I
A
E
S
I
n
t
e
r
n
at
io
n
al
Jou
r
n
al
of
A
r
t
if
ic
ia
l
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
AI
)
V
ol
.
14
, N
o.
5
,
O
c
to
be
r
2025
, pp.
3520
~
3527
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
14
.i
5
.pp
3520
-
3527
3520
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
O
b
j
e
c
t
d
e
t
e
c
t
i
on
f
or
i
n
d
oor
m
ob
i
l
e
r
ob
ot
:
d
e
e
p
l
e
ar
n
i
n
g
ap
p
r
oac
h
e
s r
e
vi
e
w
H
in
d
M
e
s
s
b
ah
, M
oh
am
e
d
E
m
h
ar
r
a
f
,
M
oh
am
m
e
d
S
ab
e
r
S
m
a
r
t
I
nf
o
r
m
a
t
i
on, C
om
m
uni
c
a
t
i
on a
nd T
e
c
hnol
ogi
e
s
L
a
bor
a
t
or
y
, N
a
t
i
ona
l
S
c
hool
of
A
ppl
i
e
d S
c
i
e
nc
e
s
, M
oha
m
m
e
d
F
i
r
s
t
U
ni
ve
r
s
i
t
y,
O
uj
da
, M
or
oc
c
o
A
r
t
ic
le
I
n
f
o
A
B
S
T
R
A
C
T
A
r
ti
c
le
h
is
to
r
y
:
R
e
c
e
iv
e
d
O
c
t
24, 2024
R
e
vi
s
e
d
J
ul
19, 2025
A
c
c
e
pt
e
d
A
ug 6, 2025
Efficient
object
detection
is
crucial
for
enabling
autonomous
indoor
robot
navigation.
This
paper
reviews
current
methodologies
and
challenges
in
the
field, wit
h a focus on
deep
learning
-
based techniques. Methods like
yo
u only
look
once
(
YOLO
)
,
region
-
based
convolutional
neural
networks
(
R
-
CNN
)
,
and
Faster
R
-
CNN
are
explored
for
their
suitability
in
real
-
time
detec
tion
in
dynamic
indoor
environments.
Deep
learning
models
are
emphasiz
ed
for
their
ability
to
improve
detection
accuracy
and
adaptability
to
v
ary
ing
conditi
ons. Key
performance m
etrics su
ch as accuracy,
speed, an
d scalabi
lity
across
different
object
types
and
environmental
scenarios
are
disc
ussed.
Additionally
,
the
integration
of
object
detection
with
navigation
systems
is
examined,
highli
ghting
the
import
ance
of
accurate
perception
for
sa
fe
and
effective
robot
movement
.
This
study
provides
insigh
ts
into
future
re
search
directions
aimed
at
advancing
the
capabilities
of
indoor
robot
navi
gation
through enha
nced dee
p learning
-
based object detection tec
hniques.
K
e
y
w
o
r
d
s
:
D
e
e
p l
e
a
r
ni
ng
I
ndoor
r
obot
na
vi
ga
ti
on
O
bj
e
c
t
de
te
c
ti
on
R
e
a
l
ti
m
e
S
e
ns
or
f
us
io
n
Y
ou only l
ook onc
e
This is an
open
acce
ss artic
le unde
r the
CC BY
-
SA
license.
C
or
r
e
s
pon
di
n
g A
u
th
or
:
H
in
d M
e
s
s
ba
h
S
m
a
r
t
I
nf
o
r
m
a
ti
on, C
om
m
uni
c
a
ti
on a
nd T
e
c
hnol
ogi
e
s
L
a
bor
a
to
r
y
,
N
a
ti
ona
l
S
c
hool
of
A
ppl
ie
d S
c
ie
nc
e
s
M
oha
m
m
e
d F
ir
s
t
U
ni
ve
r
s
it
y
B
P
669 B
d M
oha
m
m
e
d V
I
, O
uj
da
60000,
M
or
oc
c
o
E
m
a
il
:
h.m
e
s
ba
hi
@
um
p.a
c
.m
a
1.
I
N
T
R
O
D
U
C
T
I
O
N
T
he
f
ie
ld
of
r
obot
ic
s
,
pa
r
ti
c
ul
a
r
ly
in
door
na
vi
ga
ti
on,
h
a
s
e
vol
v
e
d
s
ig
ni
f
ic
a
nt
ly
ove
r
th
e
p
a
s
t
de
c
a
de
,
w
it
h
a
c
r
it
ic
a
l
e
m
pha
s
is
on
de
ve
lo
pi
ng
r
obus
t
obj
e
c
t
de
te
c
ti
on
s
ys
te
m
s
th
a
t
e
nha
nc
e
th
e
a
bi
li
ty
of
r
obot
s
to
na
vi
ga
te
c
om
pl
e
x
e
nvi
r
onm
e
nt
s
.
T
hi
s
li
te
r
a
tu
r
e
r
e
vi
e
w
e
xpl
or
e
s
ke
y
a
dva
nc
e
m
e
nt
s
in
obj
e
c
t
d
e
te
c
ti
on
te
c
hnol
ogi
e
s
a
nd
th
e
ir
im
pl
ic
a
ti
ons
f
or
in
door
r
obot
na
vi
ga
ti
on,
dr
a
w
in
g
in
s
ig
ht
s
f
r
om
r
e
le
va
nt
s
tu
di
e
s
.
T
h
e
de
ve
lo
pm
e
nt
of
a
ut
onomous
m
obi
le
r
obot
s
(
A
M
R
s
)
ha
s
r
e
vol
ut
io
ni
z
e
d
va
r
io
us
in
dus
tr
ie
s
,
in
c
lu
di
ng
hum
a
ni
ta
r
ia
n
a
s
s
is
t
a
nc
e
,
a
ut
om
ot
iv
e
,
a
gr
ic
ul
tu
r
e
,
e
du
c
a
ti
on,
a
nd
he
a
lt
hc
a
r
e
[
1]
.
A
M
R
s
a
r
e
de
s
ig
ne
d
to
ope
r
a
te
in
unpr
e
di
c
ta
bl
e
a
nd
pa
r
ti
a
ll
y
unknown
e
nvi
r
onm
e
nt
s
,
r
e
qui
r
in
g
th
e
m
to
na
vi
ga
te
c
om
pl
e
x
s
p
a
c
e
s
w
hi
le
a
voi
di
ng
obs
ta
c
le
s
.
H
ow
e
ve
r
,
one
of
th
e
m
a
in
c
h
a
ll
e
nge
s
c
onf
r
ont
in
g
A
M
R
s
is
th
e
ir
a
bi
li
ty
to
pe
r
c
e
iv
e
a
nd
in
te
r
a
c
t
e
f
f
e
c
ti
ve
ly
w
it
h
th
e
ir
s
ur
r
oundings
[
2]
.
O
bj
e
c
t
de
t
e
c
ti
on
pl
a
ys
a
n
e
s
s
e
nt
i
a
l
r
ol
e
in
A
M
R
’
s
vi
s
io
n
s
ys
te
m
s
,
e
m
pow
e
r
in
g
r
obot
s
to
c
a
r
r
y
out
in
t
r
ic
a
te
ta
s
ks
a
nd
na
vi
ga
te
va
r
io
us
c
ha
ll
e
nge
s
[
3]
.
F
or
in
s
ta
nc
e
,
gr
a
s
p
de
te
c
ti
on
is
e
s
s
e
nt
ia
l
f
or
r
obot
s
to
c
ol
le
c
t
obj
e
c
ts
in
f
r
ont
of
th
e
m
,
w
he
r
e
a
s
dyna
m
ic
obs
ta
c
le
de
te
c
ti
on
is
vi
ta
l
f
or
r
e
a
l
-
ti
m
e
na
vi
ga
ti
on
[
4
]
.
T
o
a
c
hi
e
ve
a
c
c
ur
a
te
de
te
c
ti
on,
A
M
R
s
r
e
ly
on
a
c
om
bi
na
ti
on
o
f
s
e
ns
or
s
,
in
c
lu
di
ng
na
vi
ga
ti
on,
lo
c
a
li
z
a
ti
on,
a
nd
de
te
c
ti
on
s
ys
te
m
s
.
C
ur
r
e
nt
r
e
s
e
a
r
c
h
in
di
c
a
te
s
th
a
t
s
e
ns
or
te
c
hnol
ogy,
in
c
lu
di
ng
s
e
ns
or
f
us
io
n
a
nd
th
e
u
s
e
of
m
ul
ti
pl
e
s
e
n
s
or
s
,
c
a
n
s
ig
ni
f
ic
a
nt
ly
im
pa
c
t
th
e
qua
li
ty
of
in
f
or
m
a
ti
on
pe
r
c
e
iv
e
d
by
A
M
R
s
[
5]
.
C
om
put
e
r
vi
s
io
n
is
e
s
s
e
nt
i
a
l
f
or
num
e
r
ous
a
ppl
ic
a
ti
ons
in
a
ut
om
a
ti
on
a
nd
r
obot
ic
s
,
pa
r
ti
c
ul
a
r
ly
i
n obje
c
t
de
te
c
ti
on. F
ur
th
e
r
m
or
e
, e
xpl
a
in
a
bi
li
ty
i
s
a
c
r
it
ic
a
l
r
e
qui
r
e
m
e
nt
f
or
a
lg
o
r
it
hm
s
i
n r
obot
ic
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
O
bj
e
c
t
de
te
c
ti
on f
or
i
ndoor
m
obi
le
r
obot
:
de
e
p l
e
a
r
ni
ng appr
oa
c
he
s
r
e
v
ie
w
(
H
in
d M
e
s
s
bah
)
3521
a
ppl
ic
a
ti
ons
,
a
s
it
a
id
s
in
id
e
nt
if
yi
ng
a
nd
r
e
s
ol
vi
ng
pot
e
nt
ia
l
is
s
ue
s
[
6]
.
O
bj
e
c
t
de
te
c
ti
on
te
c
hni
que
s
li
ke
f
a
c
e
,
pe
de
s
tr
ia
n,
a
nd
obs
ta
c
le
de
te
c
ti
on
r
e
ly
on
s
upe
r
vi
s
e
d
l
e
a
r
ni
ng
in
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
(
A
I
)
,
ty
pi
c
a
ll
y
de
e
p
le
a
r
ni
ng
m
e
th
ods
.
T
he
m
e
th
ods
e
m
pl
oye
d
in
c
lu
de
s
in
gl
e
-
s
ta
g
e
de
te
c
to
r
s
li
ke
you
onl
y
lo
ok
onc
e
(
Y
O
L
O
)
,
s
in
gl
e
s
hot
m
ul
ti
box
de
te
c
to
r
(
SSD
)
,
a
nd
R
e
ti
na
N
e
t,
a
s
w
e
ll
a
s
two
-
s
ta
ge
de
t
e
c
to
r
s
,
s
uc
h
a
s
c
onvolut
io
na
l
ne
ur
a
l
ne
twor
ks
(
C
N
N
)
a
nd
F
a
s
te
r
r
e
gi
on
-
ba
s
e
d
c
onvolut
io
na
l
ne
ur
a
l
ne
twor
ks
(
R
-
C
N
N
)
[
7
]
,
[
8
]
.
T
he
pe
r
f
or
m
a
nc
e
of
s
e
ns
or
s
a
nd de
e
p l
e
a
r
ni
ng
a
lg
or
it
hm
s
i
n A
M
R
s
r
e
m
a
in
s
a
t
opi
c
of
ongoing di
s
c
us
s
io
n.
C
ont
e
m
por
a
r
y
r
e
s
e
a
r
c
h
in
de
e
p
le
a
r
ni
ng
ha
s
s
ig
ni
f
ic
a
nt
ly
in
f
lu
e
nc
e
d
th
e
de
s
ig
n
of
obj
e
c
t
de
te
c
ti
on
s
ys
te
m
s
f
or
in
door
A
M
R
s
,
pa
r
ti
c
ul
a
r
ly
in
te
r
m
s
of
a
c
c
ur
a
c
y,
a
da
pt
a
bi
li
ty
,
a
nd
de
pl
oym
e
nt
e
f
f
ic
ie
nc
y.
A
r
e
f
in
e
d
a
ppr
oa
c
h
ta
r
ge
ti
ng
s
m
a
ll
obj
e
c
t
de
t
e
c
ti
on
unde
r
c
lu
tt
e
r
e
d
a
nd
c
om
pl
e
x
s
c
e
ne
s
,
known a
s
Y
O
L
O
v8
-
Q
S
D
,
w
a
s
pr
opos
e
d
in
[
9]
,
a
ddr
e
s
s
in
g
ke
y c
ha
ll
e
nge
s
in
a
ut
onomous
in
door
s
ys
te
m
s
.
Y
O
L
O
v8’
s
c
a
pa
bi
li
ti
e
s
w
e
r
e
f
ur
th
e
r
e
xt
e
nde
d
th
r
ough
it
s
a
ppl
ic
a
ti
on
to
li
ght
d
e
te
c
t
io
n
a
nd
r
a
ngi
ng
(
L
iDA
R
)
poi
nt
c
lo
ud
da
t
a
,
de
m
ons
tr
a
ti
ng
im
pr
ove
d
s
pa
ti
a
l
pr
e
c
is
io
n
f
or
obj
e
c
t
de
te
c
ti
on
in
th
r
e
e
-
di
m
e
ns
io
na
l
e
nvi
r
onm
e
nt
s
[
10]
.
A
n
opt
im
iz
e
d
im
pl
e
m
e
nt
a
ti
on
of
Y
O
L
O
v8
w
a
s
pr
opos
e
d
to
ba
la
nc
e
s
pe
e
d
a
nd
c
om
put
a
ti
ona
l
c
ons
tr
a
in
ts
w
it
hout
s
a
c
r
if
ic
in
g
de
te
c
ti
on
r
e
li
a
bi
li
ty
[
11
]
.
