
ce
th
r
o
u
g
h
o
u
t
th
e
1
,
0
0
0
tr
ain
in
g
ep
o
ch
s
.
At
ep
o
c
h
9
7
9
,
th
e
m
o
d
el
r
ea
ch
e
d
its
o
p
tim
al
p
er
f
o
r
m
an
ce
,
with
m
AP@
0
.
5
v
alu
es
o
f
0
.
9
1
9
f
o
r
b
o
u
n
d
in
g
b
o
x
d
etec
tio
n
an
d
0
.
8
1
0
f
o
r
k
e
y
p
o
in
t
p
r
ed
ictio
n
.
I
n
ad
d
itio
n
,
th
e
o
b
tain
ed
p
r
ec
is
io
n
an
d
r
ec
a
ll
v
alu
es
in
d
icate
s
tr
o
n
g
d
e
tectio
n
ca
p
ab
ilit
y
,
with
Pre
c
is
io
n
(
B
)
=
0
.
8
9
3
,
Pre
cisi
o
n
(
P)
=
0
.
8
4
8
,
R
ec
all(
B
)
=
0
.
8
2
2
,
an
d
R
ec
all(
P)
=
0
.
7
5
8
.
T
h
ese
r
esu
lts
in
d
icate
th
at
th
e
YOL
Ov
8
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
R
ea
l
-
time
fa
cia
l a
n
d
b
o
d
y
p
o
s
e
emo
tio
n
r
ec
o
g
n
itio
n
f
o
r
ch
ild
r
en
…
(
S
iti N
u
r
o
h
ma
h
w
a
ti
)
1905
ar
ch
itectu
r
e
was
ca
p
ab
le
o
f
ac
cu
r
ately
id
en
tify
in
g
b
o
d
y
l
an
d
m
ar
k
s
wh
ile
m
ain
tain
in
g
b
alan
ce
d
d
etec
tio
n
p
er
f
o
r
m
an
ce
ac
r
o
s
s
d
if
f
er
en
t
b
o
d
y
m
o
v
e
m
en
t
co
n
d
itio
n
s
.
Alth
o
u
g
h
t
h
e
m
AP@
0
.
5
–
0
.
9
5
v
alu
es
wer
e
lo
wer
th
an
th
e
c
o
r
r
esp
o
n
d
in
g
m
AP@
0
.
5
v
alu
es,
th
is
b
eh
a
v
io
r
is
ex
p
ec
ted
b
ec
au
s
e
s
tr
icter
I
o
U
th
r
esh
o
ld
s
r
eq
u
ir
e
m
o
r
e
p
r
ec
is
e
lo
ca
lizatio
n
o
f
b
o
d
y
k
ey
p
o
in
ts
.
Nev
er
t
h
eless
,
th
e
o
b
tain
e
d
r
esu
lts
co
n
f
ir
m
th
e
ef
f
ec
tiv
en
ess
an
d
r
o
b
u
s
tn
ess
o
f
th
e
p
r
o
p
o
s
ed
m
o
d
el
f
o
r
h
u
m
an
p
o
s
tu
r
e
esti
m
ati
o
n
an
d
b
o
d
y
k
e
y
p
o
i
n
t d
etec
tio
n
task
s
.
T
h
e
s
tab
le
lear
n
in
g
p
atter
n
o
b
s
er
v
ed
in
t
h
e
tr
ain
in
g
cu
r
v
es
also
in
d
icate
s
th
at
th
e
m
o
d
el
m
ain
tain
ed
g
o
o
d
g
e
n
er
aliza
tio
n
ca
p
ab
ilit
y
with
o
u
t sig
n
if
ican
t
o
v
er
f
itti
n
g
d
u
r
in
g
th
e
tr
ai
n
in
g
p
r
o
ce
s
s
.
Fig
u
r
e
6
.
T
r
ain
in
g
m
etr
ic
cu
r
v
es f
o
r
m
o
d
el
ev
alu
atio
n
o
f
b
o
d
y
k
ey
p
o
in
t d
etec
tio
n
Fig
u
r
e
7
s
h
o
ws
th
e
im
p
lem
en
tatio
n
r
esu
lts
o
f
th
e
p
r
o
p
o
s
ed
YOL
Ov
8
-
b
ased
b
o
d
y
k
ey
p
o
in
t
d
etec
tio
n
an
d
p
o
s
tu
r
al
a
n
aly
s
is
s
y
s
tem
f
o
r
em
o
tio
n
r
ec
o
g
n
itio
n
in
ch
il
d
r
en
with
ASD.
