►
From YouTube: NuPIC Geospatial Tracking Tutorial
Description
This video describes the NuPIC Geospatial Tracking application and shows how it can be used to analyze GPS data.
https://github.com/numenta/nupic.geospatial
A
Hello
new
pic,
this
is
matt
taylor,
new
pic
open
source
community,
flag
bearer.
I
am
going
to
talk
today
about
the
numenta
geospatial
tracking
application,
and
this
is
a
new
pick
application
that
will
take
geospatial
information
plot
it
on
a
map
and
do
anomaly
indications
on
every
point
in
the
route
that's
passed
in
and
give
you
a
plot
on
a
map
of
the
anomaly
score
or
indication
for
each
point
in
the
routes.
A
So
I'm
looking
right
now
at
the
repository
it
is
on
github
under
new
pic
or
excuse
me
under
numenta,
slash
geospatial,
so
you
do
have
to
have
newpic
installed
to
do
anything
useful
with
this
application.
It
is
a
python
application,
so
there
are
some
dependencies
that
you
can
pretty
easily
install
with
pip
and
to
use
it
you
simply
run
the
server.
So
when
you
clone
this
repo
or
download
it,
it
comes
already
canned
with
some
sample
data.
So
I'm
going
to
show
you
that
first,
so
here
I
am
in
my
shell.
A
Under
this
repository,
I'm
just
going
to
run
python
server,
that's
going
to
start
a
local
web
server
on
localhost
and
it's
going
to
load
up
that
sample
data.
That's
already
been
processed
with
new
pick
and
it's
showing
you
the
results
after
nupic
has
done
anomaly:
detection
on
these
routes,
so
I'm
going
to
walk
through
this.
There
are
four
routes
in
the
sample
data
set.
The
first
one
is
a
drive
from
sunnyvale
up
to
redwood
city
and
it's
all
red,
because
it
is
very
anomalous.
So
keep
in
mind.
A
This
is
a
brand
new
new
pick
model.
It's
never
seen
any
data
of
any
kind
before
and
so
it
it
is
a
high
anomaly
score
for
every
point
along
the
route.
So
we've
caught
we've
done
this
route
several
times.
So
this
is
the
second
pass
on
the
route
and,
as
you
can
see,
there
are
still
there's
a
kind
of
gradient
from
green
to
red.
That
indicates
how
anomalous
each
point
in
the
route
is
or
how
anomalous
newbik
thinks
each
point
in
the
route
is
so
in
the
second
pass.
A
It
still
has
some
high
anomalies
in
the
beginning,
but
it
gets
better
and
then
let's
look
further
along
at
the
third
track.
Here
we
go,
which
is
almost
entirely
green,
so
it
is
used
to
this
track.
At
this
point,
this
sequence
of
points
and
then
the
last
one,
it's
almost
entirely
green,
but
there's
a
point
here
right
in
the
middle.
That
is
anomalous.
A
So
let's
take
a
quick
look
at
that
and
there's
a
reason
for
that:
that's
because
whoever
was
driving
this
car
and
tracking
their
gps
location
took
a
bit
of
a
detour
and
made
an
exit
on
embarcadero
road
did
a
u-turn
and
got
back
on
the
highway.
So,
as
you
can
see,
this
is
anomalous
behavior
based
on
previous
routes
and
new
pick
does
pick
it
up
as
an
anomaly
along
this
route.
A
A
So
there
is
another
portion
of
this
application,
and
it's
at
simulate
is
the
url
just
slash
simulate,
and
there
is
a
route
simulator
here
which
is
kind
of
neat.
So
you
can
you
can
make
your
own
routes
essentially,
so
I
can
simulate
some
route
between
point
a
and
point
b
and
then
tell
it.
I'm
gonna
do
four
trips
along
this
route.
If
I
click
the
build
button,
this
will
actually
pass
this
through
new
pic
encode.
A
All
these
locations
using
the
geospatial,
coordinate,
encoder
and
create
a
model
and
pass
all
the
data
into
new
pic
and
the
result
will
be
a
display
of
the
results
from
new
pick
on
how
anomalous
it
it
thinks
each
track
is
so
here
is
the
result
from
the
simulator.
A
This
is
the
entire
track.
It
seems
you
can
kind
of
peruse
through
here
with
these
these
buttons
or
this
slider
down
at
the
bottom
and
see.
