►
From YouTube: DevoWorm #29: Image transformation, tracking/force microscopy, mu-resolution, shape characterization
Description
GSoC Updates, image transformations (inferring locations on the surface of a sphere). Paper on particle tracking algorithms and traction force microscopy. D'arcy Thompson's phenotypic warping and an insect egg dataset. 3-D printed embryos and the challenges of microresolution. Attendees: Alon Samuel, Richard Gordon, Susan Crawford-Young, Bradly Alicea, Morgan Hough, Jesse Parent, Karan Lohaan, Harikrishna Pillai, Longhui Jiang
B
B
Yeah
on
my
eyes,
yeah
our
eye
left
eye
yeah.
A
B
B
So
my
eye
gets
tired.
It's
just
like
just
leave
it
alone
for
a
couple
weeks,.
A
B
B
A
Yeah
so
yeah
welcome
to
the
meeting
I
have.
I
guess
we
have
g
suck
updates
again,
I
guess
hurry
krishna.
Do
you
want
to
give
us
an
update.
C
Onto
an
ellipsoid
by
giving
texture
coordinates
to
the
to
all
the
3d
coordinates
of
the
ellipsoid,
but
I
needed
a
slightly
different
way
because
I
don't
need
to
map
everything.
That's
on
the
image,
so,
like
only
a
specific
part
of
the
image,
so
I
was
trying
using
ue
wrapping
now,
so
I
recreated
it
using.
I
recreated
the
the
thing
which
I
wanted
using
blender.
A
C
A
C
A
C
I'll
be
finding
a
way
to
uv
lab
the
thing
which
I've
done
in
blender
the
same
way
I'll
do
the
same
thing
I'll
do
in
python.
So
that's
a
major
thing
which
I'll
be
doing
next
week.
A
Okay,
that's
good!
Thank
you,
hello,
quran,.
E
C
Doing
okay
eating
last
night,
you
know
I
was
traveling,
so
I
got
delayed
last
time.
Yes,
so,
regarding
the
updates,
you
know
I'm
still
struggling
with
the
projection
part.
So
far.
The
thing
is,
you
know,
with
those
eight
images,
all
the
image
stitching,
algorithms
that
are
there
they're
mostly
for
you
know,
panoramic
images.
You
know
images.
A
C
Not
that
short
through
that
one
camera,
so
I'm
still,
you
know
getting
better
at
the
projection
part.
So
that
thing
is
there
otherwise.
C
The
yeah,
so
this
is
this:
according
to
the
you
know
this
last
week
I
was
supposed
to
you
know
finalize
the
projection
part,
so
this
thing
is
kind
of
getting
more
delayed,
so
I'll
continue
with
it
this
week
as
well,
and
this
week's
work
would
be,
you
know,
adding
additional
features
in
in
the
form
of
you
know,
more
more
labels
or
like
we
had
you
know,
decided
upon
unless
me
and
harry
krishna
about
the
extra
features
you
know
that
we
could
add
to
the
model
itself,
zooming
interacting
with
the
model,
and
you
know,
labeling
different
parts
of
the
model,
so
we'll
be
working
on
that
as
well,
but
the
majority
of
the
attempts
will
still
go
into
this
projection.
C
A
C
Just
you
know
stitching
the
way
hari
krishna
is
doing
it
or
the
way
I
had
done
it.
You
know
when
I
had
shown
my
prototype.
That
way,
stitching
stitching
is
like
we
can.
You
know
just
go
with
that
algorithm
if
nothing
else
works,
but
the
existing
algorithms
that
are
there
three.
C
So
then,
if
you're
using
those
two
algorithms
for
you,
know
stitching
all
those
images
together
for
the
sphere.
C
D
I
was
discussing
this
with
fred
bookstein
a
bit
he
works
on.
Okay,
oh,
I
don't
know
if
you
know
the
old
pictures
from
darcy
thompson
well
over
100
years
ago,
at
any
rate
yeah,
I
think
bradley
is
badly
familiar
with
them.
Yeah
you've
got
a
suggestion
and
that
is
to
use
since
since
we're
confining
ourselves
to
a
spear,
how
about
decomposing
the
images
before
they're
overlapped
into
spherical
harmonics
and
might
that
help?