T
h
e
m
o
d
el
s
u
cc
ess
f
u
lly
d
etec
ted
h
u
m
an
b
o
d
y
p
o
s
tu
r
e
b
y
id
e
n
tify
in
g
m
u
ltip
le
b
o
d
y
k
e
y
p
o
i
n
ts
,
in
clu
d
in
g
th
e
h
ea
d
,
s
h
o
u
ld
e
r
s
,
ar
m
s
,
h
ip
s
,
k
n
ee
s
,
an
d
leg
s
,
wh
ich
ar
e
r
ep
r
esen
ted
b
y
co
n
n
ec
ted
s
k
eleta
l
lin
es.
I
n
ad
d
itio
n
,
th
e
s
y
s
tem
was
ab
le
to
class
if
y
s
ev
er
al
em
o
tio
n
al
an
d
b
e
h
av
io
r
al
s
tates,
s
u
ch
as
h
ap
p
y
,
tan
t
r
u
m
,
a
n
d
b
o
r
e
d
,
with
h
ig
h
c
o
n
f
id
en
ce
s
co
r
es.
T
h
e
r
esu
lts
d
em
o
n
s
tr
ate
th
at
th
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
ca
n
a
cc
u
r
ately
ca
p
tu
r
e
b
o
d
y
m
o
v
em
en
t
p
atter
n
s
an
d
p
o
s
tu
r
al
b
eh
a
v
io
r
s
u
n
d
er
d
if
f
er
en
t
en
v
ir
o
n
m
e
n
tal
co
n
d
itio
n
s
an
d
b
o
d
y
o
r
ien
tatio
n
s
.
T
h
ese
f
in
d
in
g
s
in
d
icate
th
at
th
e
m
o
d
el
is
ca
p
ab
le
o
f
p
er
f
o
r
m
in
g
r
o
b
u
s
t
r
ea
l
-
tim
e
b
o
d
y
m
o
v
em
e
n
t
an
aly
s
is
f
o
r
b
eh
av
io
r
al
m
o
n
ito
r
in
g
ap
p
licatio
n
s
.
Fig
u
r
e
7
.
Key
p
o
in
t
d
etec
tio
n
r
esu
lts
o
f
s
ev
er
al
em
o
tio
n
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
1
6
,
No
.
4
,
Au
g
u
s
t
20
2
6
:
1
8
9
9
-
1
912
1906
T
h
e
ex
tr
ac
te
d
b
o
d
y
k
e
y
p
o
i
n
ts
p
lay
an
im
p
o
r
tan
t
r
o
le
i
n
s
u
p
p
o
r
tin
g
m
u
ltimo
d
al
em
o
tio
n
r
e
co
g
n
itio
n
,
p
ar
ticu
lar
ly
f
o
r
ch
ild
r
en
with
ASD,
wh
o
s
e
em
o
tio
n
al
s
tates
ar
e
o
f
ten
r
ef
lecte
d
th
r
o
u
g
h
r
e
p
etitiv
e
m
o
v
em
en
ts
,
p
o
s
tu
r
e
ch
an
g
es,
an
d
n
o
n
v
er
b
al
b
o
d
y
g
estu
r
es.
C
o
m
p
ar
ed
with
co
n
v
en
tio
n
al
f
ac
ial
-
ex
p
r
ess
io
n
-
o
n
l
y
ap
p
r
o
ac
h
es,
th
e
in
teg
r
atio
n
o
f
b
o
d
y
p
o
s
tu
r
e
an
al
y
s
is
p
r
o
v
i
d
e
s
ad
d
itio
n
al
c
o
n
tex
tu
al
in
f
o
r
m
atio
n
th
at
im
p
r
o
v
es
th
e
r
eliab
ilit
y
o
f
em
o
tio
n
r
ec
o
g
n
itio
n
in
d
y
n
am
ic
class
r
o
o
m
en
v
ir
o
n
m
e
n
ts
.
T
h
er
e
f
o
r
e,
t
h
e
p
r
o
p
o
s
ed
b
o
d
y
k
ey
p
o
i
n
t
d
etec
tio
n
f
r
am
ewo
r
k
co
n
tr
ib
u
tes
to
m
o
r
e
co
m
p
r
eh
en
s
iv
e
b
eh
av
io
r
al
m
o
n
ito
r
in
g
an
d
s
u
p
p
o
r
ts
r
ea
l
-
tim
e
em
o
tio
n
al
ass
ess
m
en
t in
in
clu
s
iv
e
ed
u
ca
tio
n
al
s
ettin
g
s
.
3
.
3
.
M
o
t
io
n
pa
t
t
er
n r
ec
o
g
nit
io
n a
nd
t
em
po
ra
l e
m
o
t
io
n a
na
ly
s
is
Fig
u
r
e
8
illu
s
tr
ates
o
n
e
o
f
th
e
ch
ar
ac
ter
is
tic
b
eh
av
io
r
al
m
o
v
em
en
t
p
atter
n
s
ex
h
ib
ited
b
y
ch
ild
r
en
with
ASD
d
u
r
in
g
a
ta
n
tr
u
m
ep
is
o
d
e,
s
p
ec
if
ically
r
e
p
etitiv
e
s
elf
-
h
itti
n
g
b
eh
a
v
io
r
u
s
in
g
b
o
th
h
an
d
s
.