There's
a
lot
of
points
along
this,
so.
A
There
we
go
see
parts
of
the
route
you
can
kind
of
slide
along
and
see
as
the
object
or
whatever
it
is,
is
moving
from
place
to
place
instead
of
using
the
simulator
you
to
immediately
analyze
this
or
push
it
through
new
pick
and
get
a
result,
you
can
also
just
save
that
track
out
to
a
csv
file
which
that
can
be
then
used
later
to
run
through
newpick
from
the
command
line
and
then
start
the
server
with
that
data
to
be
displayed.
A
A
So
if
you've
got
android
or
iphone,
there
are
some
applications
available
for
getting
data.
Gps
data,
the
one
I
used
was
called
gps
kit,
a
pretty
decent
application,
cost
10
bucks.
So
that's
a
bit
of
a
bummer,
but
it
was
worth
it
to
me
to
do
this,
so
that
has
a
feature
where
you
can.
You
can
track
yourself.
However,
you
can
create
as
many
tracks
as
you
want
and
then
export
it
into
gpx
format,
which
I'll
talk
about
in
just
a
minute.
There's
also
an
app
called
a
gps
logger
for
android.
A
I
have
not
tried
this,
but
it
does
advertise
that
it
can
export
gpx
format.
The
gpx
format
itself
is
just
an
xml
format
for
location
information.
We
can
take
a
quick
look
at
this,
just
to
kind
of
show
you
what
it
looks
like.
So
here's
a
sample
gpx
file,
it's
just
xml,
it's
got
tracks,
tracks,
have
segments
and
segments
have
points.
Each
one
can
have
an
elevation
of
time
and
a
latitude
and
longitude.
A
So
that's
what
a
gpx
file
looks
like
and
that's
what
we're
we're
going
to
be
dealing
with.
When
I
tell
you
how
to
get
your
your
gps
information
into
this
geospatial
tracking
application.
A
So
I
have
a
bunch
of
gpx
files
that
I
collected
with
gpx
kit
for
iphone,
and
here
is
one
example
of
this.
This
is
a
my
dog
walking
route,
so
this
is
typically
what
I
do
every
day
I
take
the
same
route
every
day
and
we're
gonna
take
this
gpx
information.
I'm
plotting
this
in
google
earth
right
now,
which
plots
gpx
files
and
I'm
going
to
we're
going
to
take
this
data,
we're
going
to
pass
it
through
the
new
geospatial
application,
and
I'm
going
to
show
you
how
to
do
that
right
now.
A
So
let's
kill
the
server
clear
this
out.
So
I
have
my
walk
data
in
a
file
called
data,
slash
walks.
So
here's
here
are
all
my
gpx
files.
So,
as
you
can
see
their
time,
they're
dated
here
so
here's
my
walk.
You
know
friday
thursday,
tuesday
wednesday,
and
these
are
the
files
that
we're
going
to
display
and
analyze.
A
So
there
is
a
tool
that
will
convert
this
gpx
data
into
a
format
that
new
pic
geospatial
can
read
it's
in
tools.
It's
called
convert
gpx
and
I
just
point
it
to
the
directory
where
my
gpx
files
are
and
it
creates
an
output
file,
so
it
created
output,
file
and
output
walks,
and
I
can
add
a
few
options
to
this.
I
can
say
verbose,
I
can
say:
output,
temp
or
something
like
that
are
maybe
not
output
temp.
I
don't
know
if
that's
proper
option
but
sorry,
but
I
can
say.
A
Tell
me
here's
what
I'm
reading
here's
the
tracks
that
I'm
processing,
so
it
will
actually
look
through
all
of
the
gpx
files
that
are
in
that
directory
and
break
out
every
single
track,
because
one
gpx
file
can
have
many
tracks
and
create
an
output
format
that
has
them
separated
by
track.
So
that's
what
I've
done
in
my
file
is
an
output
walks,
dot
csv
the
next
step-
and
this
is
all
in
the
readme,
like
I
said,
is
I
want
to
run
this
csv
file
through
nupic
geospatial.
A
So
what
this
run
step
will
do
and
there's
there's
several
options
that
you
can
specify
here.
The
only
option
I'm
going
to
use
is
manual
sequence.
So
let's
do
and
I'll
explain
what
this
is
in
just
a
moment:
output
walks
csv
dash
m.