M
ul
ti
-
s
c
a
le
f
e
a
tu
r
e
f
us
io
n
te
c
hni
que
s
ha
ve
be
e
n
in
tr
oduc
e
d
to
im
pr
ove
th
e
de
te
c
ti
on
of
va
r
ia
bl
y
s
iz
e
d
obj
e
c
t
s
in
c
lu
tt
e
r
e
d
e
nv
ir
onm
e
nt
s
by
le
ve
r
a
gi
ng
s
e
m
a
nt
ic
in
f
or
m
a
ti
on
a
c
r
os
s
di
f
f
e
r
e
nt
s
pa
ti
a
l
r
e
s
ol
ut
io
ns
[
12]
.
D
e
pt
h
-
a
w
a
r
e
ne
s
s
ha
s
a
ls
o
be
e
n
in
te
gr
a
t
e
d
in
to
obj
e
c
t
de
te
c
ti
on
pi
pe
li
ne
s
t
o e
nha
nc
e
pe
r
f
or
m
a
nc
e
i
n s
c
e
na
r
io
s
w
it
h oc
c
lu
s
io
ns
a
nd va
r
yi
ng obje
c
t
di
s
ta
nc
e
s
[
13]
. L
ig
ht
w
e
ig
ht
de
te
c
to
r
s
s
uc
h
a
s
E
f
f
ic
ie
nt
D
e
t
a
nd
M
obi
le
N
e
t
of
f
e
r
e
f
f
ic
ie
nt
tr
a
de
-
of
f
s
be
twe
e
n
a
c
c
ur
a
c
y
a
nd
pr
oc
e
s
s
in
g
r
e
qui
r
e
m
e
nt
s
,
m
a
ki
ng
th
e
m
s
ui
ta
bl
e
f
or
e
m
be
dde
d
s
ys
t
e
m
s
[
1
4]
.
F
in
a
ll
y,
s
e
lf
-
s
upe
r
vi
s
e
d
le
a
r
ni
ng
s
tr
a
te
gi
e
s
ha
ve
be
e
n
e
m
pl
oye
d
to
im
pr
ove
m
ode
l
ge
ne
r
a
li
z
a
ti
on
in
in
doo
r
e
nvi
r
onm
e
nt
s
w
hi
le
r
e
duc
in
g
th
e
r
e
li
a
nc
e
on
a
nnot
a
te
d da
ta
s
e
ts
-
a
n e
s
s
e
nt
ia
l
s
te
p f
or
s
c
a
la
bl
e
de
pl
oym
e
nt
of
a
ut
onomous
r
obot
s
[
15]
.
T
hi
s
s
tu
dy s
e
e
ks
to
e
va
lu
a
te
th
e
pe
r
f
or
m
a
nc
e
a
nd
th
e
de
te
c
ti
on
a
c
c
ur
a
c
y
of
de
e
p
l
e
a
r
ni
ng
te
c
hni
que
s
a
ppl
ie
d t
o A
M
R
s
.
T
he
l
it
e
r
a
tu
r
e
r
e
vi
e
w
a
nd r
e
s
ul
ts
a
na
ly
s
is
a
r
e
di
s
c
us
s
e
d i
n de
ta
il
, pr
ovi
di
ng i
ns
ig
ht
s
i
nt
o t
he
c
ur
r
e
nt
s
ta
te
of
obj
e
c
t
de
t
e
c
ti
on
te
c
hni
que
s
in
A
M
R
s
.
T
hi
s
s
tu
dy
f
ol
lo
w
s
a
s
tr
uc
tu
r
e
d
a
ppr
oa
c
h
c
ons
i
s
ti
ng
of
th
r
e
e
s
e
c
ti
ons
.
S
e
c
ti
on
1
in
tr
oduc
e
s
th
e
c
on
c
e
pt
of
obj
e
c
t
de
te
c
ti
on
in
A
M
R
s
.
S
e
c
ti
on
2
a
na
ly
z
e
s
th
e
c
ur
r
e
nt
s
ta
te
of
obj
e
c
t
-
de
te
c
ti
on
te
c
hni
que
s
in
A
M
R
s
.
S
e
c
ti
on
3
pr
ovi
de
s
in
s
ig
ht
s
in
to
th
e
c
ha
ll
e
nge
s
a
nd
oppor
tu
ni
ti
e
s
f
a
c
in
g
th
e
de
v
e
lo
pm
e
nt
of
obj
e
c
t
d
e
te
c
ti
on
te
c
h
ni
que
s
in
A
M
R
s
.
F
in
a
ll
y,
s
e
c
ti
on
4
c
onc
lu
d
e
s
th
e
di
s
c
us
s
io
n.
2.
M
E
T
H
O
D
A
dva
nc
e
m
e
nt
s
in
obj
e
c
t
de
te
c
ti
on
te
c
hnol
ogi
e
s
h
a
ve
s
ig
ni
f
ic
a
nt
ly
im
pa
c
te
d
va
r
io
us
f
ie
ld
s
,
pa
r
ti
c
ul
a
r
ly
in
r
e
m
ot
e
s
e
ns
in
g.
A
ke
y
f
oc
us
ha
s
be
e
n
on
de
te
c
ti
ng
s
m
a
ll
obj
e
c
ts
w
it
hi
n
va
s
t
im
a
ge
s
,
w
hi
c
h
pr
e
s
e
nt
s
uni
qu
e
c
ha
ll
e
nge
s
due
to
f
a
c
to
r
s
li
ke
r
e
s
ol
ut
io
n
a
nd
o
bj
e
c
t
or
ie
nt
a
ti
on
[
16]
.
R
e
c
e
nt
de
v
e
lo
pm
e
nt
s
in
de
e
p
le
a
r
ni
ng
te
c
hni
qu
e
s
, s
uc
h
a
s
th
e
Y
O
L
O
s
e
r
ie
s
a
nd
S
S
D
[
17]
,
ha
ve
not
a
bl
y
im
pr
ove
d
th
e
p
e
r
f
or
m
a
nc
e
of
th
e
s
e
de
te
c
ti
on
a
lg
or
it
hm
s
[
18]
.
S
m
a
ll
obj
e
c
t
de
te
c
ti
on
is
c
a
te
gor
iz
e
d
in
to
m
ul
ti
pl
e
s
tr
a
te
gi
e
s
,
in
c
lu
di
ng
m
ul
ti
-
s
c
a
le
pr
e
di
c
ti
ons
a
nd
e
nha
nc
e
d
f
e
a
tu
r
e
r
e
s
ol
ut
io
ns
[
16]
.
R
e
s
e
a
r
c
h
e
r
s
a
r
e
a
ls
o
a
ddr
e
s
s
in
g
ir
r
e
gul
a
r
it
ie
s
in
r
e
m
ot
e
s
e
ns
in
g i
m
a
ge
s
t
ha
t
c
om
pl
ic
a
te
de
te
c
ti
on e
f
f
or
ts
.
G
a
i
ni
ng i
ns
ig
ht
i
nt
o
th
e
s
e
m
e
th
odol
ogi
e
s
not
onl
y
im
pr
ove
s
th
e
e
f
f
e
c
ti
ve
ne
s
s
of
obj
e
c
t
de
te
c
ti
on
but
a
ls
o
hi
ghl
ig
ht
s
pot
e
nt
ia
l
f
ut
ur
e
r
e
s
e
a
r
c
h
a
ve
nue
s
in
hi
gh
-
r
e
s
ol
ut
io
n e
nvi
r
onm
e
nt
s
, w
he
r
e
de
te
c
ti
ng s
m
a
ll
obj
e
c
t
s
r
e
m
a
in
s
a
s
ig
ni
f
ic
a
nt
c
ha
ll
e
ng
e
[
19]
.
2.1. T
r
ad
it
io
n
al
c
om
p
u
t
e
r
vi
s
io
n
t
e
c
h
n
iq
u
e
s
T
r
a
di
ti
ona
l
c
om
put
e
r
vi
s
io
n
m
e
th
ods
ha
ve
b
e
e
n
e
m
pl
oye
d
f
o
r
obj
e
c
t
de
te
c
ti
on
in
va
r
io
us
f
ie
ld
s
,
in
c
lu
di
ng
r
e
m
ot
e
s
e
n
s
in
g.
T
h
e
s
e
tr
a
di
ti
ona
l
te
c
hni
que
s
r
e
ly
o
n
im
a
ge
pr
oc
e
s
s
in
g
a
nd
f
e
a
tu
r
e
e
xt
r
a
c
ti
on
to
id
e
nt
if
y
obj
e
c
ts
w
it
hi
n
im
a
ge
s
[
20]
.
T
a
bl
e
1
p
r
ovi
de
s
a
n
ove
r
vi
e
w
of
th
e
m
os
t
w
id
e
ly
a
dopt
e
d
tr
a
di
ti
ona
l
c
om
put
e
r
vi
s
io
n t
e
c
hni
que
s
e
m
pl
oye
d i
n obje
c
t
de
te
c
ti
on t
a
s
k
s
.
T
a
bl
e
1
.
T
r
a
di
ti
ona
l
c
om
put
e
r
vi
s
io
n m
e
th
ods
ove
r
vi
e
w
[
21]
–
[
25]
T
e
c
hni
que
D
e
s
c
r
i
pt
i
on
E
xa
m
pl
e
s
L
i
m
i
t
a
t
i
ons
E
dge
d
e
t
e
c
t
i
on
I
de
nt
i
f
i
e
s
obj
e
c
t
e
dge
s
w
i
t
hi
n a
n i
m
a
ge
.
S
ob
e
l
op
e
r
a
t
o
r
, C
a
nny
a
l
go
r
i
t
h
m
A
f
f
e
c
t
e
d
by
i
m
a
ge
noi
s
e
a
nd
i
l
l
um
i
na
t
i
on va
r
i
a
t
i
ons
.
T
e
m
pl
a
t
e
m
a
t
c
hi
n
g
C
om
pa
r
e
s
a
gi
ve
n
i
m
a
ge
w
i
t
h
s
t
or
e
d
t
e
m
pl
a
t
e
s
t
o de
t
e
c
t
obj
e
c
t
s
.
C
r
os
s
-
c
or
r
e
l
a
t
i
on,
nor
m
a
l
i
z
e
d
c
or
r
e
l
a
t
i
on c
oe
f
f
i
c
i
e
nt
(
N
C
C
)
C
om
put
a
t
i
ona
l
l
y
e
xpe
ns
i
ve
,
poor
pe
r
f
or
m
a
nc
e
w
i
t
h
l
a
r
ge
da
t
a
ba
s
e
s
.
F
e
a
t
u
r
e
e
xt
r
a
c
t
i
on
E
xt
r
a
c
t
s
r
e
l
e
va
nt
f
e
a
t
ur
e
s
f
r
om
a
n
i
m
a
ge
t
o de
s
c
r
i
be
obj
e
c
t
s
(
t
e
xt
ur
e
, c
ol
or
, s
ha
pe
)
.
H
i
s
t
ogr
a
m
of
or
i
e
nt
e
d
gr
a
di
e
nt
s
(
H
O
G
)
,
s
c
a
l
e
-
i
nva
r
i
a
nt
f
e
a
t
ur
e
t
r
a
ns
f
or
m
(
S
I
F
T
)
C
om
put
a
t
i
ona
l
l
y
e
xpe
ns
i
ve
w
i
t
h l
a
r
ge
i
m
a
ge
da
t
a
ba
s
e
s
.
H
O
G
a
nd S
I
F
T
H
O
G
de
s
c
r
i
be
s
o
bj
e
c
t
t
e
x
t
u
r
e
,
w
hi
l
e
S
I
F
T
de
s
c
r
i
b
e
s
ob
j
e
c
t
s
ha
pe
a
nd
o
r
i
e
n
t
a
t
i
o
n
f
o
r
obj
e
c
t
r
e
c
o
gni
t
i
on
.
H
O
G
f
or
pe
de
s
t
r
i
a
n
de
t
e
c
t
i
on,
S
I
F
T
f
or
obj
e
c
t
t
r
a
c
ki
ng
M
a
y
not
pe
r
f
or
m
w
e
l
l
on
l
a
r
ge
da
t
a
s
e
t
s
,
a
nd
hi
gh
c
om
put
a
t
i
ona
l
c
os
t
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
5
,
O
c
to
be
r
2025
:
3520
-
3527
3522
2.2. De
e
p
le
ar
n
in
g ap
p
r
oac
h
e
s
R
e
c
e
nt
a
dv
a
n
c
e
m
e
nt
s
le
ve
r
a
ge
C
N
N
s
a
nd
d
e
e
p
l
e
a
r
ni
ng
p
a
r
a
di
gm
s
.
N
ot
a
bl
e
f
r
a
m
e
w
or
k
s
in
c
l
ud
e
Y
O
L
O
, S
S
D
,
a
nd
F
a
s
te
r
R
-
C
N
N
,
w
hi
c
h
h
a
v
e
e
n
a
b
le
d r
e
a
l
-
ti
m
e
o
bj
e
c
t
d
e
t
e
c
ti
o
n
w
it
h
s
ig
ni
f
ic
a
n
t
a
c
c
ur
a
c
y
[
26]
.
2.2.1. S
in
gl
e
-
s
t
age
m
od
e
ls
S
in
gl
e
-
s
ta
ge
obj
e
c
t
de
te
c
to
r
s
,
s
uc
h
a
s
Y
O
L
O
a
nd
S
S
D
,
pe
r
f
or
m
de
te
c
ti
on
in
a
s
in
gl
e
f
or
w
a
r
d
pa
s
s
.
T
he
y
a
r
e
opt
im
iz
e
d
f
or
s
pe
e
d
a
nd
a
r
e
s
ui
ta
bl
e
f
or
r
e
a
l
-
ti
m
e
a
ppl
ic
a
ti
ons
w
it
hi
n
in
door
e
nvi
r
onm
e
nt
s
[
27]
:
i)
Y
O
L
O
known
f
or
it
s
a
bi
li
ty
to
pe
r
f
or
m
r
e
a
l
-
ti
m
e
obj
e
c
t
d
e
te
c
ti
on.