Su
c
h
r
ep
etitiv
e
b
o
d
y
m
o
v
em
en
ts
r
e
p
r
esen
t
im
p
o
r
ta
n
t
n
o
n
v
er
b
al
i
n
d
icato
r
s
th
at
ca
n
p
r
o
v
id
e
ad
d
itio
n
al
co
n
tex
tu
al
in
f
o
r
m
atio
n
f
o
r
em
o
tio
n
r
ec
o
g
n
itio
n
s
y
s
tem
s
.
I
n
th
is
s
tu
d
y
,
tem
p
o
r
al
m
o
v
em
en
t
a
n
aly
s
is
was
p
er
f
o
r
m
e
d
t
o
ca
p
tu
r
e
d
y
n
am
ic
b
eh
a
v
io
r
al
p
atter
n
s
th
at
n
o
t
b
e
r
ep
r
esen
ted
th
r
o
u
g
h
f
ac
ial
ex
p
r
ess
io
n
s
.
T
o
class
if
y
s
eq
u
en
tial
m
o
v
em
en
t
p
atter
n
s
in
to
d
is
tin
ct
em
o
tio
n
al
ca
teg
o
r
ies,
a
L
STM
n
eu
r
al
n
etwo
r
k
was
em
p
lo
y
ed
d
u
e
to
its
ca
p
ab
ilit
y
to
ef
f
ec
tiv
ely
m
o
d
e
l
tem
p
o
r
al
d
ep
en
d
e
n
cies
in
s
e
q
u
en
tial
d
ata.
T
h
e
class
if
icati
o
n
p
r
o
ce
s
s
u
tili
ze
d
b
o
d
y
k
ey
p
o
in
t
in
f
o
r
m
atio
n
e
x
tr
ac
ted
f
r
o
m
1
7
a
n
ato
m
ical
la
n
d
m
ar
k
s
b
ased
o
n
th
e
C
OC
O
an
n
o
tatio
n
f
o
r
m
at.
T
h
ese
k
ey
p
o
in
ts
wer
e
co
llecte
d
co
n
tin
u
o
u
s
ly
ac
r
o
s
s
2
0
co
n
s
ec
u
tiv
e
f
r
a
m
es,
wh
er
e
ea
ch
s
e
q
u
en
ce
was
u
s
ed
as
a
s
in
g
le
in
p
u
t sam
p
le
f
o
r
th
e
L
STM
m
o
d
el
to
r
e
p
r
esen
t te
m
p
o
r
al
m
o
v
e
m
en
t b
e
h
av
io
r
.
Fig
u
r
e
8
.
Seq
u
en
tial b
eh
a
v
io
r
a
l m
o
v
em
en
t
p
atter
n
s
ac
r
o
s
s
tem
p
o
r
al
f
r
am
es
d
u
r
i
n
g
tan
tr
u
m
ep
is
o
d
es
T
h
e
p
r
o
p
o
s
ed
L
STM
-
b
ased
f
r
am
ewo
r
k
ca
teg
o
r
ize
d
b
o
d
y
m
o
v
em
en
t
p
atter
n
s
in
t
o
th
r
ee
em
o
tio
n
al
s
tates,
n
am
ely
tan
tr
u
m
,
h
a
p
p
y
,
an
d
n
eu
tr
al/b
o
r
ed
.
T
h
e
u
s
e
o
f
tem
p
o
r
al
s
eq
u
en
ce
m
o
d
elin
g
en
ab
led
th
e
s
y
s
tem
to
r
ec
o
g
n
ize
d
y
n
am
ic
b
o
d
y
ex
p
r
ess
io
n
s
m
o
r
e
ef
f
ec
tiv
ely
th
a
n
s
tatic
im
ag
e
-
b
ased
ap
p
r
o
ac
h
es.
T
h
is
ca
p
ab
ilit
y
is
p
ar
ticu
lar
ly
im
p
o
r
ta
n
t in
ASD
b
eh
av
io
r
al
a
n
aly
s
is
,
wh
er
e
em
o
tio
n
al
s
tates a
r
e
f
r
eq
u
e
n
tly
ex
p
r
ess
ed
th
r
o
u
g
h
r
ep
etitiv
e
m
o
v
e
m
en
ts
,
p
o
s
tu
r
e
ch
an
g
es,
an
d
n
o
n
v
er
b
al
g
estu
r
es r
ath
er
th
an
f
ac
ial
ex
p
r
ess
io
n
s
alo
n
e.
Fig
u
r
e
9
p
r
esen
ts
th
e
v
alid
atio
n
ac
cu
r
ac
y
an
d
v
alid
atio
n
l
o
s
s
cu
r
v
es
o
b
tain
ed
d
u
r
in
g
th
e
tr
ain
in
g
p
r
o
ce
s
s
o
f
th
e
L
STM
m
o
d
el.