So
if
you
do
not
specify
dash
m
the
logic
and
nupic
geospatial
will
break
up
sequences
for
you
automatically,
based
on
a
couple
of
different
factors,
one
of
them
being
time
deltas.
A
So,
if
it's,
if
it's,
if
it
sees
one
long
gps,
file
or
or
one
long
track
or
coordinate
information-
and
it
sees
big
gaps
between
portions
of
it,
it
will
assume
that
those
are
sequences
and
it
will
tell
nupik
to
reset
itself
between
those
kind
of
learning
each
one
as
its
own
sequence.
I
don't
want
it
to
do
that,
because
every
one
of
my
tracks
is
just
one
dog
walking
route,
so
I'm
going
to
tell
it
dash
m.
A
That
means
I've
manually
already
segregated
up
my
tracks
and
the
can
the
gpx
conversion
process
does
that
for
you,
so
I'm
going
to
tell
it
to
do
it
manually
and
now
it
is
going
to
run
those
all
through
new
pics.
So
nubik
is
actually
running
right.
Now
we
get
a
few
couple
of
debug
statements
coming
out
of
there.
If
you
want
to
see
some
more
details,
you
run
it
with
a
verbose
and
you'll
see
a
ton
of
stuff
flowing
through
the
screen,
so
you
can
see
that
it's
actually
doing
something.
A
A
A
A
So
just
like
you
saw
in
the
other
demonstration,
the
the
first
route
is
usually
pretty
anomalous,
because
it's
never
seen
anything
before
so
start
it
off
here,
and
you
can
kind
of
there
are
some
controls
down
here
so
that
you
can
track
along
with
the
route
and
see
at
each
point
where
exactly
what
color
it
is.
So
that's
the
first
route
entirely
anomalous.
Let's
move
ahead,
try
and
find
the
second
route
there
we
go.
A
This
one
is
not
so
anomalous,
but
still
anomalous.
If
you'll
notice,
I
can
show
these
at
the
same
time
my
second
route,
I
kind
of
took
this
weird
dog
leg
off
to
the
side
here,
because
one
of
my
dogs
had
done
their
business
down
here
and
I
knew
there
was
a
dumpster
on
this
little
dog
leg.
A
So
I
don't
like
to
walk
with
a
bag
full
of
dog
feces,
so
I
decided
to
dump
it
off
at
the
dumpster
and
you
can
see
there's
a
bit
of
an
anomaly
there,
because
it's
not
something
that
it
saw
the
first
track.
There
are
some
other
anomalies
along
other
parts
of
the
route
and
the
thing.
B
A
My
dogs
have
typical
places
where
they'll
stop
and
do
their
business,
but
it's
not
always
exactly
the
same
place
on
every
walk,
so
that
could
be
because
of
that
so
going
ahead
to
the
third
track
there.
It
is
this
one's
getting
a
lot
better.
So
no
real
high
anomaly
scores
on
this
one
and
then,
by
the
time
we
get
to
the
fourth
track.
B
A
A
good
representation
of
this
route
at
this
point
and
that's
cool,
so
one
thing
you
might
be
wondering
is
if
you're
familiar
with
new
pick
at
all
is
how
is
this
input
data
getting
encoded?
A
So
we
have
our
own
special
encoder
for
this.
That's
recently
been
developed,
it's
called
the
geospatial,
coordinate,
encoder
and
there's
an
entire
video
on
this.
So
if
you're
interested
in
how
the
these
locate
this
location
information
is
being
encoded
and
eventually
converted
into
sdrs
into
new
pick,
you
can
watch
this
video
of
chaitin
serper,
a
numenta
engineer
talking
about
exactly
how
the
geospatial
coordinate
encoder
works
internally,
but
that
is
a
little
off
topic
for
this.
So.
A
Do
a
couple
other
interesting
things:
I've
got
another
walk
track
that
I'm
going
to
convert
now
tools
convert
in
my
data.
I
also
have
a
walk
that
is
called
walks
reverse,
so
you
might
know
what's
going
on
here,
but
let's
I'll
just
run
it
and
still
in
output
walks,
and
I
still
want
to
manually
separate
the
sequences
based
on
track
name
so
I'll
throw
that
in
there
and
now,
let's
start
the
server
again
and
take
a
look
again.
A
A
Okay,
something's
wrong.
This
isn't
what
I
expected.
Oh,
oh,
I'm
an
idiot!