I
t
s
e
gm
e
nt
s
th
e
im
a
g
e
in
to
a
gr
id
a
nd
pr
e
di
c
ts
bounding
boxe
s
a
lo
ng
w
it
h
c
or
r
e
s
ponding
c
la
s
s
pr
ob
a
bi
li
ti
e
s
di
r
e
c
tl
y.
Y
O
L
O
’
s
f
a
s
t
in
f
e
r
e
nc
e
ti
m
e
m
a
ke
s
it
s
ui
ta
bl
e
f
or
a
ppl
ic
a
ti
on
s
r
e
qui
r
in
g
r
e
a
l
-
ti
m
e
de
c
is
io
ns
,
s
uc
h
a
s
in
door
r
obot
ic
s
[
28]
a
nd
i
i)
S
S
D
s
im
il
a
r
to
Y
O
L
O
,
S
S
D
p
e
r
f
or
m
s
ob
je
c
t
d
e
t
e
c
ti
on
in
a
s
in
g
le
f
o
r
w
a
r
d
pa
s
s
,
but
i
t
di
v
id
e
s
th
e
im
a
g
e
a
t
m
ul
t
ip
l
e
s
c
a
l
e
s
.
T
hi
s
e
nh
a
n
c
e
s
i
t
s
a
c
c
ur
a
c
y,
pa
r
ti
c
ul
a
r
l
y f
or
s
m
a
ll
e
r
ob
je
c
t
s
,
w
hi
c
h
is
c
r
u
c
i
a
l
i
n
in
d
oor
e
nvi
r
on
m
e
nt
s
[
29]
.
2.2.2. T
w
o
-
s
t
age
m
od
e
ls
T
w
o
-
s
ta
ge
m
ode
ls
,
s
u
c
h
a
s
F
a
s
te
r
R
-
C
N
N
,
f
ir
s
t
pr
opos
e
r
e
gi
ons
of
in
te
r
e
s
t
a
nd
th
e
n
c
la
s
s
if
y
th
os
e
r
e
gi
ons
.
T
he
s
e
m
ode
l
s
of
f
e
r
hi
ghe
r
a
c
c
ur
a
c
y
but
ge
ne
r
a
ll
y
r
e
qui
r
e
m
or
e
c
om
put
a
ti
ona
l
r
e
s
our
c
e
s
[
30]
:
i)
C
N
N
s
ha
ve
be
c
om
e
th
e
f
ounda
ti
on
f
or
obj
e
c
t
de
te
c
ti
on
m
o
de
ls
due
to
th
e
ir
a
bi
li
ty
to
le
a
r
n
f
e
a
t
ur
e
s
f
r
om
im
a
ge
s
a
ut
om
a
ti
c
a
ll
y.
A
r
c
hi
te
c
tu
r
e
s
s
u
c
h
a
s
A
le
xN
e
t,
V
G
G
,
a
nd
R
e
s
N
e
t
ha
ve
s
e
t
th
e
s
ta
ge
f
or
m
or
e
s
ophi
s
ti
c
a
te
d
d
e
te
c
ti
on
m
ode
ls
[
31]
a
nd
ii
)
R
-
C
N
N
th
e
ir
va
r
ia
nt
s
(
F
a
s
t
R
-
C
N
N
,
F
a
s
te
r
R
-
C
N
N
)
f
ir
s
t
ge
ne
r
a
te
r
e
gi
on
pr
opos
a
ls
a
nd
th
e
n
p
e
r
f
or
m
c
la
s
s
if
ic
a
ti
on
a
nd
bounding
box
r
e
gr
e
s
s
io
n.
T
h
e
s
e
m
ode
ls
a
r
e
known
f
or
hi
gh a
c
c
ur
a
c
y but a
r
e
c
om
put
a
ti
ona
ll
y i
nt
e
ns
iv
e
,
w
hi
c
h m
a
y l
i
m
it
t
he
ir
r
e
a
l
-
ti
m
e
a
ppl
ic
a
bi
li
ty
[
32
]
.
2.3. At
t
e
n
t
io
n
m
e
c
h
an
is
m
an
d
t
r
an
s
f
e
r
l
e
a
r
n
in
g
A
r
e
c
e
nt
a
dva
nc
e
m
e
nt
i
n obje
c
t
de
te
c
ti
on a
nd r
e
c
ogni
ti
on t
e
c
hn
ol
ogi
e
s
i
s
t
he
us
e
of
t
r
a
ns
f
e
r
l
e
a
r
ni
ng
c
om
bi
ne
d
w
it
h
a
tt
e
nt
io
n
m
e
c
ha
ni
s
m
s
.
T
hi
s
c
ut
ti
ng
-
e
dge
a
ppr
oa
c
h
e
nha
nc
e
s
th
e
m
ode
l'
s
a
bi
li
ty
to
f
oc
us
on
im
por
ta
nt
pa
r
ts
of
a
n
im
a
ge
by
dyna
m
ic
a
ll
y
w
e
ig
ht
in
g
th
e
s
ig
ni
f
ic
a
nc
e
of
di
f
f
e
r
e
nt
f
e
a
tu
r
e
s
,
e
na
bl
in
g
m
o
r
e
a
c
c
ur
a
te
a
nd
e
f
f
ic
ie
nt
obj
e
c
t
de
te
c
ti
on
[
33]
.
T
r
a
ns
f
e
r
le
a
r
ni
ng
a
ll
ow
s
m
ode
ls
pr
e
-
tr
a
in
e
d
on
la
r
ge
da
ta
s
e
ts
to
be
f
in
e
-
tu
ne
d f
or
s
pe
c
if
ic
t
a
s
ks
, s
ig
ni
f
ic
a
nt
ly
r
e
duc
in
g t
r
a
in
in
g t
im
e
a
nd i
m
pr
ovi
ng pe
r
f
or
m
a
nc
e
, e
s
pe
c
ia
ll
y i
n
li
m
it
e
d
da
ta
s
c
e
na
r
io
s
[
34]
.
T
he
in
te
gr
a
ti
on
of
a
tt
e
nt
io
n
m
e
c
h
a
ni
s
m
s
w
it
hi
n
th
e
s
e
a
r
c
hi
te
c
tu
r
e
s
r
e
pr
e
s
e
nt
s
a
m
a
jo
r
le
a
p
f
or
w
a
r
d,
im
pr
ovi
ng
th
e
pr
e
c
is
io
n
a
nd
s
pe
e
d
of
r
e
c
o
gni
ti
on
s
ys
te
m
s
in
bot
h
r
e
a
l
-
ti
m
e
a
nd
c
om
pl
e
x
e
nvi
r
onm
e
nt
s
[
35]
.
O
bj
e
c
t
de
te
c
ti
on
r
e
f
e
r
s
to
th
e
a
bi
li
ty
to
i
de
nt
if
y
a
nd
lo
c
a
te
obj
e
c
t
s
w
it
hi
n
a
n
im
a
g
e
.
T
r
a
di
ti
ona
l
m
e
th
ods
of
te
n
r
e
li
e
d
on
ha
nd
-
c
r
a
f
te
d
f
e
a
tu
r
e
s
;
how
e
ve
r
,
w
it
h
th
e
a
dve
nt
of
de
e
p
le
a
r
ni
ng,
C
N
N
s
ha
ve
be
c
om
e
t
he
s
ta
nda
r
d a
ppr
oa
c
h due
t
o t
he
ir
a
bi
li
ty
t
o a
ut
o
m
a
ti
c
a
ll
y l
e
a
r
n f
e
a
tu
r
e
s
f
r
om
da
ta
.
3.
R
E
S
U
L
T
S
A
N
D
D
I
S
C
U
S
S
I
O
N
C
om
put
e
r
vi
s
io
n
ha
s
e
vol
ve
d
dr
a
m
a
ti
c
a
ll
y,
m
ovi
ng
f
r
om
c
onv
e
nt
io
na
l
a
ppr
oa
c
he
s
to
th
e
a
dva
nc
e
d
te
c
hni
que
s
of
de
e
p
le
a
r
ni
ng.
T
r
a
di
ti
ona
l
c
om
put
e
r
vi
s
io
n,
w
hi
c
h
in
vol
ve
s
te
a
c
hi
ng
c
om
put
e
r
s
to
unde
r
s
ta
nd
im
a
ge
s
th
r
ough
s
pe
c
if
ic
,
pr
ogr
a
m
m
e
d
r
ul
e
s
,
ha
s
be
e
n
la
r
ge
l
y
r
e
pl
a
c
e
d
by
de
e
p
le
a
r
ni
ng
te
c
hni
que
s
th
a
t
e
na
bl
e
c
om
put
e
r
s
to
le
a
r
n
f
r
om
la
r
ge
a
m
ount
s
of
da
ta
,
w
hi
le
tr
a
di
ti
ona
l
m
e
th
ods
r
e
li
e
d
on
de
te
c
ti
ng
s
pe
c
if
ic
f
e
a
tu
r
e
s
li
ke
e
dge
s
,
s
ha
pe
s
,
or
te
xt
ur
e
s
u
s
in
g
a
lg
or
it
hm
s
,
de
e
p
le
a
r
ni
ng
ha
s
r
e
vol
ut
io
ni
z
e
d
im
a
ge
in
te
r
pr
e
ta
ti
on
by
e
m
pow
e
r
in
g
c
om
put
e
r
s
to
le
a
r
n
f
r
om
da
ta
a
nd
a
dj
us
t
to
a
br
oa
d
r
a
nge
of
ta
s
ks
.
T
he
k
e
y
di
f
f
e
r
e
nc
e
s
be
twe
e
n
tr
a
di
ti
ona
l
c
om
put
e
r
vi
s
io
n
a
nd
de
e
p
le
a
r
ni
ng
li
e
in
th
e
ir
da
ta
d
e
pe
nde
nc
y,
c
om
put
a
ti
ona
l
pow
e
r
,
f
le
xi
bi
li
ty
,
a
nd
a
c
c
ur
a
c
y
,
a
s
s
how
n
in
F
ig
ur
e
1
.
D
e
e
p
le
a
r
ni
ng
m
ode
ls
,
w
hi
c
h a
r
e
ba
s
e
d
on
ne
ur
a
l
ne
twor
ks
,
c
a
n
ha
ndl
e
c
om
pl
e
x
pa
tt
e
r
ns
a
nd
la
r
ge
-
s
c
a
le
im
a
ge
da
ta
,
a
nd
c
a
n
a
ut
om
a
ti
c
a
ll
y
a
dj
us
t
to
di
ve
r
s
e
ta
s
ks
w
it
hout
be
in
g
e
xpl
ic
it
ly
pr
ogr
a
m
m
e
d
f
or
e
a
c
h
n
e
w
pr
obl
e
m
.
I
n
c
ont
r
a
s
t,
tr
a
di
ti
ona
l
m
e
th
ods
r
e
qui
r
e
m
or
e
hum
a
n
gui
da
nc
e
to
de
f
in
e
f
e
a
tu
r
e
s
a
nd
a
r
e
of
te
n
le
s
s
a
c
c
ur
a
te
th
a
n
de
e
p
le
a
r
ni
ng
f
or
c
om
pl
e
x
vi
s
io
n t
a
s
ks
,
s
uc
h a
s
i
m
a
ge
r
e
c
ogni
ti
on a
nd obje
c
t
de
t
e
c
ti
on i
n va
r
ie
d c
ondi
ti
ons
[
36]
.
T
r
a
di
ti
ona
l
c
om
put
e
r
vi
s
io
n
a
nd
de
e
p
le
a
r
ni
ng
a
r
e
not
m
ut
ua
ll
y
e
xc
lu
s
iv
e
,
but
r
a
th
e
r
c
om
pl
e
m
e
nt
a
r
y
f
ie
ld
s
th
a
t
c
a
n
in
f
o
r
m
a
nd
e
nha
nc
e
e
a
c
h
ot
he
r
[
37
]
.
B
y
s
tu
dyi
n
g
tr
a
di
ti
ona
l
c
om
put
e
r
vi
s
io
n
te
c
hni
que
s
,
one
c
a
n
ga
in
a
de
e
pe
r
unde
r
s
ta
ndi
ng
of
th
e
f
unda
m
e
nt
a
l
pr
in
c
ip
le
s
of
im
a
ge
pr
oc
e
s
s
in
g
a
nd
f
e
a
tu
r
e
e
xt
r
a
c
ti
on,
w
hi
c
h
a
r
e
e
s
s
e
nt
ia
l
f
or
de
e
p
le
a
r
ni
ng
m
ode
ls
.
C
onve
r
s
e
ly
,
knowle
dge
of
de
e
p
le
a
r
ni
ng
c
a
n
pr
ovi
de
ne
w
in
s
ig
ht
s
a
nd
te
c
hni
que
s
f
or
im
pr
ovi
ng
tr
a
di
ti
ona
l
c
om
put
e
r
v
i
s
io
n
m
e
th
ods
.
U
lt
im
a
te
ly
,
th
e
in
te
r
s
e
c
ti
on
of
tr
a
di
ti
ona
l
c
om
put
e
r
vi
s
io
n
a
nd
de
e
p
le
a
r
ni
ng
c
a
n
l
e
a
d
to
m
or
e
e
f
f
e
c
ti
ve
a
nd
e
f
f
ic
ie
nt
c
om
put
e
r
vi
s
io
n
s
ol
ut
io
ns
, m
a
ki
ng one
a
m
or
e
s
ki
ll
e
d a
nd v
e
r
s
a
ti
le
e
xpe
r
t
in
t
he
f
ie
ld
.
M
ode
r
n
a
ppr
oa
c
he
s
to
obj
e
c
t
de
te
c
ti
on
c
a
n
be
c
a
te
gor
iz
e
d
in
to
s
in
gl
e
-
s
ta
ge
a
nd
two
-
s
ta
ge
de
te
c
to
r
s
[
7]
.
