►
From YouTube: Unimatrix Technical Board call 2020-07-01
A
A
B
B
B
B
And
maybe
we
could
start
with
discussing.
We
don't
have
many
points
today,
but
if
we
are
waiting
for
videotape
to
join
for
the
tv
vote,
we
could
really
start
talking
about
the
meeting
recordings
frederick.
You
suggested
that
we
should
move
to
youtube
for
the
recording
so
that
we,
I
guess
it's
more
easy
to
view
the
recordings.
If
you
have
them
on
youtube.
E
No,
they
have
to
start
their
own
service.
I
think
it's
baidu
or
whatever
it's
called,
but
I
just
know
that
it's
not
available.
B
D
B
Yeah,
if
I
get
a
suggestion
into
what
video
streaming
is
used
in
china
and
if
that
is
accessible
from
here,
I
could
upload
to
both
youtube
and
libra
service
in
china.
G
F
It
maybe
microsoft
stream.
I
will
check
if
it
there's
a
free
tier
on
microsoft
stream,
it's
possible
to
use
that.
E
E
Yeah
I
mean
it
should
be
a
relatively
easy
task
to
identify
a
placeholder,
that's
outside
the
jit.
Basically,
whether
it's
it's
you
know
could,
alternatively,
be
a
place
where
we
like
an
on
on
online
storage,
where
you
just
put
the
files
and
then
you
download
and
view,
even
though
it's
not
as
optimal
as
youtube,
but
it
would
work
as
long
as
it's
accessible
from
both
places
but
yeah.
Maybe
it's
an
action
item
for
next
meeting.
E
I'm
I'm
not
that
familiar
with
baidu,
except
I
don't
know,
of
course,
if
that
works
in
china,
and
I
assume
it
would
work
here
as
well
right,
I
never.
B
A
bit
yeah
and
while
we're
talking
about
meetings,
we
could
maybe
already
discuss
the
next
meeting.
I
know
there
is
vacation
time
coming
up
and
the
predict
you're
away
for
two
weeks.
D
Yep,
so
should
we
have
something
in
july
or
I
don't
know
when
people
have
vacation,
but
for
me
it's
okay
from.
D
B
D
E
I
Both
of
these
two
time
is
okay
with
us
in
front
of
mission.
Yeah.
B
And
manuel,
do
you
have
any
opinion
about
what
sorry
next
meeting
july,
22nd
or
29th.
H
D
D
Yeah,
so
this
was
brought
up
last
meeting,
so
the
voting
is
just
to
clarify.
This
is
the
voting
about
edge
api
for
uni
matrix
1.0.
D
So
this
is
not
to
say
that
we
will
what
we
can't
have
a
different
engine
in
future,
for
example,
moving
to
on
onyx
engine
or
something
like
that,
but
as
we
have
tried
to
explain
from
axis
side
a
couple
of
times
now
the
current
situation
of
all
these
different,
deep
learning,
chipsets
that
are
out
there
right
now,
it's
difficult
to
support
onyx
format
for
umbrella,
for
example.
D
D
So
I
think
we
would
like
to
make
another
boat
and
see
what
people
think
so
I'll
just
go
through
all
the
members,
so
from
hikvision
side
is
it
okay.
I
D
Okay,
genotech.
F
K
Agreed
a
question:
when
will
laura's
be
open
public
open
source
in
the
github.
D
I
I
can't
give
an
exact
date,
but
I
can
I
can
say
it.
Probably
it
will
happen
in
the
coming
weeks.
So
it's
not
far
away.
D
Yeah,
basically
from
our
management,
they
want
to
to
be
able
to
to
get
the
the
okay
on
investing
the
time
needed
to
for
making
the
the
release
public.
We
wanted
to
have
this
vote
to
go
through,
so
that
we,
because
there
is
a
some
resources
that
are
necessary
to
to
make
everything
their
documentation.
Everything
like
ready
for
for
to
be
public,
so
they
wanted
to
have
this
vote
done
before.
D
Yeah
so
yeah,
probably
before
end
of
august,
we
can
say
that.
D
And
what
about
from
videotech?
B
H
B
D
B
D
It's
accepted
and
the
other
vote
we
had
about
the
tensorflow
serving
api,
so
we
won't
take
up
that
vote
today,
will
give
some
more
time
for
companies
to
research
the
api
yeah.
D
Basically,
that's
all
I
had
for
this
day.
I
don't
know
if
you
have
anything
else,
you'd
like
to
bring
any
any
feedback
from
the
hackathon,
something
like
that.
D
J
Can
I
have
a
question
so
I'm
from
april
and
we
are
okay,
but
we
expect
that
our
when
release
we
expect
to
receive
some
document
to
show
the
advantage
and
some
disabilities,
maybe
some
limitation
points,
especially
for
the
optimizations
or
performance
compatibility
with
the
other
deep
learning
framework.
Yeah.
That's
already
my.
J
Question
or
yes,
questions
are
on
comments,
so
I
expect
to
have
the
when
you
release
the
narrate,
so
it
comes
out
with
some
comparisons
to
show
the
advantage
and
some
limitations
of
the
road.
D
Like
yeah,
like.
D
Yeah
yeah
there
are
so
many
different
inference
engines.