T
h
e
v
alid
atio
n
ac
cu
r
ac
y
i
n
cr
e
ased
s
ig
n
if
ican
tly
d
u
r
in
g
th
e
i
n
itial
tr
ain
in
g
p
h
ase,
r
is
in
g
f
r
o
m
ap
p
r
o
x
im
atel
y
4
7
%
to
7
3
%
with
in
th
e
f
ir
s
t
f
iv
e
ep
o
ch
s
.
Fo
llo
win
g
th
is
s
tag
e,
th
e
m
o
d
el
d
em
o
n
s
tr
ated
s
tab
le
lear
n
in
g
b
eh
av
io
r
b
ef
o
r
e
co
n
tin
u
in
g
to
im
p
r
o
v
e
a
n
d
r
ea
c
h
in
g
a
p
ea
k
v
alid
atio
n
ac
cu
r
ac
y
o
f
9
5
%
at
ep
o
c
h
1
7
.
T
h
e
ac
c
u
r
ac
y
v
alu
es
r
em
ai
n
ed
r
elativ
ely
s
tab
le
d
u
r
in
g
th
e
r
em
ain
i
n
g
tr
ain
in
g
e
p
o
ch
s
,
in
d
icatin
g
co
n
s
is
ten
t m
o
d
el
p
e
r
f
o
r
m
a
n
ce
an
d
ef
f
ec
tiv
e
tem
p
o
r
al
f
ea
tu
r
e
lea
r
n
in
g
.
Simu
ltan
eo
u
s
ly
,
th
e
v
alid
atio
n
lo
s
s
ex
h
ib
ited
a
g
r
ad
u
al
d
e
cr
ea
s
e
th
r
o
u
g
h
o
u
t
th
e
tr
ain
in
g
p
r
o
ce
s
s
,
r
ef
lectin
g
ef
f
ec
tiv
e
m
in
im
izatio
n
o
f
p
r
ed
ictio
n
e
r
r
o
r
s
d
u
r
in
g
m
o
d
el
o
p
tim
izatio
n
.
T
h
e
in
v
er
s
e
r
elatio
n
s
h
ip
b
etwe
en
in
cr
ea
s
in
g
v
alid
atio
n
ac
cu
r
ac
y
an
d
d
ec
r
ea
s
in
g
v
alid
atio
n
lo
s
s
s
u
g
g
ests
th
at
t
h
e
p
r
o
p
o
s
ed
L
STM
m
o
d
el
s
u
cc
ess
f
u
lly
ca
p
tu
r
e
d
t
em
p
o
r
al
m
o
v
em
en
t
p
atter
n
s
with
o
u
t
ex
h
ib
itin
g
s
ig
n
if
ican
t
o
v
er
f
itti
n
g
b
e
h
av
io
r
.
T
h
ese
f
in
d
in
g
s
in
d
icate
th
at
t
h
e
tem
p
o
r
al
m
o
d
elin
g
ap
p
r
o
a
ch
co
n
tr
i
b
u
ted
s
u
b
s
tan
tially
to
im
p
r
o
v
in
g
em
o
tio
n
class
if
icatio
n
p
er
f
o
r
m
an
ce
b
ased
o
n
b
eh
av
i
o
r
al
m
o
v
em
en
t se
q
u
en
ce
s
.
Fig
u
r
e
1
0
p
r
esen
ts
th
e
c
o
n
f
u
s
io
n
m
atr
ix
o
f
th
e
p
r
o
p
o
s
ed
tem
p
o
r
al
em
o
tio
n
class
if
icatio
n
m
o
d
el.
T
h
e
ex
p
er
im
en
tal
r
esu
lts
d
em
o
n
s
tr
ate
ex
ce
llen
t
class
if
icatio
n
p
er
f
o
r
m
an
ce
f
o
r
th
e
n
eu
tr
al/
b
o
r
ed
a
n
d
ta
n
tr
u
m
ca
teg
o
r
ies,
wh
er
e
all
n
eu
t
r
al/b
o
r
ed
s
am
p
les
an
d
f
iv
e
o
u
t
o
f
s
ix
tan
tr
u
m
s
am
p
les
wer
e
co
r
r
e
ctly
class
if
ied
.