S
in
gl
e
-
s
ta
ge
de
te
c
to
r
s
,
s
uc
h
a
s
Y
O
L
O
a
nd
S
S
D
,
pe
r
f
or
m
de
te
c
ti
on
in
a
s
in
gl
e
s
te
p,
c
om
bi
ni
ng
c
la
s
s
if
ic
a
ti
on
a
nd
bounding
box
r
e
gr
e
s
s
io
n.
T
he
s
e
m
ode
ls
a
r
e
ty
pi
c
a
ll
y
f
a
s
te
r
a
nd
s
im
pl
e
r
,
m
a
ki
ng
th
e
m
s
ui
ta
bl
e
f
or
r
e
a
l
-
ti
m
e
a
ppl
ic
a
ti
ons
,
but
th
e
y
m
a
y
c
om
pr
om
is
e
on
a
c
c
ur
a
c
y
due
to
th
e
ir
s
tr
e
a
m
li
ne
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
O
bj
e
c
t
de
te
c
ti
on f
or
i
ndoor
m
obi
le
r
obot
:
de
e
p l
e
a
r
ni
ng appr
oa
c
he
s
r
e
v
ie
w
(
H
in
d M
e
s
s
bah
)
3523
a
r
c
hi
te
c
tu
r
e
[
38]
.
I
n
c
ont
r
a
s
t,
two
-
s
ta
ge
de
te
c
to
r
s
,
li
ke
R
-
C
N
N
a
nd
F
a
s
te
r
R
-
C
N
N
,
ope
r
a
te
in
two
pha
s
e
s
:
f
ir
s
t,
th
e
y
ge
ne
r
a
te
r
e
gi
on
of
in
te
r
e
s
t
(
R
oI
)
pr
opos
a
ls
,
a
nd
t
he
n
th
e
s
e
r
e
gi
on
s
a
r
e
f
ur
th
e
r
c
la
s
s
if
ie
d
a
nd
r
e
f
in
e
d. W
hi
le
t
hi
s
m
e
th
od p
r
ovi
de
s
hi
ghe
r
a
c
c
ur
a
c
y, pa
r
ti
c
ul
a
r
ly
t
hr
ough R
oI
pooli
ng, i
t
c
om
e
s
a
t
th
e
c
os
t
of
in
c
r
e
a
s
e
d c
om
put
a
ti
ona
l
ti
m
e
a
nd
c
om
pl
e
xi
ty
[
30]
.
O
bj
e
c
t
de
te
c
ti
on
is
de
f
in
e
d
a
s
id
e
nt
if
yi
ng
obj
e
c
t
in
s
ta
nc
e
s
f
r
om
pr
e
de
f
in
e
d
c
a
te
gor
ie
s
w
it
hi
n a
gi
ve
n
r
e
gi
on,
a
s
di
s
c
u
s
s
e
d
by
[
39]
.
T
hi
s
a
ppr
oa
c
h
e
m
ph
a
s
iz
e
s
de
te
c
ti
ng
a
w
id
e
va
r
ie
ty
of
na
tu
r
a
l
obj
e
c
ts
,
a
voi
di
ng
li
m
it
a
ti
ons
to
s
pe
c
if
ic
c
a
te
gor
ie
s
li
ke
f
a
c
e
s
,
tr
e
e
s
,
or
ve
hi
c
le
s
. D
e
s
pi
te
th
e
r
a
nge
of
pot
e
nt
ia
l
obj
e
c
ts
,
r
e
s
e
a
r
c
h
e
f
f
or
ts
ha
ve
la
r
ge
ly
f
oc
us
e
d
on
hi
ghl
y
s
tr
uc
tu
r
e
d
obj
e
c
ts
(
e
.g.,
f
a
c
e
s
,
a
ir
pl
a
ne
s
)
a
nd
a
r
ti
c
ul
a
te
d
obj
e
c
ts
s
u
c
h
a
s
a
ni
m
a
l
s
.
O
bj
e
c
t
de
te
c
ti
on
s
uppor
ts
va
r
io
us
a
ppl
ic
a
ti
ons
,
in
c
l
udi
ng
f
a
c
ia
l
r
e
c
ogni
ti
on,
a
ut
onomous
dr
iv
in
g,
a
nd be
ha
vi
or
a
na
ly
s
is
[
40]
.
I
n
la
r
ge
-
s
c
a
le
s
ur
ve
il
la
nc
e
s
ys
te
m
s
,
a
c
c
ur
a
te
obj
e
c
t
tr
a
c
ki
ng
r
e
li
e
s
on
e
f
f
e
c
ti
ve
m
ot
io
n
e
s
ti
m
a
ti
on
a
nd
c
om
pe
ns
a
ti
on
te
c
hni
que
s
,
a
s
not
e
d
by
[
41]
.
T
he
s
tu
dy
pr
opos
e
d
a
h
a
r
dw
a
r
e
a
r
c
hi
te
c
tu
r
e
in
c
or
por
a
ti
ng
r
e
a
l
-
ti
m
e
m
ot
io
n
de
te
c
ti
on,
e
s
ti
m
a
ti
on,
a
nd
c
om
pe
ns
a
ti
on,
ut
il
iz
in
g
a
K
ogge
-
S
to
ne
a
dde
r
to
e
nha
nc
e
ope
r
a
ti
ona
l
s
pe
e
d. A
lt
hough the
m
e
th
od pr
oj
e
c
te
d a
4.21%
f
a
ls
e
de
te
c
ti
on r
a
te
, e
xpe
r
im
e
nt
a
l
r
e
s
ul
ts
i
ndi
c
a
te
d
a
n
11.91%
r
a
te
.
A
ddi
ti
o
na
ll
y
,
Z
he
ng
e
t
al
.
[
42]
pr
opo
s
e
d
a
c
o
s
t
-
e
f
f
e
c
ti
v
e
,
i
nt
e
gr
a
t
e
d
r
o
bot
ic
s
y
s
t
e
m
u
s
i
ng
C
a
r
te
s
ia
n
a
nd
a
r
ti
c
u
la
t
e
d
c
onf
i
gur
a
ti
on
s
f
o
r
ob
je
c
t
de
t
e
c
ti
on
i
n
a
gr
ic
ul
t
ur
a
l
e
nvi
r
on
m
e
nt
s
.
H
o
w
e
ve
r
,
th
e
d
e
s
i
gn
f
a
c
e
s
c
h
a
l
le
ng
e
s
d
ue
t
o
l
i
m
it
e
d
a
c
c
ur
a
c
y,
n
e
c
e
s
s
it
a
ti
ng
hu
m
a
n
c
o
ll
a
bor
a
ti
on
t
o
a
c
h
ie
ve
op
ti
m
a
l
p
e
r
f
or
m
a
nc
e
.
T
a
bl
e
2
pr
ovi
de
s
a
s
um
m
a
r
y
of
th
e
a
dva
nt
a
g
e
s
,
di
s
a
dva
nt
a
g
e
s
,
a
nd
e
xa
m
pl
e
s
of
s
in
gl
e
-
s
ta
ge
,
two
-
s
ta
ge
de
te
c
to
r
s
,
a
nd
t
r
a
ns
f
e
r
le
a
r
ni
ng
w
it
h
a
tt
e
nt
io
n
;
hi
ghl
i
ght
in
g
th
e
ir
r
e
s
pe
c
ti
ve
tr
a
de
-
of
f
s
in
s
pe
e
d,
a
c
c
ur
a
c
y,
a
nd
c
om
put
a
ti
ona
l
c
os
t,
pa
r
ti
c
ul
a
r
ly
in
in
door
na
vi
g
a
ti
on
ta
s
ks
.
W
hi
le
bot
h
a
ppr
oa
c
he
s
ha
ve
th
e
ir
r
e
s
pe
c
ti
ve
dr
a
w
ba
c
k
s
,
two
-
s
ta
ge
de
te
c
to
r
s
ty
pi
c
a
ll
y
of
f
e
r
s
upe
r
io
r
a
c
c
ur
a
c
y.
O
n
th
e
ot
he
r
ha
nd,
s
in
gl
e
-
s
ta
ge
de
te
c
to
r
s
a
r
e
ge
ne
r
a
ll
y
f
a
s
te
r
,
a
s
th
e
y
a
voi
d
th
e
c
om
pl
e
xi
ty
o
f
m
ul
ti
pl
e
s
ta
ge
s
.
T
he
im
pr
ove
d
a
c
c
ur
a
c
y
of
two
-
s
ta
ge
de
te
c
to
r
s
c
a
n be
a
tt
r
ib
ut
e
d t
o t
he
i
nc
lu
s
io
n of
r
e
gi
on pr
opos
a
l
ne
twor
ks
(
R
P
N
)
or
R
oI
pooli
ng.
F
ig
ur
e
1. D
e
e
p
le
a
r
ni
ng vs
. t
r
a
di
ti
ona
l
c
om
put
e
r
vi
s
io
n a
ppr
oa
c
he
s
[
37]
T
a
bl
e
2
. C
om
pa
r
is
on b
e
twe
e
n
s
in
gl
e
-
s
ta
ge
a
nd t
w
o
-
s
t
a
ge
de
te
c
t
or
[
43]
, [
44
]
T
ype
H
ow
i
t
w
or
ks
A
dva
nt
a
ge
s
D
i
s
a
dva
nt
a
ge
s
E
xa
m
pl
e
s
S
i
ngl
e
-
s
t
a
ge
de
t
e
c
t
or
A
s
i
ngl
e
-
l
a
ye
r
f
e
e
d
-
f
or
w
a
r
d
ne
t
w
or
k
t
ha
t
pe
r
f
or
m
s
obj
e
c
t
c
l
a
s
s
i
f
i
c
a
t
i
on
a
nd
r
e
gr
e
s
s
i
on t
o t
he
boundi
ng boxe
s
.
S
i
m
pl
e
r
a
nd f
a
s
t
e
r
f
or
de
t
e
c
t
i
on.
M
a
y ha
ve
r
e
duc
e
d
c
om
put
a
t
i
ona
l
a
c
c
ur
a
c
y.
Y
O
L
O
,
Y
O
L
O
v3,
S
S
D
,
R
e
t
i
na
N
e
t
T
w
o
-
s
t
a
ge
de
t
e
c
t
or
U
s
e
s
t
w
o
ne
t
w
or
ks
.
T
he
f
i
r
s
t
ge
ne
r
a
t
e
s
a
s
pa
r
s
e
R
oI
,
f
ol
l
ow
e
d
by
c
l
a
s
s
i
f
i
c
a
t
i
on
a
nd r
e
gr
e
s
s
i
on.
O
f
f
e
r
s
i
m
pr
ove
d
a
c
c
ur
a
c
y t
hr
ough t
he
us
e
of
R
oI
pool
i
ng.
I
nc
r
e
a
s
e
d
c
om
put
a
t
i
ona
l
t
i
m
e
due
t
o m
ul
t
i
pl
e
s
t
a
ge
s
.
R
-
C
N
N
,
C
a
s
c
a
de
R
-
C
N
N
,
F
a
s
t
e
r
R
-
C
N
N
T
r
a
ns
f
e
r
l
e
a
r
ni
ng w
i
t
h
a
t
t
e
nt
i
on
U
t
i
l
i
z
e
s
a
pr
e
-
t
r
a
i
ne
d
m
ode
l
,
f
i
ne
-
t
une
d
on
a
s
pe
c
i
f
i
c
t
a
s
k,
c
om
bi
ne
d
w
i
t
h
a
t
t
e
nt
i
on
m
e
c
ha
ni
s
m
s
t
o
hi
ghl
i
ght
i
m
por
t
a
nt
r
e
gi
ons
of
a
n i
m
a
ge
.
R
e
duc
e
s
t
r
a
i
ni
ng
t
i
m
e
a
nd r
e
qui
r
e
s
f
e
w
e
r
da
t
a
;
a
t
t
e
nt
i
on
e
nha
nc
e
s
f
oc
u
s
on
ke
y i
m
a
ge
a
r
e
a
s
f
or
i
m
pr
ove
d a
c
c
ur
a
c
y.
M
a
y s
t
i
l
l
r
e
qui
r
e
s
ubs
t
a
nt
i
a
l
c
om
put
a
t
i
ona
l
r
e
s
our
c
e
s
;
c
om
pl
e
x
a
r
c
hi
t
e
c
t
ur
e
.
D
e
t
e
c
t
i
on
t
r
a
ns
f
or
m
e
r
(
D
E
T
R
)
,
E
f
f
i
c
i
e
nt
D
e
t
3.1. Ch
al
le
n
ge
s
I
ndo
or
e
n
vi
r
o
nm
e
nt
s
pr
e
s
e
nt
uni
qu
e
c
h
a
l
le
ng
e
s
f
or
o
bj
e
c
t
d
e
t
e
c
ti
on
du
e
t
o t
he
f
o
ll
o
w
i
ng f
a
c
to
r
s
[
4
5]
:
−
O
c
c
lu
s
io
ns
:
obj
e
c
ts
m
a
y be
pa
r
ti
a
ll
y hi
dde
n be
hi
nd othe
r
obj
e
c
t
s
, m
a
ki
ng de
te
c
ti
on dif
f
ic
ul
t.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
5
,
O
c
to
be
r
2025
:
3520
-
3527
3524
−
V
a
r
yi
ng
li
ght
in
g
c
ondi
ti
ons
:
I
ndoor
li
ght
in
g
c
a
n
c
ha
nge
dr
a
m
a
ti
c
a
ll
y
ba
s
e
d
on
ti
m
e
of
da
y,
a
r
ti
f
ic
ia
l
li
ght
s
our
c
e
s
, a
nd s
h
a
dow
in
g e
f
f
e
c
ts
.