Yes,
the
I
guess
the
advantages
is
that
it's
it
can
support
it's
an
open
engine
that
can
support
the
the
hardware
is
that
we.
J
Yeah-
and
I
think
there
are
some
optimizations
phase
in
narrow-
that's
right,
so
maybe
every
members
want
to
know
the
how
much
it
can
optimize
something
like
that.
D
Yeah,
the
current
the
work
that's
going
into
lard
right
now,
except
for
supporting
a
new
like
umbrella.
D
It's
got
the
work
that
is
going
on
right
now
is
it's
about
the
conversions,
the
pre-processing
and
post-processing.
So
basically
you
don't
have
to
do
that
in
opencv,
for
example,
because
opencv
is
not
really
the
color
conversions
is
not
really
optimized
in
opencv
yeah,
okay,.
I
D
I
was
just
going
to
say
it
was
something
that
I
think
when
we
do
the
new
release
of
lord.
I
think
one
important
aspect
are
the
examples,
so
because
right
now
we
only
have
one
example
and
it's
using
the
not
using
the
larot
api.
Actually
it's
using
the
tensorflow
serving
api,
so
I
think
with
the
release.
There
should
be
a
couple
of
examples,
so
you
understand,
and
then
maybe
from
that
you
can
see
what
the
limitations
are
right
now
with
acceleration
and
so
on.
D
You
mean
hacking
hacking
on
an
example
yeah
I've
been
I've
been
meaning
to
do
my
own
example.
Actually,
because
right
now
we
only
have
a
python
example,
so
I
I
would
like
to
have
a
c
plus
plus
example,
using
opencv
and
power.
Okay,.
B
Wouldn't
that
be
a
good,
great
idea
to
have
a
some
sort
of
mini
hackathon,
and
so
people
could
participate
in
that
and
also
learned
how
to
collaborate
using
the
tools
that
you
used
to
hack
upon.
D
H
I'm
not
really
sure
if
I
have
the
sorry
for
the
background
noise.
There
is
a
lot
going
on.
I'm
not
really
sure
if
I
have
the
all
the
technical
expertise
related
to
the
example
to
host
something,
but
I
think
it
would
be
a
good
idea.
We
could
think
of
it.
Sometime.
B
Yeah
because
since
we
have
a
problem
we've
we
have
a
limited
experience
when
we're
doing
the
hackathon,
and
this
could
be
a
good
event
both
to
learn
more
about
narrow.
G
H
We
intend
to
invite
yon
in
one
of
our
next
sessions.
Could
it
be
more
useful
if
he
sort
of
walks
us
through
and
then
companies
will
have
questions
around
the
performance
of
larrow
as
opposed
to
other
influence
engine?
So
I
suppose
he
would
be
better
able
to
respond
to
the
queries
as
well
when
it
comes
down
to
performance
issues.
D
I
I
think
now,
since
we
have
now,
we
have
this
vote,
I
think
from
the
lara
team.
They
are
more
willing
to
participate,
so
I
mean,
for
example,
in
december
we
had
two
other
guys
that
was
not
present
in
the
and
the
hackathon
we
just
finished.
So
that's
was
anders
and
another
guy
don't
remember
his
name.
So
those
guys
are
really
good
on
all
the
details
about
what
you
can
do
and
cannot
do.
D
So
I
think
we
should
definitely
have
maybe
some
more
guys
from
lard
if
we,
if
we're
gonna,
have
such
an
event.
D
D
Yeah,
so
regarding
your
examples
from
ipro,
is
it
is
it
okay
to
because
now
they
are
uploaded
to
the
meetings
project?
I
guess
so.
D
Maybe
we
could
containerize
those
examples
and
put
them
under
the
other
ex
there's
one
more.
Other
example.
Is
that
something
that's
that.
J
D
D
H
H
Yeah,
I
think
it
would
make
more
sense
if
laura
gets
open
source
first
and
then
someone
from
the
lara
team
tags
along
in
in
the
any
hackathon.
D
Okay,
dora,
I'm
sorry,
I
interrupted
you,
you
were
gonna,
say
something.
I
D
Doesn't
a
tool,
a
tool
like
like
postman
validated.
D
D
A
I
Okay,
I
want
to,
after
the
validation-
and
I
will
put
it
on
the
github
and
maybe
colleagues
can
give
some
comments
on
the
specification
on
the
definition
of
this
apis
and
for
the
pdc
operation.
I
I
If
other
colleagues
have
some
some
use
cases
about
the
pdg
operate,
I
hope
they
can
share
with
me
and
we
do
discussion
and
specify
the
entity
of
api.
I
D
Okay,
so
yeah,
please
send
a
when
you
have
put
it
in
the
kit,
so
please
send
a
you,
can
send
it
to
the
list.
If
you
want
some
comments
or
you
can
maybe
best
ways
to
create
a
merch
request
and
then
people
can
comment
there.
I
C
Just
make
sure
they
say
they
they
seem
with
the
with
the
with
the
china
platform.
Is
that
you
you
want
you
want
to
put
you
want
to
find
a
place
to
put
the
audio
file,
or
do
you
want
somewhere
to
do
some
live
streaming.
B
B
J
B
Yeah
then,
I
have
to
thank
everybody
who
participated
and
see
you
on
the
next
meeting.
I
will
send
out
an
invitation
for
the
date
we
we
said
the
29th
of
july.