D
Now
I
don't
know
what
spherical
harmonic
like
on
a
patch
over
the
surface
rather
than
the
whole
surface,
but
I
think
that
might
be
quite
interesting.
C
So
if
you're
taking
like
an
image
of
a
sphere,
I
think
it's
it's
like
you're,
taking
an
image
of
a
negative
inversion
kind
of
a
spare
thing
you
know
like,
but
when
we
have
lenses,
you
know
a
fish
eye
lens.
So
the
way
a
concave
lens
would,
you
know,
bought
the
image
and
the
way
a
convex
lens.
Would
you
know,
walk
the
image
that
negative
or
positive
inversion?
C
D
D
C
D
C
Easily,
if
pitching
algorithms
kind
of
deal
with
overlaps
pretty
well,
but
they
don't
deal
with
these
inversions
very
well.
You
know
like
so
it's
like
if
you're
taking
a
picture
of
a
plane,
it's
easy
if
you're
taking
a
picture
of
something
like
that
is
more
concave
or
convicts,
it
kind
of
you
know,
starts
giving
back
your
results.
Okay,.
D
You
could
approximate
overlap
by
saying
I'm
going
to
make
an
image
so
spherical
harmonics
are
the
average
of
the
two
spherical
harmonic
amplitudes
for
each
so-called
frequency
or
whatever
you
want
to
use.
D
Maybe
yeah
more
images
or
somehow
introducing
iteration
into
just
two
images:
okay,
okay,
so
the
point
is
that
if
you
have
two
overlapping
images,
there's
spherical
components,
the
the
amplitude
of
the
spherical
harmonics
should
be
similar
in
some
sense:
yeah
yeah,
okay.
E
Can
I
can
I
just
ask,
what
do
you
mean
to
decompose
an
image
to
spherical
harmonics?
Do
you
mean,
like
the
functions
yeah
certain
functions
to
see
the
coefficients.
D
E
I
think
I
think
I
I'm
missing
some
some
background
on
the
property.
E
I'm
familiar
with
decomposing
an
image
to
like
frequencies.
D
If
right
right,
these
are
the
corresponding
series,
decomposition
of
a
spherical
function
with
a
constant
radius.
D
I
see
and
they
produce
a
spherical
harmonics
produce.
What
do
you
call
it?
A
a
set
of
component
systems
which
span
the
whole
space.
In
other
words,
you
can
represent
any
spherical
surface
by
an
infinite
expansion
of
the
spherical
spherical
harmonics,
so
in
a
in
other
words,
their
basis
function
for
this
for
functions
on
a
sphere.
D
B
A
So
this
would
be
a
very
similar
thing
where
you'd
have
like
images
that
were
sort
of
tiling
a
sphere
and
then
you're
finding
like
you're
trying
to
press
reference
locations,
I
guess
but
yeah.
I
think
it
would
be
nice
to
try,
especially
if
we
get
around
some
of
these
challenges
with
some
of
the
other
things
that
the
algorithm.
I
know
that,
but
what
kron's
talking
about
a
lot
of
the
algorithms
assume
like
a
panoramic
view
like
they'll,
have
these
360
videos
where
you
know
you
have.
A
D
Are
any
of
the
panoramic
software
based
on
sine
and
cosine
coefficients.
D
When
you
match
they
match
in
real
space
or
in
the
space
of
the
amplitudes
of
the
sine
and
cosine
coefficients.
C
No,
no,
I
mean
the
algorithm
does
use
these
things,
but
not
in
that
general
sense.
It's
like
it's
more
about
you
know,
feature
tracking
and
then
stitching
based
on
that
feature.
Instead
of
you
know,
oh
okay,.
D
A
I've
heard
of
his
name-
I
don't
know
much
about
the
world.
Okay,.
D
D
Yeah
yeah:
that's
what
inspired
it
originally
now
he's
he's
retired.
Now
I
tried
to
entice
him
to
come
to
our
group
and
give
us
a
lecture
yeah
if
it
would
probably
help.