Fo
r
th
e
h
ap
p
y
ca
teg
o
r
y
,
th
e
m
o
d
el
co
r
r
ec
tly
id
e
n
tifie
d
f
o
u
r
o
u
t
o
f
f
iv
e
s
am
p
les,
wh
ile
o
n
e
s
am
p
le
was
m
is
class
if
ied
as
tan
tr
u
m
d
u
e
to
s
im
ilar
ities
in
b
o
d
y
m
o
v
e
m
en
t
in
ten
s
ity
an
d
tr
an
s
itio
n
al
b
eh
av
io
r
al
p
atter
n
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
R
ea
l
-
time
fa
cia
l a
n
d
b
o
d
y
p
o
s
e
emo
tio
n
r
ec
o
g
n
itio
n
f
o
r
ch
ild
r
en
…
(
S
iti N
u
r
o
h
ma
h
w
a
ti
)
1907
T
h
e
co
n
f
u
s
io
n
m
atr
ix
an
aly
s
i
s
in
d
icate
s
th
at
th
e
p
r
o
p
o
s
ed
m
o
d
el
was
ca
p
ab
le
o
f
d
is
tin
g
u
is
h
in
g
r
e
p
etitiv
e
b
eh
av
io
r
al
m
o
v
em
en
ts
ass
o
ciate
d
with
d
if
f
er
en
t
em
o
tio
n
al
s
tates
in
ch
ild
r
en
with
ASD.
T
h
e
r
elativ
ely
s
m
all
n
u
m
b
er
o
f
m
is
class
if
icatio
n
s
d
em
o
n
s
tr
ates
th
e
ef
f
ec
tiv
e
n
ess
o
f
t
h
e
tem
p
o
r
al
f
ea
tu
r
e
ex
tr
a
ctio
n
m
ec
h
an
is
m
i
n
ca
p
tu
r
in
g
s
eq
u
en
tial
m
o
tio
n
c
h
ar
ac
ter
is
tics
.
Fu
r
th
er
m
o
r
e,
th
e
r
esu
lts
co
n
f
ir
m
th
at
i
n
teg
r
ati
n
g
b
o
d
y
m
o
v
em
e
n
t
an
aly
s
is
with
tem
p
o
r
al
m
o
d
elin
g
im
p
r
o
v
es
th
e
r
o
b
u
s
tn
ess
o
f
em
o
tio
n
r
ec
o
g
n
itio
n
u
n
d
er
d
y
n
am
ic
class
r
o
o
m
co
n
d
itio
n
s
.
Fig
u
r
e
9
.
Valid
atio
n
ac
cu
r
ac
y
an
d
v
alid
atio
n
lo
s
s
tr
en
d
s
d
u
r
in
g
L
T
SM
m
o
d
el
tr
ain
in
g
Fig
u
r
e
1
0
.
C
o
n
f
u
s
io
n
m
atr
ix
o
f
tem
p
o
r
al
e
m
o
tio
n
class
if
icatio
n
r
esu
lts
T
ab
le
2
s
u
m
m
ar
izes
t
h
e
p
er
f
o
r
m
an
ce
o
f
th
e
L
STM
-
b
ased
tem
p
o
r
al
e
m
o
tio
n
r
ec
o
g
n
itio
n
m
o
d
el
in
ter
m
s
o
f
p
r
ec
is
io
n
,
r
ec
all,
F1
-
s
co
r
e,
an
d
o
v
er
all
class
if
icatio
n
ac
cu
r
ac
y
.
T
h
e
n
eu
tr
al/b
o
r
ed
ca
teg
o
r
y
ac
h
iev
ed
p
er
f
ec
t
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
s
co
r
e
v
alu
es
o
f
1
.
0
0
,
in
d
icatin
g
h
ig
h
l
y
ac
cu
r
ate
an
d
s
tab
le
r
ec
o
g
n
itio
n
p
er
f
o
r
m
an
ce
f
o
r
n
o
n
-
a
ctiv
e
b
eh
av
io
r
al
s
tates.
T
h
e
tan
tr
u
m
ca
teg
o
r
y
also
d
em
o
n
s
tr
ated
s
tr
o
n
g
class
if
icatio
n
ca
p
ab
ilit
y
with
a
p
r
ec
is
io
n
v
alu
e
o
f
1
.
0
0
an
d
an
F1
-
s
co
r
e
o
f
0
.
9
1
,
wh
ile
th
e
h
a
p
p
y
ca
teg
o
r
y
ac
h
iev
e
d
an
F1
-
s
co
r
e
o
f
0
.
8
9
d
esp
ite
m
in
o
r
m
is
class
if
icatio
n
in
s
tan
ce
s
.
Ov
er
all,
th
e
p
r
o
p
o
s
ed
m
o
d
el
a
ch
iev
ed
a
class
if
icatio
n
ac
cu
r
ac
y
o
f
9
3
.
3
3
%,
with
m
ac
r
o
-
a
v
e
r
ag
ed
an
d
weig
h
ted
-
av
er
a
g
e
F1
-
s
co
r
es
o
f
0
.
9
3
.