−
D
yna
m
ic
obj
e
c
ts
:
obj
e
c
ts
in
m
ot
io
n,
s
uc
h
a
s
pe
opl
e
or
r
obot
s
,
c
r
e
a
te
a
ddi
ti
ona
l
c
ha
ll
e
nge
s
f
or
de
te
c
ti
on
a
lg
or
it
hm
s
, pa
r
ti
c
ul
a
r
ly
i
n e
nvi
r
onm
e
nt
s
w
he
r
e
r
obot
s
m
us
t
a
voi
d c
ol
li
s
io
ns
.
−
R
e
a
l
-
ti
m
e
pr
oc
e
s
s
in
g:
d
e
pl
oyi
ng
a
dva
nc
e
d
m
ode
ls
on
m
obi
le
r
obot
s
r
e
qui
r
e
s
opt
im
iz
a
ti
on
f
or
li
m
it
e
d
c
om
put
a
ti
ona
l
r
e
s
our
c
e
s
w
it
hout
s
a
c
r
if
ic
in
g a
c
c
ur
a
c
y.
I
n
s
um
m
a
r
y,
w
h
il
e
d
e
e
p
le
a
r
n
in
g
a
ppr
oa
c
h
e
s
s
u
c
h
a
s
C
N
N
-
b
a
s
e
d
a
r
c
hi
te
c
t
ur
e
s
(
e
.g
.,
Y
O
L
O
,
R
C
N
N
)
ha
ve
r
e
vol
ut
io
ni
z
e
d
obj
e
c
t
de
te
c
ti
on,
th
e
y
c
om
e
w
it
h
th
e
dr
a
w
ba
c
k
of
s
ig
ni
f
ic
a
nt
c
om
put
a
ti
ona
l
c
om
pl
e
xi
ty
.
T
hi
s
de
m
a
nd
f
or
hi
gh
pr
oc
e
s
s
in
g
pow
e
r
m
a
ke
s
th
e
ir
de
pl
oym
e
nt
on
e
m
be
dde
d
s
ys
t
e
m
s
,
of
te
n
u
s
e
d
in
in
door
A
M
R
s
,
im
pr
a
c
ti
c
a
l
due
to
li
m
it
e
d
ha
r
dw
a
r
e
r
e
s
our
c
e
s
.
A
s
a
r
e
s
ul
t,
a
m
or
e
f
e
a
s
ib
le
s
ol
ut
io
n
i
s
th
e
in
te
gr
a
ti
on
of
c
la
s
s
ic
a
l
m
e
th
ods
,
s
u
c
h
a
s
f
e
a
tu
r
e
e
xt
r
a
c
ti
on,
w
it
h
de
e
p
le
a
r
ni
ng
te
c
hni
que
s
.
T
hi
s
c
om
bi
na
ti
on
e
na
bl
e
s
th
e
us
e
of
obj
e
c
t
de
te
c
ti
on
m
ode
ls
on
r
e
s
our
c
e
-
c
ons
tr
a
in
e
d
e
m
be
dde
d
s
ys
te
m
s
,
pr
ovi
di
ng
a
ba
la
nc
e
be
twe
e
n
pe
r
f
or
m
a
nc
e
a
nd
e
f
f
ic
ie
nc
y
w
hi
le
ove
r
c
om
in
g
h
a
r
dw
a
r
e
li
m
it
a
ti
ons
.
A
ddr
e
s
s
in
g
th
e
s
e
c
h
a
ll
e
nge
s
r
e
qui
r
e
s
m
ode
ls
c
a
pa
bl
e
of
hi
gh ge
ne
r
a
li
z
a
ti
on, a
s
w
e
ll
a
s
t
he
us
e
of
s
e
ns
or
f
us
io
n t
e
c
hni
que
s
(
e
.g., c
om
bi
ni
ng c
a
m
e
r
a
da
ta
w
it
h L
I
D
A
R
or
de
pt
h s
e
ns
or
s
)
t
o i
m
pr
ove
de
te
c
ti
on r
e
li
a
bi
li
ty
[
46]
.
3.2.
P
e
r
f
or
m
an
c
e
o
f
ob
j
e
c
t
d
e
t
e
c
t
io
n
m
od
e
ls
O
bj
e
c
t
de
te
c
ti
on
m
ode
ls
a
r
e
c
om
m
onl
y
e
va
lu
a
te
d
b
a
s
e
d
on
pr
e
c
is
io
n,
r
e
c
a
ll
,
a
nd
ot
he
r
m
e
tr
ic
s
,
w
it
h
F
a
s
te
r
R
-
C
N
N
a
nd
S
S
D
a
m
ong
th
e
m
o
s
t
a
c
c
ur
a
t
e
f
or
in
door
a
ppl
ic
a
ti
ons
.
H
ow
e
ve
r
,
th
e
r
e
a
r
e
in
h
e
r
e
nt
tr
a
de
-
of
f
s
be
twe
e
n
a
c
c
ur
a
c
y
a
nd
in
f
e
r
e
nc
e
ti
m
e
,
w
hi
c
h
a
r
e
e
s
pe
c
ia
ll
y
c
r
it
ic
a
l
f
or
r
e
a
l
-
ti
m
e
A
M
R
s
.
M
ode
ls
li
ke
Y
O
L
O
,
a
lt
hough
s
li
ght
ly
le
s
s
a
c
c
ur
a
te
,
of
te
n
s
tr
ik
e
th
e
b
e
s
t
ba
la
nc
e
f
or
r
e
a
l
-
ti
m
e
a
ppl
ic
a
ti
ons
,
m
a
ki
ng
th
e
m
w
e
ll
-
s
ui
te
d f
or
i
ndoor
e
nvi
r
onm
e
nt
s
w
he
r
e
qui
c
k de
c
is
io
n
s
a
r
e
ne
c
e
s
s
a
r
y
[
47]
.
I
n
a
ddi
ti
on
to
pr
e
c
is
io
n
a
nd
r
e
c
a
ll
,
ot
he
r
im
por
ta
nt
pe
r
f
o
r
m
a
nc
e
m
e
tr
ic
s
in
c
lu
de
m
e
a
n
a
ve
r
a
ge
pr
e
c
is
io
n
(
m
A
P
)
,
w
hi
c
h
a
s
s
e
s
s
e
s
a
c
c
ur
a
c
y
a
c
r
os
s
va
r
io
us
c
la
s
s
e
s
,
a
nd
in
te
r
s
e
c
ti
on
ov
e
r
uni
on
(
I
oU
)
,
w
hi
c
h
m
e
a
s
ur
e
s
th
e
ove
r
la
p
b
e
twe
e
n
pr
e
di
c
t
e
d
a
nd
gr
ound
tr
ut
h
boun
di
ng
boxe
s
.
F
ur
th
e
r
m
or
e
,
th
e
F
1
s
c
or
e
,
w
hi
c
h
c
om
bi
ne
s
pr
e
c
i
s
io
n
a
nd
r
e
c
a
ll
,
s
e
r
ve
s
a
s
a
ba
la
n
c
e
d
in
di
c
a
to
r
of
th
e
s
y
s
te
m
'
s
ove
r
a
ll
pe
r
f
or
m
a
nc
e
.
T
he
s
e
m
e
tr
ic
s
pr
ovi
de
a
c
om
pr
e
he
ns
iv
e
a
s
s
e
s
s
m
e
nt
of
th
e
a
c
c
ur
a
c
y,
r
obus
tn
e
s
s
,
a
nd
r
e
li
a
bi
li
ty
of
obj
e
c
t
de
te
c
ti
on
s
ys
te
m
s
i
n
A
M
R
s
[
48]
.
M
or
e
ove
r
,
s
c
a
la
bi
li
ty
a
nd
a
da
pt
a
bi
li
ty
a
r
e
im
por
ta
nt
c
ons
i
de
r
a
ti
ons
.
W
hi
le
S
S
D
a
nd
Y
O
L
O
e
f
f
ic
ie
nt
ly
de
te
c
t
obj
e
c
t
s
in
di
ve
r
s
e
e
nvi
r
onm
e
nt
s
,
e
n
s
ur
in
g
hi
gh
pe
r
f
or
m
a
nc
e
a
c
r
os
s
di
f
f
e
r
e
nt
li
ght
in
g
c
ondi
ti
ons
,
va
r
yi
ng
obj
e
c
t
or
ie
nt
a
ti
ons
,
a
nd
pot
e
nt
ia
l
oc
c
lu
s
io
ns
r
e
m
a
in
s
a
c
ha
ll
e
ng
e
.
A
dva
nc
e
d
te
c
hni
que
s
s
uc
h
a
s
c
ont
e
xt
ua
l
r
e
a
s
oni
ng
a
nd
m
ul
ti
-
s
c
a
le
f
e
a
tu
r
e
e
xt
r
a
c
ti
on
c
a
n
im
pr
ove
de
te
c
ti
on
a
c
c
ur
a
c
y,
pa
r
ti
c
ul
a
r
ly
in
c
om
pl
e
x i
ndoor
s
e
tt
in
gs
[
38]
.
L
a
s
tl
y,
th
e
im
pa
c
t
of
ha
r
dw
a
r
e
li
m
it
a
ti
ons
on
pe
r
f
or
m
a
nc
e
m
us
t
be
a
c
knowle
dge
d.
R
e
a
l
-
ti
m
e
de
te
c
ti
on
s
y
s
te
m
s
m
u
s
t
ba
la
n
c
e
be
tw
e
e
n
li
ght
w
e
ig
ht
m
od
e
ls
f
or
de
pl
oym
e
nt
on
e
m
be
dd
e
d
de
vi
c
e
s
a
nd
he
a
vi
e
r
,
m
or
e
a
c
c
ur
a
te
m
od
e
ls
f
or
s
e
r
ve
r
-
ba
s
e
d
pr
oc
e
s
s
in
g
.
T
h
is
ba
la
nc
e
is
e
s
pe
c
ia
ll
y
im
por
ta
nt
in
r
e
s
our
c
e
-
c
ons
tr
a
in
e
d e
nvi
r
onm
e
nt
s
w
he
r
e
i
nf
e
r
e
nc
e
s
pe
e
d i
s
c
r
uc
ia
l
f
or
de
c
is
io
n
-
m
a
ki
ng.
3.3.
F
u
t
u
r
e
d
ir
e
c
t
io
n
s
T
he
f
ut
ur
e
of
obj
e
c
t
de
te
c
ti
on i
n i
ndoor
A
M
R
s
m
a
y i
nvol
ve
[
49]
:
−
F
us
io
n
of
s
e
ns
or
s
:
c
om
bi
ni
ng
da
ta
f
r
om
c
a
m
e
r
a
s
,
L
i
D
A
R
,
a
nd
de
pt
h
s
e
ns
or
s
c
a
n
pr
ovi
de
r
ic
he
r
c
ont
e
xt
ua
l
in
f
or
m
a
ti
on, i
m
pr
ovi
ng de
te
c
ti
on r
obus
tn
e
s
s
.
−
S
e
lf
-
s
upe
r
vi
s
e
d
le
a
r
ni
ng
:
a
ppr
oa
c
hi
ng
th
e
is
s
ue
of
li
m
it
e
d
la
be
le
d
da
ta
s
e
t
s
th
r
ough
s
e
lf
-
s
upe
r
vi
s
e
d
or
s
e
m
i
-
s
upe
r
vi
s
e
d l
e
a
r
ni
ng t
e
c
hni
qu
e
s
c
oul
d e
nh
a
nc
e
m
ode
l
tr
a
in
in
g.
−
I
nt
e
r
pr
e
ta
bi
li
ty
:
a
s
r
obot
s
ope
r
a
te
c
lo
s
e
to
hum
a
ns
,
e
nha
nc
in
g
th
e
in
te
r
pr
e
ta
bi
li
ty
of
m
a
c
hi
ne
le
a
r
ni
n
g
m
ode
ls
be
c
om
e
s
ne
c
e
s
s
a
r
y f
or
t
r
us
t
in
de
c
is
io
n
-
m
a
ki
ng pr
oc
e
s
s
e
s
.
4.
C
O
N
C
L
U
S
I
O
N
T
hi
s
pa
p
e
r
ha
s
pr
ovi
de
d a
c
om
pr
e
he
ns
iv
e
r
e
vi
e
w
of
de
e
p
le
a
r
ni
ng
-
ba
s
e
d
obj
e
c
t
de
t
e
c
ti
on
te
c
hni
que
s
f
or
in
door
m
obi
le
r
obot
na
vi
ga
ti
on.
I
t
e
xa
m
in
e
d
th
e
tr
a
ns
it
io
n
f
r
om
tr
a
di
ti
ona
l
c
om
put
e
r
vi
s
io
n
m
e
th
ods
to
s
ta
te
-
of
-
th
e
-
a
r
t
de
e
p
le
a
r
ni
ng
m
ode
ls
,
w
it
h
a
pa
r
ti
c
ul
a
r
e
m
ph
a
s
is
on
th
e
Y
O
L
O
f
a
m
il
y,
tr
a
ns
f
or
m
e
r
-
ba
s
e
d
a
r
c
hi
te
c
tu
r
e
s
, a
nd mul
ti
-
s
e
ns
or
f
us
io
n s
tr
a
te
gi
e
s
. W
hi
le
t
he
s
e
a
p
pr
oa
c
he
s
ha
ve
l
e
d t
o nota
bl
e
a
dva
nc
e
m
e
nt
s
i
n
de
te
c
ti
on
a
c
c
ur
a
c
y,
r
e
a
l
-
ti
m
e
pe
r
f
or
m
a
nc
e
,
a
nd
de
pl
oym
e
nt
f
e
a
s
ib
il
it
y,
s
e
ve
r
a
l
li
m
it
a
ti
ons
r
e
m
a
in
,
pa
r
ti
c
ul
a
r
ly
in
ha
ndl
in
g
dyna
m
ic
in
door
e
nvi
r
onm
e
nt
s
,
li
m
it
e
d
a
nnot
a
te
d
da
ta
s
e
ts
,
a
nd
c
om
put
a
ti
ona
l
c
ons
tr
a
in
ts
on
e
m
be
dde
d
pl
a
tf
or
m
s
.