If
you
is
that
all
right
yeah,
I
could.
A
B
Yeah
yeah
this
is
that
that's
a
a
software
package
called
first,
that's
in
fsl
and
brian
brian
patnode
did
that
together
with
mark
jacobson,
but
I
know
that
they
they
used
some
of
the
you
know
they
were
inspired
by
his
work.
B
A
B
That,
what's
in
the
eg,
electro
and
stuff,
while
grouping
yeah
it's
when
you're
you're,
if
you
see
my,
I
guess
my
my
image
isn't
when
you
got
the
head
covered
in
electronics,.
B
Well,
because
you
don't
you,
don't
have
you
know,
you
can't
sample
the
neck
right,
so
you
don't
have
the
sphere
properly
sampled
and
so
there's
there's
a
polar
average
reference
effect
correction.
That
is
a
paper
from
back
in
the
90s.
B
Did
this
and
and
so
you're,
basically
assuming
yeah
you're,
assuming
a
sphere
and
you're
extrapolating
on
the
sphere?
Okay,
so
these
other
locations,
yeah,
there's
a
lot
of
spherical
harmonic
stuff
on
pg.
D
C
D
So
there
we
got
two
papers
going,
one
for
just
standard
montage
with
sine
and
cosine
expansion
instead
of
feature
analysis
and
I'm
trying
to
do
it
with
the
embryo.
A
Okay,
yeah
well,
thank
you
quran
for
that
update.
That
was
hopefully
worked,
some
issues
out
for
you
and
then
thank
you
very
krishna
for
your
update
as
well.
We
have
some
things
in
the
chat
here.
I
think
this
is
just
morgan,
saying
hello
from
apostle
robles,
which
is
in
california,
then
talking
about
spherical,
harmonics
and
eeg,
some
other
things
here.
That's.
A
Yeah,
so
thanks
yeah,
I
know
that
morgan's
on
the
road
and
jesse
is
here
so
hello
did
either
of
you
have
any
updates.
You
want
to
share
with
us
or
just
comments
morgan's
already.
D
Oh
for
those
of
you
who
are
new
to
this,
I'm
bouncing
up
and
down
because
I'm
on
a
treadmill.
D
A
D
D
D
Okay,
so
we're
going
to
stretch
your
brains.
A
D
A
E
Yes,
no
not
that
I
have
some
questions.
What
I
did
after
the
meeting
I
found
a
video
of
of
the
bar,
but
correctly
there.
E
And
yeah,
just
after,
like
some
maybe
like
checking
about
it
with
with
dick,
I
I
wanted
to
kind
of
clip
just
a
short
like,
maybe
like
a
10
or
30
seconds
of
it.
It's
gonna
be
just
static
because
some
of
them,
just
like
the
the
point
of
view
just
moves
around.
E
So
I
downloaded
like
a
movie
from
youtube
just
one
of
them
and
I
I
just
clipped
it
to
like
the
what
I
need.
I
wrote
the
code
and
then
after
that
you
send
me
the
link
to
the
vaselyria
repo
that
there
is
some
default
people
on
diva
worm.
So
I'm
just
gonna
add
that
afterwards,
so
yeah,
that's
a
get
like
that.
Far
so
much.
I
kind
of
also
didn't
didn't
see
on
it
so
much.
E
A
I
think
so,
yes.
E
E
I
still
didn't
install
like
the
repos
on
my
kinda
like
machine,
so
I
kinda,
like
I
kinda,
like
check
everything
properly,
so
I
also
kind
of
plan
to
do
that
as
well
yeah,
and
I
think
I
think,
after
that
I
was
having
like
some
thoughts
about
like
how
the
main
question
about
like
proving
the
bastille
area,
that
they
have
like
just
a
smooth
action.
D
Would
connect
with
center
of
mass
or
center
of
something
that
can
just
mention?
I
can
find
that
along
one
comment
about
that.
The
first
observation
of
the
jerky
movement
of
single
diatoms
was
done
in
1979
at
10
frames
per
second,
okay,
so,
okay,
since
most
of
your
data,
is
probably
30
frames
per
second.
Yes,
yes,.
