►
Description
A quick video to show the new models you can build with the updated ML.NET Model Builder.
Announcement blog post - https://devblogs.microsoft.com/dotnet/ml-net-model-builder-updates/
Model Builder introduction video - https://www.youtube.com/watch?v=2M75v9z9R2U
ML.NET Playlist - https://www.youtube.com/watch?v=8gVhJKszzzI&list=PLl_upHIj19Zy3o09oICOutbNfXj332czx
Contact:
Twitter: https://twitter.com/JWood/
Blog: https://jonwood.co/
Gear used (affiliate links):
Mic - https://amzn.to/2YEXtxI
Mouse - https://amzn.to/2ZtASoQ
A
Hey
everyone
recently,
the
m/l
net
model
builder
tool
has
been
updated
and
it
includes
a
couple
of
cool
new
things.
So,
in
this
video
I
wanted
to
go
over
the
two
new
types
of
models
you
can
build
for
you,
image,
classification
and
recommendation.
First
of
all,
if
you
haven't
heard
of
the
internet
model
builder,
it
is
a
visual
studio,
UI
tool
that
you
can
use
to
build
models
for
you
and
it
generates
the
code
and
it
get
started
on
a
page
forward.
A
You
can
click
the
download
button
and
it
takes
you
to
the
business
to
do
marketplace
where
you
can
download
it.
If
you
already
have
it
installed
and
you
want
to
update
it,
just
go
to
tools,
extensions
and
updates'
and
then
and
the
updates
part
here.
You
will
see
it
listed.
Here's
an
update,
alright!
So
let's
get
started.
A
I
have
just
a
regular
console
project
here
and
let's
go
into
the
model
builder
here,
so
I
do
add,
and
machine
learning
now
tell
you
what
scenario
to
pick
out
so
first,
let's
look
at
the
recommendation
model
data
can
build
and
I'll
give
it
a
file
and
go
knocks.
Let
what
column
it
needs
to
predict
and
that's
going
to
be.
A
The
ratings
:
select
the
user
column
and
in
the
item
column
it's
gonna,
be
user,
IDs
and
then
book
IDs
for
the
suit
book
recommendation
model
and
I
don't
want
to
build
in
a
slight
train
and
I.
Give
it
the
number
of
seconds
that
I
wanted
to
train,
and
here
I'll
just
give
it
30
seconds
and
as
a
change,
you
can
see
the
output
in
the
output
pane
here.
So
we'll
just
let
this
run
and
be
back
in
a
bit.
A
A
You
can
click
predict
here
and
it
gives
us
a
predicted
rating
for
the
book
for
that
user
and
does
give
us
top
recommendations
for
that
user
as
well,
and
we
click
code
to
generate
our
code
and
then
add
the
projects
to
our
solution
and
then
over.
Here
we
have
our
model
builder
projects
here
and
then
we
can
just
go
mark.
The
console
app
is
our
startup
project
here
and
we
can
run
it
real,
quick
and
the
code
that
generates
is
much
the
same,
that
it
did
before
then
good.
A
A
So
it
gives
you
the
best
model
that
I
can
within
the
training
town
that
you
give
it
sorry,
let's
go
look
at
the
other
model
that
we
can
build
in
the
new
model
builder
and
a
new
console
project
here
and
I'm
gonna,
add
another
machine
learning
project
and
we
have
image
classification
as
one
of
our
choices.
So
we
split
the
folder
of
our
data
and
it
has
to
be
in
this
folder
structure
where
the
folder
defines
what
label
it
is
and
then,
within
that
folder
or
all
the
images
for
that
label.
A
Alright,
so
I
get
this
one
data
and
we
have
a
folder
of
red
and
white
one,
and
so
each
it's
gonna
be
a
photo
of
that
specific
one
within
it.
Yeah
just
one
the
model
to
determine
what
type
of
one
is
in
the
photo,
so
I'll
click
train,
and
here
we
have
the
option
of
training
locally.
The
counselor
training
is
your,
but
I'll
just
keep
keep
it
locally
here
and
notice.
A
We
don't
specify
the
number
of
seconds
to
train
and
that's
because
that
it
automatically
sets
the
training
time
based
on
the
size
of
our
data
set.
So
I
click
start
training
and
then
well
that
train
for
as
long
as
it
takes
and
we'll
come
back
when
it's
done
all
right.
So
our
training
got
complete
here
and
so
well.
I'll
click
on
buy,
evaluate
you
see,
have
an
overall
accuracy
of
a
hundred
percent,
which
is
it
was
nice,
and
here
we
can
try
our
model
again.
A
We
can
select
an
image
and
I
have
a
test
folder
here.
So
let's
look
at
a
one.
I
said:
94%
read
so
that's
nice.
We
can
try
another
image.
Just
try
white
one
and
99%
white,
sir,
so
far
so
good
on
our
image
classification
model
here
and
we
can
add
the
code
to
our
solution
and
we
can
go
over
and
when
this
as
startup
funds
it
and
run
it
real,
quick.
A
It's
going
to
use
one
of
those
images
as
a
simple
image
and
it's
a
natural
label,
red
predicted
label
red
and
it
gives
the
two
scores
so
95
percent
red
and
about
five
percent
white
all
right.
So
that's
just
a
bit
of
a
taste
of
what's
new
in
an
inland
inet
model
builder
feel
free
to
play
with
it.
I
hope
you
enjoyed
the
video
and
I'll
see
you
all
next
time,
Thanks.