I
n
our
f
ut
ur
e
w
o
r
k,
w
e
a
i
m
to
pr
io
r
it
iz
e
th
e
de
ve
lo
pm
e
nt
of
li
ght
w
e
ig
ht
ye
t
hi
gh
-
pe
r
f
or
m
a
nc
e
obj
e
c
t
de
te
c
ti
on
m
ode
ls
s
ui
ta
bl
e
f
or
r
e
s
our
c
e
-
c
ons
tr
a
in
e
d
in
door
e
nvi
r
onm
e
nt
s
.
W
e
w
il
l
a
ls
o
e
xpl
or
e
s
e
lf
-
s
upe
r
vi
s
e
d
l
e
a
r
ni
ng
te
c
hni
que
s
to
r
e
duc
e
de
pe
nd
e
nc
e
on
a
nnot
a
te
d
d
a
ta
s
e
t
s
a
nd
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
O
bj
e
c
t
de
te
c
ti
on f
or
i
ndoor
m
obi
le
r
obot
:
de
e
p l
e
a
r
ni
ng appr
oa
c
he
s
r
e
v
ie
w
(
H
in
d M
e
s
s
bah
)
3525
in
ve
s
ti
ga
te
a
dva
nc
e
d
m
ul
ti
m
oda
l
s
e
ns
or
f
us
io
n
to
im
pr
ove
pe
r
c
e
pt
ua
l
r
obus
tn
e
s
s
.
F
ur
th
e
r
m
or
e
,
w
e
pl
a
n
to
de
s
ig
n
s
im
pl
if
ie
d
a
nd
ge
ne
r
a
li
z
a
bl
e
f
r
a
m
e
w
or
ks
th
a
t
e
n
a
bl
e
r
e
li
a
bl
e
de
pl
oym
e
nt
in
dyna
m
ic
r
e
a
l
-
w
or
ld
in
door
s
e
tt
in
gs
, t
he
r
e
by br
id
gi
ng t
he
ga
p be
twe
e
n t
he
or
e
ti
c
a
l
a
d
va
nc
e
m
e
nt
s
a
nd pr
a
c
ti
c
a
l
a
ppl
ic
a
ti
ons
.
F
U
N
D
I
N
G
I
N
F
O
R
M
A
T
I
O
N
T
he
a
ut
hor
s
s
t
a
te
no f
undi
ng i
s
i
nvol
ve
d.
A
U
T
H
O
R
C
O
N
T
R
I
B
U
T
I
O
N
S
S
T
A
T
E
M
E
N
T
T
hi
s
jo
ur
na
l
us
e
s
th
e
C
ont
r
ib
ut
or
R
ol
e
s
T
a
xonomy
(
C
R
e
d
iT
)
to
r
e
c
ogni
z
e
in
di
vi
dua
l
a
ut
hor
c
ont
r
ib
ut
io
ns
, r
e
duc
e
a
ut
hor
s
hi
p di
s
put
e
s
,
a
nd f
a
c
il
it
a
te
c
ol
la
bo
r
a
ti
on.
N
am
e
o
f
A
u
t
h
or
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
H
in
d M
e
s
s
ba
h
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
M
oha
m
e
d E
m
ha
r
r
a
f
✓
✓
✓
✓
✓
✓
✓
✓
M
oha
m
m
e
d S
a
be
r
✓
✓
✓
✓
✓
C
:
C
onc
e
pt
ua
l
i
z
a
t
i
on
M
:
M
e
t
hodol
ogy
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
da
t
i
on
Fo
:
Fo
r
m
a
l
a
na
l
ys
i
s
I
:
I
nve
s
t
i
ga
t
i
on
R
:
R
e
s
our
c
e
s
D
:
D
a
t
a
C
ur
a
t
i
on
O
:
W
r
i
t
i
ng
-
O
r
i
gi
na
l
D
r
a
f
t
E
:
W
r
i
t
i
ng
-
R
e
vi
e
w
&
E
di
t
i
ng
Vi
:
Vi
s
ua
l
i
z
a
t
i
on
Su
:
Su
pe
r
vi
s
i
on
P
:
P
r
oj
e
c
t
a
dm
i
ni
s
t
r
a
t
i
on
Fu
:
Fu
ndi
ng a
c
qui
s
i
t
i
on
C
O
N
F
L
I
C
T
O
F
I
N
T
E
R
E
S
T
S
T
A
T
E
M
E
N
T
T
he
a
ut
hor
s
s
t
a
te
no c
onf
li
c
t
of
i
nt
e
r
e
s
t.
D
A
T
A
A
V
A
I
L
A
B
I
L
I
T
Y
D
a
ta
a
va
il
a
bi
li
ty
doe
s
not
a
ppl
y t
o t
hi
s
p
a
pe
r
a
s
no ne
w
da
ta
w
e
r
e
c
r
e
a
te
d or
a
na
ly
z
e
d i
n t
hi
s
s
tu
dy.
R
E
F
E
R
E
N
C
E
S
[
1]
A
.
B
.
R
a
s
hi
d
a
nd
M
.
A
.
K
.
K
a
us
i
k,
“
A
I
r
e
vol
ut
i
oni
z
i
ng
i
ndus
t
r
i
e
s
w
or
l
dw
i
de
:
a
c
om
pr
e
he
ns
i
ve
ove
r
vi
e
w
of
i
t
s
di
ve
r
s
e
a
ppl
i
c
a
t
i
ons
,”
H
y
b
r
i
d A
dv
anc
e
s
, vol
. 7, D
e
c
. 2024, doi
:
10.1016/
j
.hyba
dv.2024.100277.
[
2]
A
.
L
oga
na
t
ha
n
a
nd
N
.
S
.
A
hm
a
d,
“
A
s
ys
t
e
m
a
t
i
c
r
e
vi
e
w
on
r
e
c
e
nt
a
dva
nc
e
s
i
n a
ut
onom
ous
m
obi
l
e
r
obot
na
vi
ga
t
i
on,”
E
ngi
ne
e
r
i
ng
Sc
i
e
nc
e
and T
e
c
hnol
ogy
, an I
nt
e
r
nat
i
onal
J
ou
r
nal
, vol
. 40, A
pr
. 2023, doi
:
10.1016/
j
.j
e
s
t
c
h.2023.101343.
[
3]
J
.
F
u,
L
.
Z
ong,
Y
.
L
i
,
K
.
L
i
,
B
.
Y
a
ng,
a
nd
X
.
L
i
u,
“
M
ode
l
a
da
pt
i
on
obj
e
c
t
de
t
e
c
t
i
on
s
ys
t
e
m
f
or
r
obot
,”
i
n
2020
39t
h
C
hi
ne
s
e
C
ont
r
ol
C
onf
e
r
e
nc
e
(
C
C
C
)
, J
ul
. 2020, pp. 3659
–
3664, doi
:
10.23919/
C
C
C
50068.2020.9189674.
[
4]
S
.
A
bdul
-
K
ha
l
i
l
,
S
.
A
bdul
-
R
a
hm
a
n,
S
.
M
ut
a
l
i
b,
S
.
I
.
K
a
m
a
r
ud
i
n,
a
nd
S
.
S
.
K
a
m
a
r
uddi
n,
“
A
r
e
vi
e
w
on
obj
e
c
t
de
t
e
c
t
i
on
f
or
a
ut
onom
ous
m
obi
l
e
r
obot
,”
I
A
E
S I
nt
e
r
nat
i
onal
J
ou
r
nal
of
A
r
t
i
f
i
c
i
al
I
nt
e
l
l
i
ge
nc
e
(
I
J
-
A
I
)
, vol
. 12, no. 3, pp. 1033
–
1043,
S
e
p. 2023
,
doi
:
10.11591/
i
j
a
i
.v12.i
3.pp1033
-
1043.
[
5]
Y
.
L
i
u,
S
.
W
a
ng,
Y
.
X
i
e
,
T
.
X
i
ong,
a
nd
M
.
W
u,
“
A
r
e
vi
e
w
of
s
e
ns
i
ng
t
e
c
hnol
ogi
e
s
f
or
i
ndoor
a
ut
onom
ous
m
obi
l
e
r
obot
s
,”
Se
ns
or
s
, vol
. 24, no. 4, F
e
b. 2024, doi
:
10.3390/
s
24041222.
[
6]
K
.
O
ka
r
m
a
,
“
A
ppl
i
c
a
t
i
on
s
of
c
om
put
e
r
vi
s
i
on
i
n
a
ut
om
a
t
i
on
a
nd
r
obot
i
c
s
,”
A
ppl
i
e
d
Sc
i
e
nc
e
s
,
vol
.
10,
no.
19,
S
e
p.
2020,
doi
:
10.3390/
a
pp10196783.
[
7]
J
.
K
a
ng,
S
.
T
a
r
i
q,
H
.
O
h,
a
nd
S
.
S
.
W
oo,
“
A
s
ur
ve
y
of
de
e
p
l
e
a
r
ni
ng
-
ba
s
e
d
o
bj
e
c
t
de
t
e
c
t
i
on
m
e
t
hods
a
nd
da
t
a
s
e
t
s
f
or
ove
r
he
a
d
i
m
a
ge
r
y,”
I
E
E
E
A
c
c
e
s
s
, vol
. 10, pp. 20118
–
20134, 2022, doi
:
10.1109/
A
C
C
E
S
S
.2022.3149052.
[
8]
M
.
C
a
r
r
a
nz
a
-
G
a
r
c
í
a
,
J
.
T
or
r
e
s
-
M
a
t
e
o,
P
.
L
a
r
a
-
B
e
ní
t
e
z
,
a
nd
J
.
G
a
r
c
í
a
-
G
ut
i
é
r
r
e
z
,
“
O
n
t
he
pe
r
f
or
m
a
nc
e
of
one
-
s
t
a
ge
a
nd
t
w
o
-
s
t
a
ge
obj
e
c
t
de
t
e
c
t
or
s
i
n a
ut
onom
ous
ve
hi
c
l
e
s
us
i
ng c
a
m
e
r
a
da
t
a
,”
R
e
m
ot
e
Se
n
s
i
ng
, v
ol
. 13, no. 1, D
e
c
. 2020, doi
:
10.3390/
r
s
13010089.
[
9]
H
.
W
a
ng,
C
.
L
i
u,
Y
.
C
a
i
,
L
.
C
he
n,
a
nd
Y
.
L
i
,
“
Y
O
L
O
v8
-
Q
S
D
:
a
n
i
m
pr
ove
d
s
m
a
l
l
obj
e
c
t
de
t
e
c
t
i
on
a
l
gor
i
t
hm
f
or
a
ut
onom
ous
ve
hi
c
l
e
s
ba
s
e
d
on
Y
O
L
O
v8,
”
I
E
E
E
T
r
ans
ac
t
i
ons
on
I
ns
t
r
um
e
nt
at
i
on
and
M
e
as
ur
e
m
e
nt
,
vol
.
73,
pp.
1
–
16,
2024,
doi
:
10.1109/
T
I
M
.2024.3379090.
[
10]
S
.
B
e
he
r
a
,
B
.
A
n
a
nd,
a
nd
P
.
R
a
j
a
l
a
ks
hm
i
,
“
Y
ol
oV
8
ba
s
e
d
nove
l
a
ppr
oa
c
h
f
or
obj
e
c
t
de
t
e
c
t
i
on
on
L
i
D
A
R
poi
nt
c
l
oud,”
i
n
I
E
E
E
V
e
hi
c
ul
ar
T
e
c
hnol
ogy
C
onf
e
r
e
nc
e
, J
un. 2024, pp. 1
–
5, doi
:
10.1109/
V
T
C
2024
-
S
pr
i
ng62846.2024.10683316.
[
11]
S
. B
a
s
ha
a
nd G
.
R
a
m
,
“
R
e
a
l
-
t
i
m
e
obj
e
c
t
de
t
e
c
t
i
on i
n l
ow
-
l
i
ght
e
nvi
r
onm
e
nt
s
us
i
ng Y
O
L
O
v8:
a
c
a
s
e
s
t
udy w
i
t
h a
c
us
t
om
d
a
t
a
s
e
t
,
”
I
nt
e
r
nat
i
onal
J
our
nal
of
E
ngi
ne
e
r
i
ng R
e
s
e
ar
c
h and
, vol
. 13, 2024, doi
:
10.17577/
I
J
E
R
T
V
13I
S
100050.
[
12]
A
.
A
l
ot
a
i
bi
,
H
.
A
l
a
t
a
w
i
,
A
.
B
i
nnouh,
L
.
D
uw
a
yr
i
a
t
,
T
.
A
l
hm
i
e
da
t
,
a
nd
O
.
M
. A
l
i
a
,
“
D
e
e
p
l
e
a
r
ni
ng
-
ba
s
e
d
vi
s
i
on
s
ys
t
e
m
s
f
or
r
obot
s
e
m
a
nt
i
c
na
vi
ga
t
i
on:
a
n e
xpe
r
i
m
e
nt
a
l
s
t
udy,”
T
e
c
hnol
ogi
e
s
, vol
. 12, no. 9, 2024
, doi
:
10.3390/
t
e
c
hnol
ogi
e
s
12090157.
[
13]
E
.
H
e
i
ke
l
a
nd
L
.
E
s
pi
nos
a
-
L
e
a
l
,
“
I
ndoor
s
c
e
ne
r
e
c
ogni
t
i
on
vi
a
obj
e
c
t
de
t
e
c
t
i
o
n
a
nd
T
F
-
I
D
F
,”
J
our
nal
of
I
m
agi
ng
,
vol
.
8,
no.
8,
2022, doi
:
10.3390/
j
i
m
a
gi
ng8080209.
[
14]
M
.
A
f
i
f
,
R
.
A
ya
c
hi
,
Y
.
S
a
i
d,
a
nd
M
.