E
D
E
D
E
D
The
reason
we
have
to
do
this
carefully
is
that
smooth
motion
of
diatoms
may
be
an
optical
illusion:
okay,
okay,
therefore,
we
can't
trust
our
eyes.
D
Or
like
a
geometry
that
should
do
the
only
problem
you
you
might
have
is
if
a
diatom,
if
the
view
of
a
diatom
changes,
then
its
two-dimensional
image
may
not
be
the
same.
Yes,
okay,
yeah,
so
you
you
want.
Basically
what
you
want
is
colonies
or
a
portion
of
a
colony.
D
D
D
E
Okay,
okay,
so
I
think
it's
gonna
be
more.
I
don't
know
if
it's
gonna
be
a
tracking
question,
maybe
more
because
the
most
tracking
models
they
kind
of
assume
a
smooth
motion
to
begin
with,
or
they
have
like
they
have
their
own
filters,
assume
smooth
motions,
I'm
familiar
with
common
filter,
that's
connects
for
tracking,
and
that
assumes
like
smooth
motion
anyway.
E
So
I
think
it's
going
to
be
more
of
like
a
just
a
detection
problem
frame
to
frame
and
that's
kind
of
like
it,
then
just
connect
together.
Okay,
good.
D
E
Yeah,
okay
and
we
got
an
assumption
like
each
cell.
It
doesn't,
it
doesn't
change
like
the
formation,
it
doesn't
change
unless
it's.
E
Morgan,
you
wrote
about
eurov7,
yes
got
it.
I
think
I
think
every
like
yeah
detection,
yeah
network
can
just
yeah
gonna
be
able
to
use
it
hope
they
have
some
yeah
pre-trained
networks
and
diatoms
or
a
city,
but
I
think
not.
B
Yeah
yeah,
I
mean,
I
know
that,
like
you
know
like
pg,
is
you
know,
there's
the
more
kind
of
typical
microscopy
stuff,
but
but
this
sounds
like
a
pretty
pretty
typical
object.
Recognition.
Yes,.
D
A
That's
good
well,
thank
you,
elon
for
the
update.
If
you
need
anything,
let
us
know
we
have.
You
know
the
institutional
knowledge,
except
because
we've
been
doing
this
for
a
couple
years,
so
you
know
we
can,
if
you
see
something
that
maybe
doesn't
make
sense
or
something
let
us
know.
A
I
was
gonna
give
out.
I
wanted
to
kind
of
get
an
update
of
the
okay,
so
yeah
where's.
A
A
Was
just
kind
of
raining
a
lot
there
yeah
I
got
caught
in
the
rain
one
day,
but
yeah
it's!
It
is
winter
yeah.
Well,
it's
yeah!
So
yeah.
I
guess
we
have
the
gnns
project
long
when
he's
involved
in
that
I
don't
know
if
he
can
put
anything
in
the
chat
about
what
where
they
are
in
that
process.
I
know
we
got
some
updates
last
week
from
I
think
from
wataru
and
they're
progressing.
A
The
otaro,
I
think,
is
getting
through
stage
one
and
they
have
the
stage
one
stage
two
and
stage
three
yeah
and
so
stage.
One
I
think
is
getting
to
coming
to
completion
stage.
Two
is
something
that
jia
hong
is
working
on
independently.
I
think
he's
making
good
progress
on
that
and
then
stage
three
I
guess
they're
starting
to
work
on
now.
So
hopefully
you
know
we'll
get
some
I'll
I'll
elicit
some
updates
as
well
for
the
week,
so
we'll
keep.
On
top
of
that.
A
So
it's
you
know
as
a
time
when
you
kind
of
want
to
get
everything
sort
of
you
know
start
to
think
about
like
generating
documentation,
and
you
know
working
on
feet,
adding
features
to
the
interface
or
whatever
you
want
to
do,
bring
it
to
a
completion
point,
so
you
don't
need
to
like
you
know
you
can
deviate
from
what
you
proposed
in
the
final
thing,
but
you
do
need
to
have
something
that
you
can
submit
that
can
be
executed,
so,
whether
that's
like
a
notebook
like
a
colab
notebook
or
if
it's
like
a
repository
that
they
can
just
install,
I
mean
because
they're
going
to
check
this
and
see
if
it
runs
or
whatever.