Excuse
me
when
I
did
the
run.
I
did
it
on
the
old
walks
file.
It's
actually
called
walks
reverse
now,
so
I
was
actually
running
that
against
the
old
output
file.
So
now
we're
running
it
on
the
reverse
file
and.
A
I
did
one
walk
after
all,
those
other
walks
where
I
did
the
reverse
of
that
route,
because
I
wanted
to
see
what
newpick
thought
of
that.
So,
let's
start
and
see
what
newport
thought
of
that.
So
the
server
has
started
and
I'm
going
to
start
with
the
first
track
again
completely
an
anomalous.
The
second
one-
and
this
looks
like
it's
the
second
and
the
third
kind
of
combined.
So
second
and
third,
are
the
same
pretty
much
as
last
time.
A
The
fourth
track
is
in
here
somewhere
I'll
have
to
trust
others
right
there
and
then
a
couple
days
later.
I
did
this
track.
So
if
you
look
closely
at
this,
I'm
I
am
starting
in
the
same
place,
but
I'm
taking
a
left
instead
of
a
right,
and
that
is
anomalous.
That
is
not
something
that
lupic
has
ever
seen
before.
So,
as
you
can
see,
there's
decent
levels
of
anomalies
along
the
entire
route.
A
Until
I
get
back
to
the
the
bottom
leg
of
the
route,
when
I'm
heading
back
to
the
house
and
that's
all
entirely
green,
because
that's
a
typical
track
that
I'd
seen
before
so
that's
something
that
you
pick
will
definitely
pick
up
if
you're,
on
the
exact
in
the
exact
same
location
somewhere,
you've
been
before,
but
you're
not
moving
in
the
same
direction.
A
A
A
Output
walks
production
manual,
sequence,
resets
and
let's
take
a
few
seconds
here
and
then
we'll
start
the
server
again
python
server.
You
will
see
immediately
what
I'm
doing
here,
but
let's,
let's
hide
that
for
a
sec,
all
right,
first
route,
same
one,
the
second
third,
fourth
routes.
All
the
same,
so
it's
learning
the
same
pattern
basically,
but
I'm
trying
to
trip
it
up
on
this
last
one.
So,
let's.
A
A
A
So
this
I
call
like
the
abduction
case.
So
if
this
were
something
like
I
had
attached
to
my
daughter
and
when
she
was
walking
the
dogs,
if
we
theoretically
had
some
type
of
application
that
was
live,
feeding,
gps
information
from
her
location
into
new
pic
and
I
was
getting
notified
whenever
there
were
significant
anomalies.
A
This
would
notify
me
almost
immediately
that
something
went
wrong
on
the
dog
walk,
so
I
think
that's
pretty
pretty
interesting
and
could
be
very
applicable
to
a
lot
of
different
scenarios
or
situations.
A
So
that
is
that's
pretty
much
it.
So
I've
gone
over
getting
your
own
data
files,
your
own
gps
data
into
this
sample
application,
nupix
geospatial
tracking
application.
We've
talked
a
little
bit
about
gpx.
I
talked
a
little
bit
about
the
geospatial
encoder
but,
like
I
said,
if
you're
interested
in
seeing
how
that
data
is
encoded,
you
really
should
check
out
that
video
about
it.
A
That
goes
in
quite
a
bit
of
detail
and
it'll,
probably
help
you
understand
exactly
why
the
anomaly
scores
went
up
so
high
when
my
speed
changed
so
dramatically
on
that
last
abduction
route.
So
I
hope
you
enjoyed
this.
If
you
want
to
play
around
with
this,
please
do
I'd
love
to
see
what
it
what
types
of
things
you
guys
in
the
community
come
up
with.
There's
a
lot
of
potential
applications
for
for
this
type
of
technology,
just
think
about
logistics,
logistics,
trucks,
deliveries,
typical
routes
for
ups.
A
You
know
trains
planes,
whatever
all
those
stuff
that
have
typical
patterns
within
their
routes.
This
could
be
useful
for
live
anomaly
detection
on
any
of
those
patterns
when
something
goes
awry
during
the
typical
workday
or
what
have
you
so
thank
you
for
watching,
and
I
hope
you
guys
have
fun
with
this.
It
was
fun
to
play
around
with
I'm
really
excited
about
it
and
be
sure
to
check
out
the
video
about
the
geospatial
coordinate
encoder
thanks.