A
t
r
i
,
“
A
n
e
va
l
ua
t
i
on
o
f
E
f
f
i
c
i
e
nt
D
e
t
f
o
r
obj
e
c
t
de
t
e
c
t
i
on
us
e
d
f
or
i
ndoor
r
obot
s
a
s
s
i
s
t
a
n
ce
na
vi
ga
t
i
on,”
J
our
nal
of
R
e
al
-
T
i
m
e
I
m
age
P
r
oc
e
s
s
i
ng
, vol
. 19, no. 3, pp. 651
–
66
1, 2022, doi
:
10.1007/
s
11554
-
022
-
01212
-
4.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
5
,
O
c
to
be
r
2025
:
3520
-
3527
3526
[
15]
C
.
T
a
ng,
B
.
A
bba
t
e
m
a
t
t
e
o,
J
.
H
u,
R
.
C
h
a
ndr
a
,
R
.
M
a
r
t
í
n
-
M
a
r
t
í
n,
a
nd
P
.
S
t
one
,
“
D
e
e
p
r
e
i
nf
or
c
e
m
e
nt
l
e
a
r
ni
ng
f
or
r
obot
i
c
s
:
a
s
ur
ve
y
of
r
e
a
l
-
w
or
l
d
s
uc
c
e
s
s
e
s
,
”
A
nnual
R
e
v
i
e
w
of
C
ont
r
ol
,
R
obot
i
c
s
,
and
A
ut
onom
ous
Sy
s
t
e
m
s
,
vol
.
8,
no.
1,
pp.
153
–
188,
2025,
doi
:
10.1146/
a
nnur
e
v
-
c
ont
r
ol
-
030323
-
022510.
[
16]
Z
. X
i
ul
i
ng, W
. H
ui
j
ua
n, S
. Y
u, C
. G
a
ng, Z
. S
uhua
, a
nd Y
. Q
ua
nbo,
“
S
t
a
r
t
i
ng f
r
om
t
he
s
t
r
uc
t
ur
e
:
a
r
e
vi
e
w
of
s
m
a
l
l
obj
e
c
t
de
t
e
c
t
i
o
n
ba
s
e
d on de
e
p l
e
a
r
ni
ng,”
I
m
age
and V
i
s
i
on C
om
put
i
ng
, vol
. 146, 2024, doi
:
10.1016/
j
.i
m
a
vi
s
.2024.105054.
[
17]
Y
.
S
un,
Z
.
S
un,
a
nd
W
.
C
he
n,
“
T
he
e
vol
ut
i
on
of
obj
e
c
t
de
t
e
c
t
i
on
m
e
t
hods
,”
E
ngi
ne
e
r
i
ng
A
ppl
i
c
at
i
ons
of
A
r
t
i
f
i
c
i
al
I
nt
e
l
l
i
ge
nc
e
,
vol
. 133, J
ul
. 2024, doi
:
10.1016/
j
.e
nga
ppa
i
.2024.108458.
[
18]
L
.
W
a
ng,
L
.
Z
.
C
he
n,
B
.
P
e
ng,
a
nd
Y
.
T
.
L
i
n,
“
I
m
pr
ove
d
Y
O
L
O
v5
a
l
gor
i
t
hm
f
or
r
e
a
l
-
t
i
m
e
pr
e
di
c
t
i
on
of
f
i
s
h
yi
e
l
d
i
n
a
l
l
c
a
g
e
s
c
hool
s
,”
J
ou
r
nal
of
M
ar
i
ne
Sc
i
e
nc
e
and E
ngi
ne
e
r
i
ng
, vol
. 12, no. 2, 2024, doi
:
10.3390/
j
m
s
e
12020195.
[
19]
S
.
G
ui
,
S
.
S
ong,
R
.
Q
i
n,
a
nd
Y
.
T
a
ng,
“
R
e
m
ot
e
s
e
n
s
i
ng
obj
e
c
t
de
t
e
c
t
i
on
i
n
t
he
de
e
p
l
e
a
r
ni
ng
e
r
a
—
a
r
e
vi
e
w
,”
R
e
m
ot
e
Se
ns
i
ng
,
vol
. 16, no. 2, 2024, doi
:
10.3390/
r
s
16020327.
[
20]
M
.
A
.
H
a
m
e
e
d
a
nd
Z
.
A
.
K
ha
l
a
f
,
“
A
s
ur
ve
y
s
t
udy
i
n
obj
e
c
t
de
t
e
c
t
i
on:
a
c
om
pr
e
he
ns
i
ve
a
na
l
y
s
i
s
of
t
r
a
di
t
i
ona
l
a
nd
s
t
a
t
e
-
of
-
t
he
-
a
r
t
a
ppr
oa
c
he
s
,”
B
as
r
ah R
e
s
e
ar
c
he
s
S
c
i
e
nc
e
s
, vol
. 50, no. 1, 2024, doi
:
10.56714/
b
j
r
s
.50.1.5.
[
21]
R
. S
un
e
t
al
.
, “
S
ur
ve
y of
i
m
a
ge
e
dge
de
t
e
c
t
i
on,”
F
r
ont
i
e
r
s
i
n Si
gnal
P
r
oc
e
s
s
i
ng
, vol
. 2, 2022, doi
:
10.3389/
f
r
s
i
p.2022.826967.
[
22]
D
.
T
.
N
guye
n,
W
.
L
i
,
a
nd
P
.
O
gunbona
,
“
A
n
i
m
p
r
ove
d
t
e
m
pl
a
t
e
m
a
t
c
hi
ng
m
e
t
hod
f
or
obj
e
c
t
de
t
e
c
t
i
on,”
i
n
A
s
i
an
C
onf
e
r
e
nc
e
on
C
om
put
e
r
V
i
s
i
on
, 2009, pp. 193
–
202, doi
:
10.1007/
978
-
3
-
642
-
12297
-
2_19.
[
23]
N
.
M
a
na
ki
t
s
a
,
G
.
S
.
M
a
r
a
s
l
i
di
s
,
L
.
M
oy
s
i
s
,
a
nd
G
.
F
.
F
r
a
gul
i
s
,
“
A
r
e
vi
e
w
of
m
a
c
hi
ne
l
e
a
r
ni
ng
a
nd
de
e
p
l
e
a
r
ni
ng
f
or
obj
e
c
t
de
t
e
c
t
i
on,
s
e
m
a
nt
i
c
s
e
gm
e
nt
a
t
i
on,
a
nd
hum
a
n
a
c
t
i
on
r
e
c
ogni
t
i
on
i
n
m
a
c
hi
ne
a
nd
r
obot
i
c
vi
s
i
on,”
T
e
c
hnol
ogi
e
s
,
vol
.
12,
no.
2
,
2024, doi
:
10.3390/
t
e
c
hnol
ogi
e
s
12020015.
[
24]
Z
.
Z
ou,
K
.
C
he
n,
Z
.
S
hi
,
Y
.
G
uo,
a
nd
J
.
Y
e
,
“
O
bj
e
c
t
de
t
e
c
t
i
on
i
n
20
y
e
a
r
s
:
a
s
ur
ve
y,”
P
r
oc
e
e
di
ng
s
of
t
he
I
E
E
E
,
vol
.
111,
no.
3,
pp. 257
–
276, 2023, doi
:
10.1109/
J
P
R
O
C
.2023.3238524.
[
25]
P
.
K
.
G
os
w
a
m
i
a
nd
G
.
G
os
w
a
m
i
,
“
A
c
om
pr
e
he
n
s
i
ve
r
e
vi
e
w
on
r
e
a
l
t
i
m
e
obj
e
c
t
de
t
e
c
t
i
on
us
i
ng
d
e
e
p
l
e
a
r
ni
ng
m
ode
l
,”
i
n
P
r
oc
e
e
di
ngs
of
t
he
2022
11t
h I
nt
e
r
nat
i
onal
C
onf
e
r
e
n
c
e
on
Sy
s
t
e
m
M
ode
l
i
ng an
d A
dv
anc
e
m
e
nt
i
n
R
e
s
e
ar
c
h T
r
e
nds
, SM
A
R
T
2022
,
2022, pp. 1499
–
1502, doi
:
10.1109/
S
M
A
R
T
55829.2022.10046972.
[
26]
N
.
A
l
s
ha
r
a
bi
,
“
R
e
a
l
-
t
i
m
e
obj
e
c
t
de
t
e
c
t
i
on
ove
r
vi
e
w
:
a
dva
nc
e
m
e
nt
s
,
c
ha
l
l
e
nge
s
,
a
nd
a
ppl
i
c
a
t
i
ons
,”
J
our
nal
of
A
m
r
an
U
ni
v
e
r
s
i
t
y
,
vol
. 3,
pp. 267
-
278
,
2023, doi
:
10.59145/
j
a
us
t
.v3i
6.73.
[
27]
S
.
M
.
A
a
m
i
r
,
H
.
Ma
,
M
.
A
.
A
.
K
ha
n
,
a
nd
M
.
A
a
qi
b,
“
R
e
a
l
t
i
m
e
obj
e
c
t
de
t
e
c
t
i
on
i
n
oc
c
l
ude
d
e
nvi
r
onm
e
nt
w
i
t
h
ba
c
kgr
ound
c
l
ut
t
e
r
i
ng
e
f
f
e
c
t
s
us
i
ng
de
e
p
l
e
a
r
ni
ng,”
T
he
7t
h
I
nt
e
r
nat
i
onal
W
or
k
s
hop
on
A
d
v
anc
e
d
C
om
put
at
i
onal
I
nt
e
l
l
i
ge
nc
e
and
I
nt
e
l
l
i
ge
nt
I
nf
or
m
at
i
c
s
(
I
W
A
C
I
I
I
2021)
,
2021,
pp. 1
-
6
.
[
28]
G
.
L
a
va
nya
a
nd
S
.
D
.
P
a
nde
,
“
E
nha
nc
i
ng
r
e
a
l
-
t
i
m
e
obj
e
c
t
de
t
e
c
t
i
on
w
i
t
h
Y
O
L
O
a
l
gor
i
t
hm
,”
E
A
I
E
ndor
s
e
d
T
r
ans
ac
t
i
ons
on
I
nt
e
r
ne
t
of
T
hi
ngs
, vol
. 10, 2024, doi
:
10.4108/
e
e
t
i
ot
.4541.
[
29]
W
.
L
i
u
e
t
al
.
,
“
S
S
D
:
s
i
ngl
e
s
hot
m
ul
t
i
box
de
t
e
c
t
or
,”
i
n
E
ur
ope
an
c
onf
e
r
e
nc
e
on
c
om
put
e
r
v
i
s
i
on
,
2016,
pp.
21
–
37,
doi
:
10.1007/
978
-
3
-
319
-
46448
-
0_2.
[
30]
L
.
D
u,
R
.
Z
ha
ng,
a
nd
X
.
W
a
ng,
“
O
ve
r
vi
e
w
of
t
w
o
-
s
t
a
ge
obj
e
c
t
de
t
e
c
t
i
on
a
l
gor
i
t
hm
s
,”
J
our
nal
of
P
hy
s
i
c
s
:
C
onf
e
r
e
nc
e
Se
r
i
e
s
,
vol
. 1544, no. 1, M
a
y 2020, doi
:
10.1088/
1742
-
6596/
1544/
1/
012033.
[
31]
K
.
M
a
ha
j
a
n
a
nd
P
.
M
a
n
e
,
“
O
bj
e
c
t
de
t
e
c
t
i
on
us
i
ng
c
onvol
ut
i
ona
l
ne
ur
a
l
ne
t
w
or
ks
(
C
N
N
s
)
:
a
s
t
udy,
i
m
pl
e
m
e
nt
a
t
i
on
a
nd
c
ha
l
l
e
nge
s
,”
I
nt
e
r
nat
i
onal
J
ou
r
nal
of
A
r
c
hi
t
e
c
t
ur
e
, E
ngi
ne
e
r
i
ng, and C
ons
t
r
uc
t
i
on
, vol
. 12, no. 3, pp. 308
–
317, 2023.
[
32]
R
.
G
i
r
s
hi
c
k,
J
.
D
ona
hue
,
T
.
D
a
r
r
e
l
l
,
a
nd
J
.
M
a
l
i
k,
“
R
e
gi
on
-
ba
s
e
d
c
onvol
ut
i
ona
l
ne
t
w
or
ks
f
or
a
c
c
ur
a
t
e
obj
e
c
t
de
t
e
c
t
i
on
a
nd
s
e
gm
e
nt
a
t
i
on,”
I
E
E
E
T
r
ans
ac
t
i
ons
on
P
at
t
e
r
n
A
nal
y
s
i
s
and
M
ac
hi
ne
I
nt
e
l
l
i
ge
nc
e
,
vol
.
38,
no.
1,
pp.
142
–
158,
2016,
doi
:
10.1109/
T
P
A
M
I
.2015.2437384.
[
33]
J
.
Y
a
ng,
D
.
C
he
n,
a
nd
H
.
S
hi
,
“
A
obj
e
c
t
de
t
e
c
t
i
on
m
e
t
hod
ba
s
e
d
on
a
t
t
e
nt
i
on
m
e
c
ha
ni
s
m
a
nd
r
e
i
nf
or
c
e
m
e
nt
l
e
a
r
ni
ng,”
i
n
2022
18t
h
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
on
C
om
put
at
i
onal
I
nt
e
l
l
i
ge
nc
e
and
Se
c
ur
i
t
y
,
C
I
S
2022
,
2022,
pp.
229
–
233,
doi
:
10.1109/
C
I
S
58238.2022.00055.
[
34]
W
.
L
i
,
K
.
L
i
u,
L
.
Z
ha
ng,
a
nd
F
.
C
he
ng,
“
O
bj
e
c
t
de
t
e
c
t
i
on
ba
s
e
d
on
a
n
a
da
pt
i
v
e
a
t
t
e
nt
i
on
m
e
c
ha
ni
s
m
,”
Sc
i
e
nt
i
f
i
c
R
e
por
t
s
,
vol
.
10,
no. 1, 2020, doi
:
10.1038/
s
41598
-
020
-
67529
-
x.