A
So
that's
that's
what
they're
looking
for,
but
I
mean
the
the
point
being
is
that
you
know
when
it
comes
to
a
completion
on
some
of
the
work.
Now
you
can
com
work
on
it
afterwards.
I
just
want
to
come
to
some
sort
of
stopping
point
for
the
submission,
and
so
I
mean
you
know
if
it's
like
something
that
their
longer
term
things
that
we
want
to
do,
that's
fine.
A
We
don't
have
to
implement
them
in
this
next
few
weeks,
but
we
do
want
to
get
to
a
point
where
we
can
submit
something
and
then
just
you
know
continue
on
from
there
if
necessary
and
yeah.
So
I
mean
you
know
after
we're.
After
we're
done,
I
think
we're
gonna
have
some
nice
tools.
A
We
have
the
divalern
repository
where
the
diva
learn
program,
which
has
been
developed
over
the
past
few
years,
we're
going
to
try
to
incorporate
the
gnn
into
this
into
divalern,
so
we're
going
to
have
like
a
new
module.
A
Then
the
spherical
representations,
the
digital
microspheres.
We
have
two
approaches
here:
rival
approaches
are
in
krsna's
quran
and
we'll
try
to
stitch
those
together
into
one
not
now,
but
we'll
do
that.
Maybe
later,
but
right
now
we'll
have
like
you
know
two
products
that
are
sort
of
complementary
and
then
eventually
we'll
be
able
to
put
that
together
into
something
that
people
can
use.
A
And
you
know
the
reason
we're
doing
this
kind
of
rival
or
approach
is
because
you
know
there
might
be
one
approach,
that's
better
for
some
types
of
embryos
and
the
other
approach
might
be
better.
For
other
types
of
embryos,
so
it's
nice
to
give
people
that
option
and
especially
when
you
have
the
diversity
of
the
biological
world
we
know
like,
as
we've
talked
about
in
our
meetings,
it's
quite
vast
and
one
you
know
one
set
of
techniques,
don't
necessarily
work
for
another
model
organism.
So.
A
I
I
in
so
in
some
of
the
other
in
the
other
group,
I'm
working
with
on
gsoc
projects.
A
We
had
mentioned
that
we
were
going
to
do
like
a
presentation
at
some
point
in
the
future
on
their
projects
like
a
summary
of
their
projects
and
at
some
point,
not
not
necessarily
before
september
7th
I'd
like
to
get
our
gsoc
students
and
I'd
like
to
give
them,
you
know
give
them
a
short,
have
a
give,
a
short
presentation
or
kind
of
do
a
short
review
of
their
finished
product.
A
So
we
can
advertise
it
for
next
year
and
if
we,
you
know,
we
did
advertise
it
for
various
things,
so
that
those
are
just
things
coming
up.
I
want
to
keep
people
aware
of.
D
Right
yeah
paso
robles
was
on
earlier.
B
Yeah
yeah
happy
too,
as
as
well
as
they're,
more
their
more
recent.
More
recent
work
is
really
just
yeah.
Our
harmonic
harmonic
analysis
on
on
connectomes
is
kind
of
an
extension
of
that:
okay,
okay
and
yeah,
that
that
takes
the
stuff
to
2022.
B
A
Yeah,
so
I'm
going
to
share
my
screen.
I
want
to
go
over
some
things
that
I
kind
of
present
for
the
week.
Let's
see
if
my
screen
shares.
A
Okay,
it
says
it's
sharing.
I
think
it
is
all
right.
So
the
first
thing
I
wanted
to
mention
was
something
that
elon
brought
up
in
his
update
that
there
was
some
activity
in
the
diva
learn
repository.
A
So
I
put
this
in
the
slack
and
now
kumar,
wataru,
sushmanth
and
nawan
were
all
on
this
issue,
and
so
then
this
is
the
issue
here.
It's
number
69,
so
this
is
diva
learn.
This
is
the
diva
learn
repository
and
the
diva
learn
organization.
