<s>
OpenVINO	B-Application
toolkit	O
(	O
Open	O
Visual	O
Inference	O
and	O
Neural	O
network	O
Optimization	O
)	O
is	O
a	O
free	O
toolkit	O
facilitating	O
the	O
optimization	O
of	O
a	O
deep	B-Algorithm
learning	I-Algorithm
model	O
from	O
a	O
framework	O
and	O
deployment	O
using	O
an	O
inference	O
engine	O
onto	O
Intel	O
hardware	O
.	O
</s>
<s>
The	O
toolkit	O
has	O
two	O
versions	O
:	O
OpenVINO	B-Application
toolkit	O
,	O
which	O
is	O
supported	O
by	O
open	O
source	O
community	O
and	O
Intel	O
Distribution	O
of	O
OpenVINO	B-Application
toolkit	O
,	O
which	O
is	O
supported	O
by	O
Intel	O
.	O
</s>
<s>
OpenVINO	B-Application
was	O
developed	O
by	O
Intel	O
.	O
</s>
<s>
The	O
toolkit	O
is	O
cross-platform	B-Operating_System
and	O
free	O
for	O
use	O
under	O
Apache	B-Application
License	I-Application
version	O
2.0	O
.	O
</s>
<s>
The	O
toolkit	O
enables	O
a	O
write-once	O
,	O
deploy-anywhere	O
approach	O
to	O
deep	B-Algorithm
learning	I-Algorithm
deployments	O
on	O
Intel	O
platforms	O
,	O
including	O
CPU	B-General_Concept
,	O
integrated	O
GPU	B-Architecture
,	O
Intel	O
Movidius	O
VPU	B-General_Concept
,	O
and	O
FPGAs	B-Architecture
.	O
</s>
<s>
The	O
toolkit	O
’s	O
is	O
a	O
cross-platform	B-Operating_System
tool	O
that	O
transforms	O
a	O
trained	O
model	O
from	O
the	O
original	O
framework	O
to	O
OpenVINO	B-Application
format	O
(	O
IR	O
)	O
and	O
optimizes	O
it	O
for	O
future	O
inference	O
on	O
supported	O
devices	O
.	O
</s>
<s>
As	O
a	O
result	O
,	O
Model	O
Optimizer	O
produces	O
two	O
files	O
:	O
*	O
.bin	O
and	O
*	O
.xml	B-Protocol
,	O
which	O
contain	O
weights	O
and	O
model	O
structures	O
respectively	O
.	O
</s>
<s>
The	O
toolkit	O
’s	O
is	O
a	O
C++	B-Language
library	O
for	O
inferring	O
input	O
on	O
devices	O
and	O
getting	O
results	O
.	O
</s>
<s>
OpenVINO	B-Application
has	O
different	O
sample	O
types	O
:	O
classification	O
,	O
object	O
detection	O
,	O
style	O
transfer	O
,	O
speech	O
recognition	O
,	O
etc	O
.	O
</s>
<s>
All	O
these	O
models	O
are	O
available	O
for	O
learning	O
purpose	O
or	O
for	O
development	O
deep	B-Algorithm
learning	I-Algorithm
software	O
.	O
</s>
<s>
Open	O
Model	O
Zoo	O
is	O
licensed	O
under	O
Apache	B-Application
License	I-Application
version	O
2.0	O
.	O
</s>
<s>
Along	O
with	O
the	O
primary	O
components	O
of	O
model	O
optimization	O
and	O
runtime	O
within	O
Intel®	O
Distribution	O
of	O
OpenVINO	B-Application
toolkit	O
,	O
the	O
toolkit	O
also	O
includes	O
a	O
user-friendly	O
web	O
browser	O
interface	O
called	O
the	O
to	O
aid	O
in	O
model	O
analysis	O
and	O
experimentation	O
;	O
a	O
tool	O
called	O
the	O
to	O
accelerate	O
inference	O
by	O
converting	O
models	O
into	O
low-precision	O
and	O
that	O
do	O
not	O
require	O
re-training	O
(	O
e.g.	O
,	O
post-training	O
quantization	O
)	O
;	O
and	O
,	O
additional	O
add-ons	O
,	O
such	O
as	O
the	O
to	O
aid	O
in	O
streaming	O
analytics	O
pipeline	O
interoperability	O
,	O
the	O
to	O
enable	O
scalability	O
via	O
a	O
serving	O
microservice	O
,	O
like	O
the	O
,	O
and	O
the	O
,	O
an	O
online	O
interactive	O
video	O
and	O
image	O
annotation	O
tool	O
.	O
</s>
<s>
OpenVINO	B-Application
has	O
two	O
webpages	O
:	O
one	O
for	O
another	O
for	O
.	O
</s>
<s>
OpenVINO	B-Application
is	O
written	O
in	O
C++	B-Language
and	O
Python	B-Language
.	O
</s>
<s>
OpenVINO	B-Application
runs	O
on	O
the	O
following	O
desktop	O
operation	O
systems	O
:	O
Windows	B-Application
,	O
Linux	B-Application
and	O
MacOS	B-Application
.	O
</s>
<s>
OpenVINO	B-Application
also	O
runs	O
on	O
.	O
</s>