[
35]
A
.
D
e
vi
e
t
al
.
,
“
T
r
a
ns
f
e
r
l
e
a
r
ni
ng
f
or
obj
e
c
t
de
t
e
c
t
i
on
i
n
r
e
m
ot
e
s
e
ns
i
ng
i
m
a
ge
s
w
i
t
h
Y
O
L
O
,”
J
our
nal
of
E
l
e
c
t
r
i
c
al
Sy
s
t
e
m
s
,
vol
. 20, no. 3, pp. 980
–
989, A
pr
. 2024, doi
:
10.52783/
j
e
s
.1412.
[
36]
N
.
O
’
M
a
hony
e
t
al
.
,
“
D
e
e
p
l
e
a
r
ni
ng
vs
.
t
r
a
di
t
i
ona
l
c
om
put
e
r
vi
s
i
on,”
A
dv
anc
e
s
i
n
I
nt
e
l
l
i
ge
nt
S
y
s
t
e
m
s
and
C
om
put
i
ng
,
vol
.
943,
pp. 128
–
144, 2020, doi
:
10.1007/
978
-
3
-
030
-
17795
-
9_10.
[
37]
O
pe
nC
V
U
ni
ve
r
s
i
t
y,
“
W
hy
t
r
a
di
t
i
ona
l
c
om
put
e
r
vi
s
i
on
t
hr
i
ve
s
a
l
ongs
i
de
de
e
p
l
e
a
r
ni
ng:
a
c
ount
e
r
poi
nt
t
o
d
e
e
p
l
e
a
r
ni
ng
dom
i
na
nc
e
,”
L
i
nk
e
dI
n,
2024.
A
c
c
e
s
s
e
d:
S
e
p.
10,
2024.
[
O
nl
i
ne
]
.
A
va
i
l
a
bl
e
:
ht
t
ps
:
/
/
w
w
w
.l
i
nke
di
n.c
om
/
pul
s
e
/
w
hy
-
t
r
a
di
t
i
ona
l
-
c
om
put
e
r
-
vi
s
i
on
-
t
hr
i
ve
s
-
a
l
ongs
i
de
-
de
e
p
-
w
v0i
f
[
38]
D
.
S
.
B
a
c
e
a
a
nd
F
.
O
ni
g
a
,
“
S
i
ngl
e
s
t
a
ge
a
r
c
hi
t
e
c
t
ur
e
f
or
i
m
pr
ove
d
a
c
c
ur
a
c
y
r
e
a
l
-
t
i
m
e
obj
e
c
t
de
t
e
c
t
i
on
on
m
obi
l
e
de
vi
c
e
s
,”
I
m
ag
e
and V
i
s
i
on C
om
put
i
ng
, vol
. 130, F
e
b. 2023, doi
:
10.1016/
j
.i
m
a
vi
s
.2022.104613.
[
39]
L
.
A
z
i
z
,
M
.
S
.
B
.
H
.
S
a
l
a
m
,
U
.
U
.
S
he
i
kh,
a
nd
S
.
A
yub,
“
E
xpl
or
i
ng
de
e
p
l
e
a
r
ni
ng
-
ba
s
e
d
a
r
c
hi
t
e
c
t
ur
e
,
s
t
r
a
t
e
gi
e
s
,
a
ppl
i
c
a
t
i
on
s
a
n
d
c
ur
r
e
nt
t
r
e
nds
i
n
ge
ne
r
i
c
obj
e
c
t
de
t
e
c
t
i
on:
a
c
om
pr
e
he
ns
i
ve
r
e
vi
e
w
,”
I
E
E
E
A
c
c
e
s
s
,
vol
.
8,
pp.
170461
–
170495,
2020,
do
i
:
10.1109/
A
C
C
E
S
S
.2020.3021508.
[
40]
B
.
R
.
K
i
r
a
n
e
t
al
.
,
“
R
e
a
l
-
t
i
m
e
dyna
m
i
c
obj
e
c
t
de
t
e
c
t
i
on
f
or
a
ut
onom
ous
dr
i
vi
ng
us
i
ng
pr
i
or
3D
-
m
a
ps
,”
i
n
P
r
o
c
e
e
di
ngs
of
t
h
e
E
ur
ope
an c
onf
e
r
e
nc
e
on c
om
put
e
r
v
i
s
i
on (
E
C
C
V
)
w
or
k
s
hop
s
, 2019, pp. 567
–
5
82, doi
:
10.1007/
978
-
3
-
030
-
11021
-
5_35.
[
41]
Z
. C
he
n, R
. Z
ha
o, X
. G
uo, J
. X
i
e
, a
nd
X
. H
a
n, “
M
ovi
ng obj
e
c
t
de
t
e
c
t
i
on i
n f
r
e
e
l
y m
ovi
ng c
a
m
e
r
a
vi
a
gl
oba
l
m
ot
i
on c
om
pe
ns
a
t
i
on
a
nd l
oc
a
l
s
pa
t
i
a
l
i
nf
or
m
a
t
i
on f
us
i
on,”
Se
ns
or
s
, vol
. 24, no. 9, A
pr
. 2024, doi
:
10.3390/
s
24092859.
[
42]
Z
.
Z
he
ng,
Y
.
H
u,
X
.
L
i
,
a
nd
Y
.
H
ua
ng,
“
A
ut
onom
ous
na
vi
ga
t
i
on
m
e
t
hod
of
j
uj
ube
c
a
t
c
h
-
a
nd
-
s
ha
ke
ha
r
ve
s
t
i
ng
r
obot
ba
s
e
d
on
c
onvol
ut
i
ona
l
ne
ur
a
l
ne
t
w
or
ks
,”
C
om
put
e
r
s
and
E
l
e
c
t
r
oni
c
s
i
n
A
gr
i
c
ul
t
ur
e
,
vol
.
215,
no.
1,
D
e
c
.
2023,
doi
:
10.1016/
j
.c
om
pa
g.2023.108469.
[
43]
L
.
S
hi
ne
a
nd
C
.
V
.
J
i
j
i
,
“
C
om
pa
r
a
t
i
ve
a
na
l
ys
i
s
of
t
w
o
s
t
a
g
e
a
nd
s
i
ngl
e
s
t
a
g
e
de
t
e
c
t
or
s
f
or
a
nom
a
l
y
de
t
e
c
t
i
on,”
i
n
2021
12t
h
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
on
C
o
m
put
i
ng
C
om
m
uni
c
at
i
on
and
N
e
t
w
o
r
k
i
ng
T
e
c
hnol
ogi
e
s
,
I
C
C
C
N
T
2021
,
2021,
pp.
1
–
6,
doi
:
10.1109/
I
C
C
C
N
T
51525.2021.9580079.
[
44]
R
.
R
a
j
a
nd
A
.
K
o
s
,
“
S
t
udy
of
hum
a
n
–
r
obot
i
nt
e
r
a
c
t
i
ons
f
or
a
s
s
i
s
t
i
ve
r
obot
s
us
i
ng
m
a
c
hi
n
e
l
e
a
r
ni
ng
a
nd
s
e
ns
or
f
us
i
o
n
t
e
c
hnol
ogi
e
s
,”
E
l
e
c
t
r
oni
c
s
, vol
. 13, no. 16, A
ug. 2024, doi
:
10.3390/
e
l
e
c
t
r
oni
c
s
13163285.
[
45]
M
.
A
f
i
f
,
R
.
A
ya
c
hi
,
a
nd
M
.
A
t
r
i
,
“
I
ndoor
obj
e
c
t
s
d
e
t
e
c
t
i
on
s
y
s
t
e
m
i
m
pl
e
m
e
nt
a
t
i
on
us
i
ng
m
ul
t
i
-
gr
a
phi
c
pr
oc
e
s
s
i
ng
uni
t
s
,”
C
l
us
t
e
r
C
om
put
i
ng
, vol
. 25, no. 1, pp. 469
–
483, 2022, doi
:
10.1007/
s
10586
-
021
-
03419
-
9.
[
46]
H
.
L
i
u,
C
.
W
u,
a
nd
H
.
W
a
ng,
“
R
e
a
l
t
i
m
e
obj
e
c
t
de
t
e
c
t
i
on
us
i
ng
L
i
D
A
R
a
nd
c
a
m
e
r
a
f
us
i
on
f
or
a
ut
onom
ous
dr
i
vi
ng,”
Sc
i
e
nt
i
f
i
c
R
e
por
t
s
, vol
. 13, no. 1, 2023, doi
:
10.1038/
s
41598
-
023
-
35170
-
z.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
O
bj
e
c
t
de
te
c
ti
on f
or
i
ndoor
m
obi
le
r
obot
:
de
e
p l
e
a
r
ni
ng appr
oa
c
he
s
r
e
v
ie
w
(
H
in
d M
e
s
s
bah
)
3527
[
47]
T
.
M
a
he
ndr
a
ka
r
e
t
al
.
,
“
P
e
r
f
or
m
a
nc
e
s
t
udy
of
Y
O
L
O
v5
a
nd
f
a
s
t
e
r
R
-
C
N
N
f
or
a
ut
onom
ous
na
vi
ga
t
i
on
a
r
ound
non
-
c
oope
r
a
t
i
ve
t
a
r
ge
t
s
,”
i
n
I
E
E
E
A
e
r
os
pac
e
C
onf
e
r
e
nc
e
P
r
oc
e
e
di
ngs
, 2022, pp. 1
–
12, doi
:
10.1
109/
A
E
R
O
53065.2022.9843537.
[
48]
J
.
R
e
z
a
i
e
,
“
M
a
s
t
e
r
i
ng
obj
e
c
t
de
t
e
c
t
i
on
m
e
t
r
i
c
s
:
f
r
om
I
oU
t
o
m
A
P
,”
M
e
di
um
,
2
024.
A
c
c
e
s
s
e
d
:
S
e
p.
12,
2024
.
[
O
nl
i
ne
]
.
A
va
i
l
a
bl
e
:
ht
t
ps
:
/
/
m
e
di
um
.c
om
/
@
r
e
z
a
i
e
.j
/
m
a
s
t
e
r
i
ng
-
obj
e
c
t
-
de
t
e
c
t
i
on
-
m
e
t
r
i
c
s
-
f
r
om
-
i
ou
-
to
-
m
a
p
-
12b42e
f
78d4b
[
49]
Y
.
D
a
i
,
D
.
K
i
m
,
a
nd
K
.
L
e
e
,
“
A
n
a
dva
nc
e
d
a
ppr
oa
c
h
t
o
ob
j
e
c
t
de
t
e
c
t
i
on
a
nd
t
r
a
c
ki
ng
i
n
r
obo
t
i
c
s
a
nd
a
ut
onom
ous
ve
hi
c
l
e
s
us
i
ng
Y
O
L
O
ov8 a
nd
L
i
D
A
R
da
t
a
f
us
i
on,
”
E
l
e
c
t
r
oni
c
s
, vol
. 13, no. 12, J
un. 2024, doi
:
10.3390/
e
l
e
c
t
r
oni
c
s
13122250.
B
I
O
G
R
A
P
H
I
E
S
O
F
A
U
T
H
O
R
S
Hind
Messbah
received
her
master’s
degree
in
Big
Data
and
her
bachelor’s
degree
in
Computer
Enginee
ring
f
rom
the
International
University
of
Rabat,
Morocco,
in
the
year
2018.
Presently,
she
holds
the
position
of
technica
l
leader
with
in
the
consulting
industry.
With
a
career
spanning
over
five
years,
she
has
garnered
extensi
ve
expertise
as
a
data
engineer,
boasting
a
track
record
of
successful
project
implementation
s
across
diverse
sectors,
encompass
ing
telecomm
unicatio
ns,
insurance,
retail,
and
banking.
Her
research
interests
include
big
data,
artificial
intelligence,
robotics,
and
the
internet
of
t
hings
(IoT
)
.
She
can
be
contacted
at email
:
h.mesbahi@
ump.ac.ma.
Mohamed
Emharraf
is
a
Profes
sor
of
Robotics
at
National
Sch
ool
of
Applied
Scienc
es,
Mohame
d
First
Univer
sity,
Oujda,
Morocc
o.
He
rece
iv
ed
his
Ph.D.
in
2017
from
CEDOC
-
EMPO.
His
research
interests
include
indoor
robot
cont
rol,
smart
agricultural,
computer
engineerin
g,
human
-
computer
-
interactio
n,
and
artificia
l
intelligence
.
He
has
published 29 papers in
peer
-
reviewed journal
s and
conference proceed
ings
.
He has also
served
as
a
reviewer
for
several
scientific
journals
and
as
a
program
commi
ttee
member.
He
can
be
contacted
at email
:
m.emharra
f@
ump.ac.ma
.
Mohammed
Saber
is
currently
a
Full
Professor
(PES)
in
the
Department
of
Electronics,
Computer
Science,
and
Telecommunications
at
th
e
Nati
onal
School
of
Applied
Scienc
es
at
Mohamme
d
First
Univer
sity,
Oujda,
Morocc
o
(2013)
.
He
received
a
Ph
.
D
.
in
Computer
Science
at
the
Faculty
of
Science
s,
Oujda,
Moro
cco,
in
July
2012,
an
enginee
r
degree
in
Network
and
Telecommunication
at
the
National
School
of
Applied
Sciences,
in
July
2004,
and
a
Licence
degree
in
Electronics
at
the
Faculty
of
Sci
ences,
in
July
2002,
all
from
Mohammed
First
University,
Oujda.
He
is
currently
the
Director
of
the
Smart
Information,
Communi
cation
and
Technologies
Laboratory
(SmartI
CT
Lab).
His
interests
include
network
security
(intrusion
detection
systems,
eval
uation
of
s
ecurity
components,
and
security
IoT),
AI,
robotics,
and
embedded
systems
.
He
can
b
e
contacted
at
email:
m.saber@ump.ac.ma
.
Evaluation Warning : The document was created with Spire.PDF for Python.