►
From YouTube: Slingshot Phase 2.3 - AetherImages
Description
Join us for the Slingshot 2.3 closing ceremony on Jun 9th 7pm PT / Jun 10 2am UTC. Click below to register for the event!
https://protocol.zoom.us/meeting/register/tJwlfuGrpz8vEteJ_jVPWPbp7aNmitFwWxv3
A
Data
set
presentation
for
ether
images
for
the
slingshot
phase,
2.3
competition.
We
try
to
keep
this
under
four
minutes.
So
what
is
open
images?
What
does
the
data
set?
Look
like?
How
have
we
prepared
it
for
the
storage
of
the
filecoin
network?
How
can
you
retrieve
it
and
what
are
the
post-processing
steps
to
yield
a
complete
data
set?
So,
first
of
all,
what
is
the
google
open
images
data
set?
A
Essentially,
it's
a
bunch
of
tagged
images,
split
into
training,
test
and
validation
folders
for
you
to
help
you
train
up
neural
networks
that
endeavor
to
observe
and
understand.
What's
going
on
in
in
the
real
world
short
version,
it
looks
kind
of
like
this,
where
cats,
dogs
and
all
sorts
of
other
things
in
the
real
world
that
are
tagged
for
you,
as
well
as
they've
got
a
bunch
of
them
with
masks
and
things
like
that
are
very
helpful
for
edge
detection
and
the
rest
when
you
download
it.
A
The
home
page
looks
like
this
and,
if
you
click
into
the
downloads,
it'll
look
like
this,
this
page
that
we're
on
and
if
you
get
into
the
downloads,
it
looks
like
this,
so
we
use
these
aws
links
to
download
the
data
sets
the
three
of
them
one
for
each
each
folder,
train,
validation
and
test
once
done
downloading
the
the
folder
looks
like
like
this
I'll,
give
you
a
quick
console
example
here.
This
is
our
source
data
folder.
Here
we've
got
all
these
three
things.
A
If
we
go
to
test
and
we
show
you
directory
listing
there,
you'll
see
it
comes
as
a
bunch
of
jpegs,
it's
really
bad
for
storing
on
the
network.
So
I'm
going
to
show
you
one
thing
here:
real
quick,
though
just
an
md5
of
one
of
these,
so
we
can
show
you
at
the
end
that
the
data
set
you've
downloaded
or
you've
been
able
to
retrieve
from
the
network
is
the
same.
A
We're
going
to
run
mb5
in
the
absolute
last
file
there
ending
in
bbea,
and
this
ends
in
cf205
you'll
see
that
again
at
the
end
of
this
presentation.
So
how
do
we
prepare
it?
We
basically
zipped
up
all
these
files
indicating
the
maximum
file
size
to
try
to
fit
into
that
32
gig
sector
for
filecoin
network.
A
Just
get
in
here
we'll
see
these
are
the
these.
Are
the
prepared
files
there,
so
we
stored
those
on
the
network
and
now,
if
you
want
to
retrieve
them,
what
do
you
do?
Well,
you
go
to
our
project
website,
which
is
here
alan
boyd.wixsite.com,
stroke,
ether,
boxy
and
you
scroll
down
to
project
number
four
open
images,
data
set
and
click.
The
open
images
deal
list
that'll
bring
you
to
our
air
table,
which
shows
you
everything
that
you
need
to
know
in
order
to
be
able
to
download
the
correct,
correct
files.
A
A
A
You
just
need
the
oh,
I'm
in
any
case
that
results
in
a
zip
that
you
end
up
then
unzip
or
combining
into
the
into
the
largest
one
and
then
unzipping
the
the
entire
thing.
That's
a
long
process
here,
but
let
me
show
you
what's
next,
let's
just
confirm
the
final
md5.
Here
we
go
to
the
folder
that
we
downloaded
that
into
and
extracted
it
to,
which
is
the
storage
that
I've
retrieved,
open
images.
A
Then
the
final
image
data
set
and
then
the
same
file
that
we
rendered
md5
on
before
and
that'll
show
you
here
again
cf-205.
So
there
you
go.
There's
our
four-minute
video!
If
you
have
any
questions
or
would
you
like
some
help,
downloading
stuff
feel
free
to
ping
me
on
slack,
it's
rob
mac
and
we
hope
you
enjoy
the
open
images
dataset.
Thank
you
very.