T
h
ese
r
esu
lts
d
em
o
n
s
tr
ate
th
at
th
e
in
teg
r
atio
n
o
f
b
o
d
y
k
ey
p
o
in
t
ex
tr
ac
tio
n
an
d
te
m
p
o
r
al
s
eq
u
en
ce
m
o
d
elin
g
u
s
in
g
L
ST
M
p
r
o
v
id
es
r
eliab
le
p
e
r
f
o
r
m
an
ce
f
o
r
e
m
o
tio
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
1
6
,
No
.
4
,
Au
g
u
s
t
20
2
6
:
1
8
9
9
-
1
912
1908
r
ec
o
g
n
itio
n
b
ased
o
n
m
o
v
e
m
en
t
p
atter
n
s
.
C
o
m
p
ar
ed
with
co
n
v
en
ti
o
n
al
s
tatic
im
ag
e
-
b
ased
em
o
tio
n
r
ec
o
g
n
itio
n
ap
p
r
o
ac
h
es,
th
e
p
r
o
p
o
s
ed
tem
p
o
r
al
f
r
am
ewo
r
k
o
f
f
er
s
im
p
r
o
v
e
d
co
n
tex
tu
a
l
u
n
d
e
r
s
tan
d
in
g
o
f
b
eh
av
io
r
al
e
x
p
r
ess
io
n
s
an
d
s
u
p
p
o
r
ts
r
ea
l
-
tim
e
em
o
tio
n
al
m
o
n
ito
r
in
g
in
in
clu
s
iv
e
ed
u
c
atio
n
al
en
v
ir
o
n
m
en
ts
f
o
r
ch
ild
r
en
with
ASD.
T
ab
le
2
.
E
v
alu
atio
n
m
et
r
ics f
o
r
L
T
SM
m
o
d
el
p
er
f
o
r
m
an
ce
M
e
t
r
i
c
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
sc
o
r
e
S
u
p
p
o
r
t
N
e
u
t
r
a
l
/
B
o
r
e
d
1
.
0
0
1
.
0
0
1
.
0
0
5
H
a
p
p
y
0
.
8
0
1
.
0
0
0
.
8
9
4
Ta
n
t
r
u
m
1
.
0
0
0
.
8
3
0
.
9
1
6
A
c
c
u
r
a
c
y
0
.
9
3
0
.
9
3
0
.
9
3
15
M
a
c
r
o
a
v
g
0
.
9
3
0
.
9
4
0
.
9
3
15
W
e
i
g
h
t
e
d
a
v
g
0
.
9
5
0
.
9
3
0
.
9
3
15
3
.
4
.
Co
m
pa
ra
t
iv
e
a
na
ly
s
is
wit
h e
x
is
t
ing
m
et
ho
ds
T
o
ev
alu
ate
th
e
e
f
f
ec
tiv
en
ess
o
f
t
h
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
,
a
co
m
p
a
r
ativ
e
an
al
y
s
is
was
co
n
d
u
cte
d
ag
ain
s
t
s
ev
er
al
e
x
is
tin
g
em
o
ti
o
n
r
ec
o
g
n
itio
n
m
eth
o
d
s
r
ep
o
r
ted
in
p
r
ev
i
o
u
s
s
tu
d
ies.
T
h
e
c
o
m
p
ar
is
o
n
f
o
c
u
s
es
o
n
k
ey
asp
ec
ts
,
in
clu
d
in
g
d
etec
tio
n
ap
p
r
o
ac
h
,
in
p
u
t
m
o
d
ality
,
r
ea
l
-
tim
e
ca
p
ab
ilit
y
,
an
d
class
if
icatio
n
p
er
f
o
r
m
an
ce
.
T
h
is
a
n
aly
s
is
aim
s
to
id
e
n
tify
t
h
e
a
d
v
an
tag
es
an
d
lim
itatio
n
s
o
f
th
e
p
r
o
p
o
s
ed
m
u
ltimo
d
a
l
f
r
am
ewo
r
k
in
co
m
p
ar
is
o
n
wit
h
co
n
v
en
tio
n
al
s
in
g
le
-
m
o
d
alit
y
em
o
tio
n
r
ec
o
g
n
itio
n
s
y
s
tem
s
.
T
ab
le
3
p
r
esen
ts
a
co
m
p
ar
at
iv
e
an
aly
s
is
b
etwe
en
th
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
an
d
s
ev
er
a
l
ex
is
tin
g
em
o
tio
n
r
ec
o
g
n
itio
n
a
p
p
r
o
ac
h
es.
Pre
v
io
u
s
s
tu
d
ies
h
av
e
p
r
im
ar
ily
f
o
cu
s
ed
o
n
s
in
g
le
-
m
o
d
ality
em
o
tio
n
r
ec
o
g
n
itio
n
,
p
ar
ticu
lar
ly
f
ac
ia
l
-
ex
p
r
ess
io
n
-
b
ased
an
aly
s
is
.