A
This
is
the
issue
tracker,
so
we
have
a
number
of
issues
here.
If
you
want
to
address
some
of
these
issues,
you're
free
to
do
so,
you
know
you
just
put
in
a
you
put
in
you
basically
say
I
want
to
address
this.
You
put
in
a
comment,
so
there
are
a
lot
of
different
maintenance
things
here.
There's
long-term
vision,
there's
some
of
these
things
that
are
just
like
unit
tests
that
aren't
consistent,
they're
enhancements
that
we
might
want.
A
You
know
they're
different
things
like
that.
So
at
this
number
69,
which
is
here,
this
is
the
bug.
Collab
demo
requires
multiple
upgrades,
and
this
is
something
that
was
going
on
with
like
I
guess
it
was
a
compatibility
issue,
so
maya
who's.
The
maintainer
here
put
out
some
points
here.
This
is
a
beginner-friendly
issue
that
he
proposed
for
people,
because
you
know
not.
A
He
doesn't
have
time
to
do
the
actual
work
and
I
don't-
and
you
know
so-
we
want
to
get
people
involved
and
then
alan
said
he
wanted
to
take
a
look
at
it.
He
took
a
look,
then
sushma
got
involved
here
and
he's
basically
going
through
and
then
my
oak
said
shows
them
where
the
co-lab
notebook
is,
and
they
took
a
look
than
a
not.
I
think
he
resolved
it
at
some
to
some
degree.
A
Then
alan
was
discussing
this
with
him
and
not
then
so
yeah.
They
basically
went
through
this,
and
I
I
don't
know
elan.
Did
this
get
resolved
or
is
there
more
to
do
here
or
what.
E
Yeah,
I
think
anand
said
something
that
it's
he
managed
to
make
it
work,
but
I
didn't
I
didn't.
I
didn't
look
at
it
thoroughly
after
they
said
because,
after
a
while,
I
understood
that
they
need
to
download
both
repos
the
diva
learn
and
they
kind
of
like
the
other
one
that
the
notebook
is
at.
So
I
can
just
connect
them
both
yeah.
So
I
still
need
to
check
the
if
it
kind
of
works,
because
anand
said
that
now
it's
kind
of
worked
for
him.
E
E
A
I
mean
this
is
something
we
might
consider
making
a
another
issue
on
for
like
if
there's
something
that
doesn't
get
resolved,
you
can
make
another
issue
for
like
like
put
it
in
the
documentation
or
how
it
might
be
clarified
or
something
I
mean
just
I'm
just
trying
to
suggest
best
practices
here,
because
it's
always
good
to
have
like
you
know
when
people
do
contributions
that
there's
like
follow-up
on
them.
Otherwise
it's
kind
of
like
it
gets
lost,
but
then
that's
good,
and
so
there's
this,
maybe
there's
a
pull.
A
A
A
Well,
my
yoke
is
the
main
mayuk
is
the
maintainer
he's,
usually
the
one
who
approves
the
pull
request,
because
this
is
sort
of
the
thing
that
he
originally
put
together.
Like
you
know,
he
was
able
he
was
the
one
who
sort
of
proposed
this
this
version
of
divalern
and
then
you
know,
he's
he's
been
able
to
work
on
maintaining
it,
and
you
know
he
has
a
pretty
good
technical
sense
of
what
needs
to
be
done
for
it.
A
So
I
leave
it
up
to
him
as
to
like
approving
call
requests
yeah,
but
we,
you
know,
we
always
need
people
if
people
want
to
take
a
part
of
it
on,
and
you
know
it's
like
you
know,
when
you
build
the
software
things
go
out
of
date,
things.
Don't
always
you
know,
we
always
have
issues
with
people
trying
to
use
it,
and
so
that
that's
why
it's
important
to
have
these
sort
of
issues
and
follow-ups,
and
things
like
that
yeah.
Well,
thank
you,
yeah.
That
was
great.
A
Okay,
yes,
so
that
that's
some
nice
open
source
activity.
I
also
found
a
paper
on
particle
tracking
and
we
talked
about
this
with
respect
to
some
of
the
stuff
going
on
in
the
gnns
project.