Fo
r
in
s
tan
ce
,
T
alaa
t
[
1
6
]
in
te
g
r
ated
C
NN
-
b
ased
f
ac
ial
em
o
tio
n
r
ec
o
g
n
itio
n
with
I
o
T
tech
n
o
lo
g
y
f
o
r
r
ea
l
-
ti
m
e
m
o
n
ito
r
i
n
g
.
Nev
e
r
th
eless
,
th
e
s
y
s
tem
r
eq
u
ir
e
d
d
ir
ec
t f
ac
ial
o
r
ien
tatio
n
to
war
d
th
e
ca
m
er
a
a
n
d
d
i
d
n
o
t
u
tili
ze
b
o
d
y
m
o
v
e
m
en
t f
ea
tu
r
es,
lim
itin
g
its
r
o
b
u
s
tn
ess
in
d
y
n
am
ic
class
r
o
o
m
en
v
ir
o
n
m
en
ts
.
Similar
ly
,
Ho
s
n
e
y
et
a
l
.
[
2
9
]
d
ev
elo
p
ed
a
n
atten
tio
n
-
e
n
h
an
ce
d
YOL
Ov
8
f
r
am
ewo
r
k
f
o
r
f
ac
ial
em
o
tio
n
r
ec
o
g
n
itio
n
,
ac
h
ie
v
in
g
h
ig
h
class
if
icatio
n
ac
cu
r
ac
y
o
f
9
7
.
2
%.
Ho
wev
er
,
th
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
r
elied
m
ai
n
ly
o
n
f
ac
ial
ex
p
r
ess
io
n
s
an
d
lack
ed
tem
p
o
r
al
b
e
h
av
io
r
al
an
aly
s
is
.
I
n
ad
d
itio
n
,
Z
h
an
g
et
a
l.
[
3
0
]
u
tili
ze
d
YO
L
Ov
8
with
an
atten
tio
n
m
ec
h
an
is
m
to
a
n
aly
ze
b
o
d
y
b
eh
av
i
o
r
,
ac
h
iev
in
g
r
ea
l
-
tim
e
p
er
f
o
r
m
an
ce
.
Ho
wev
e
r
,
th
e
m
eth
o
d
f
o
c
u
s
ed
p
r
i
m
ar
ily
o
n
b
e
h
av
io
r
r
ec
o
g
n
itio
n
r
ath
er
th
an
co
m
p
r
eh
e
n
s
iv
e
em
o
tio
n
al
s
tate
class
if
icat
io
n
.
Mo
r
e
r
ec
en
tly
,
San
g
ee
th
a
et
a
l.
[
3
1
]
p
r
o
p
o
s
ed
a
m
u
ltimo
d
al
S
p
atio
-
tem
p
o
r
al
m
u
ltimo
d
al
em
o
tio
n
r
ec
o
g
n
itio
n
n
etwo
r
k
(
ST
-
ME
R
N)
in
te
g
r
atin
g
L
STM
,
T
C
N,
an
d
B
iLST
M
-
Atten
tio
n
ar
ch
itectu
r
es
with
f
ac
ial
lan
d
m
ar
k
s
,
g
a
ze
,
ac
tio
n
u
n
its
(
AUs),
g
eo
m
etr
ic
f
ea
tu
r
es,
an
d
laten
t
p
o
s
e
d
escr
ip
to
r
s
.
Alth
o
u
g
h
th
e
f
r
a
m
ewo
r
k
ac
h
iev
ed
h
ig
h
class
if
icatio
n
ac
cu
r
ac
y
o
f
9
4
.
3
%,
th
e
ap
p
r
o
ac
h
in
v
o
lv
e
d
h
ig
h
co
m
p
u
tatio
n
al
co
m
p
le
x
ity
,
d
ep
e
n
d
en
cy
o
n
lar
g
e
an
n
o
tate
d
d
a
tasets
,
an
d
lim
ite
d
lig
h
tweig
h
t I
o
T
-
b
ased
d
e
p
lo
y
m
en
t f
o
r
p
r
ac
tical
r
ea
l
-
tim
e
m
o
n
ito
r
in
g
ap
p
licatio
n
s
.
T
ab
le
3
.
C
o
m
p
a
r
ativ
e
an
aly
s
is
b
etwe
en
th
e
p
r
o
p
o
s
ed
f
r
am
ew
o
r
k
with
ex
is
tin
g
m
eth
o
d
s
S
t
u
d
y
M
e
t
h
o
d
M
o
d
a
l
i
t
y
R
e
a
l
-
Ti
me
C
a
p
a
b
i
l
i
t
y
A
c
c
u
r
a
c
y
Li
mi
t
a
t
i
o
n
Ta
l
a
a
t
[
1
6
]
C
N
N
-
b
a
s
e
d
f
a
c
i
a
l
e
mo
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
a
n
d
I
o
T
F
a
c
i
a
l
e
x
p
r
e
ss
i
o
n
Y
e
s
9
0
.
5
%
R
e
q
u
i
r
e
s
d
i
r
e
c
t
f
a
c
e
o
r
i
e
n
t
a
t
i
o
n
a
n
d
d
o
e
s
n
o
t
u
t
i
l
i
z
e
b
o
d
y
m
o
v
e
me
n
t
f
e
a
t
u
r
e
s
H
o
sn
e
y
e
t
a
l
.
[
2
9
]
A
t
t
e
n
t
i
o
n
-
e
n
h
a
n
c
e
d
Y
O
LO
v
8
F
a
c
i
a
l
e
x
p
r
e
ss
i
o
n
Y
e
s
9
7
.
2
%
R
e
l
i
e
s
o
n
l
y
o
n
f
a
c
i
a
l
e
x
p
r
e
s
si
o
n
s
a
n
d
l
a
c
k
s
t
e
m
p
o
r
a
l
m
o
v
e
me
n
t
a
n
a
l
y
s
i
s
Zh
a
n
g
e
t
a
l
.
[
3
0
]
Y
O
LO
v
8
w
i
t
h
a
t
t
e
n
t
i
o
n
m
e
c
h
a
n
i
sm
B
o
d
y
b
e
h
a
v
i
o
r
Y
e
s
9
1
.
3
%
F
o
c
u
ses
o
n
b
e
h
a
v
i
o
r
r
e
c
o
g
n
i
t
i
o
n
r
a
t
h
e
r
t
h
a
n
e
m
o
t
i
o
n
a
l
s
t
a
t
e
c
l
a
ssi
f
i
c
a
t
i
o
n
S
a
n
g
e
e
t
h
a
e
t
a
l
.
[
3
1
]
M
u
l
t
i
m
o
d
a
l
S
T
-
M
E
R
N
:
LS
TM
+
TC
N
+
B
i
LST
M
-
A
t
t
e
n
t
i
o
n
F
a
c
i
a
l
l
a
n
d
mar
k
s,
p
o
s
e
,
g
a
z
e
,
A
u
s,
g
e
o
me
t
r
i
c
f
e
a
t
u
r
e
s,
a
n
d
l
a
t
e
n
t
p
o
se
d
e
s
c
r
i
p
t
o
r
s
Y
e
s
9
4
.
3
%
H
i
g
h
c
o
m
p
u
t
a
t
i
o
n
a
l
c
o
m
p
l
e
x
i
t
y
,
d
e
p
e
n
d
e
n
c
y
o
n
l
a
r
g
e
a
n
n
o
t
a
t
e
d
d
a
t
a
se
t
s,
a
n
d
l
i
mi
t
e
d
l
i
g
h
t
w
e
i
g
h
t
I
o
T
-
b
a
s
e
d
d
e
p
l
o
y
m
e
n
t
f
o
r
p
r
a
c
t
i
c
a
l
r
e
a
l
-
t
i
m
e
m
o
n
i
t
o
r
i
n
g
a
p
p
l
i
c
a
t
i
o
n
s.
P
r
o
p
o
se
d
M
e
t
h
o
d
Y
O
LO
v
8
+
K
e
y
p
o
i
n
t
D
e
t
e
c
t
i
o
n
+
LST
M
+
I
o
T
F
a
c
i
a
l
e
x
p
r
e
ss
i
o
n
a
n
d
b
o
d
y
mo
v
e
me
n
t
Y
e
s
9
2
.
0
%
(
f
a
c
i
a
l
e
x
p
r
e
ss
i
o
n
)
9
5
.
0
%
(
m
o
v
e
me
n
t
p
a
t
t
e
r
n
)
P
e
r
f
o
r
ma
n
c
e
ma
y
b
e
a
f
f
e
c
t
e
d
b
y
u
n
s
t
a
b
l
e
l
i
g
h
t
i
n
g
c
o
n
d
i
t
i
o
n
s
I
n
co
n
tr
ast,
th
e
p
r
o
p
o
s
ed
f
r
a
m
ewo
r
k
in
teg
r
ates
YOL
Ov
8
-
b
ased
f
ac
ial
ex
p
r
ess
io
n
r
ec
o
g
n
itio
n
,
b
o
d
y
k
ey
p
o
i
n
t
ex
tr
ac
tio
n
,
an
d
L
ST
M
-
b
ased
tem
p
o
r
al
m
o
d
elin
g
to
im
p
r
o
v
e
em
o
tio
n
r
ec
o
g
n
iti
o
n
p
er
f
o
r
m
an
ce
in
r
ea
l
-
wo
r
ld
class
r
o
o
m
e
n
v
ir
o
n
m
en
ts
.
T
h
e
in
teg
r
atio
n
o
f
f
ac
ial
ex
p
r
ess
io
n
an
d
b
o
d
y
m
o
v
em
en
t
a
n
aly
s
is
Evaluation Warning : The document was created with Spire.PDF for Python.