A
A
So
it's
it's
definitely
related
to
sort
of
the
the
motion
tracking
and
the
cell
tracking
that
we
do
in
the
project
here
in
different
ways,
and
so
this
is
a
nice
new
paper
on
this.
So
this
the
abstract
I'll
just
go
through
the
abstract,
maybe
some
of
the
images
it
just
you'll
need
to
read
the
whole
paper.
A
If
you
want
to
get
a
really
good
technical
appreciation
for
it,
the
abstract
reads:
deformation
measurement
is
a
key
process
in
traction,
force
microscopy,
which
is
tfm,
and
so
I'm
not
familiar
with
traction
force
microscopy,
but
it's
this
detect
a
specialized
technique
that
people
are
using
conventionally
particle
image,
volt
velocity
symmetry
piv.
I
know
what
that
is,
or
correlation
based
particle
tracking
velocimetry
cptv
have
been
used
for
such
a
purpose
using
simulated,
bead
images,
and
I
guess
they
show
those
in
the
paper.
A
So
when
they're
tracking-
something
in
the
image,
you
know
they're
trying
to
build
this
displacement
vector
and
they're,
showing
these
methods
fail
in
in
some
ways,
especially
for
large
displacement
vectors
here
to
redeem
the
potential
large
vectors,
we
propose
a
two-step
deformation
tracking
algorithm
that
combines
cptv,
which
is
this
method
in
particle
tracking
zelicometry,
which
performs
better
for
small
displacements
than
piv
methods.
A
newly
designed
retracting
algorithm
that
exploits
statistically
confident
vectors
from
the
initial
cptv
to
guide
the
selection
correlation
peak,
which
are
not
necessarily
a
global
maximum.
A
So
this
new
method
named
cptv
retracting
or
a
c
ptvr,
was
able
to
track
more
than
92
of
large
vectors,
whereas
conventional
methods
could
track
43
to
77
percent.
Of
these
correspondingly
traction
force.
Reconstructed
from
ptvr
should
better
recovery
of
large
traction
than
the
old
methods.
This.
This
method
applied
on
the
experimental,
bead
images,
has
shown
a
better
resolving
power
of
the
traction
with
different
sized
cell
matrix
adhesions
than
the
conventional
methods.
A
So
this
is
just
variations
in
the
in
the
thing
that
you're
imaging
altogether
cptvr
method
enhances,
enhances
the
accuracy
of
tfm,
which
is
this
traction
force
microscopy
in
the
case
of
large
deformations
present
in
soft
substrates,
which
are
like
biological
tissues.
A
So
I
guess
they
can
simulate
this
with
beads
in
a
gel,
so
you
can
do
this
with
or
without
cells
as
an
input
and
it
quantifies
the
gel
deformation
and
reconstructs
the
traction
field.
So
this
is
measuring
like
this
is
using
the
deformation
knowledge
of
this
gels
elastic
modulus.
So
it's
using
physical
parameters
to
show
these
these
tracks,
so
this
is
traction
reconstruction.
This
is
kind
of
one
of
these
techniques.
A
That
is,
you
know
it's
useful,
I
think
in
development,
because
you're
dealing
with
a
lot
of
these
type
of
phenomena
so
like
with
cell
migration,
especially
you'd,
probably
be
able
to
use
as
well
so
yeah
they're,
really
the
older
methods,
the
adopted
methods
from
experimental,
fluid
mechanics
which
have
been
well
established
and
generated
a
sufficiently
accurate
displacement
field
for
small
to
intermediate
deformation
levels,
but
the
large
deformation
levels
they
haven't
been
able
to
use
because
they're
using
these
techniques
that
they've
kind
of
brought
in
from
another
field.
A
Let
me
see
if
we
have
any
questions.
I
don't
know
if
we
have
any
questions
in
the
chat,
but
yeah
we'll
be
a
couple
here,
just
waiting
for
it
to
load.
A
Oh
yeah,
okay,
it
was
just
a
couple
things
from
morgan.
A
Okay,
this
is
it
so
here
here's
some
images
now.
So
these
are
the
typical
viv
methods,
a
topology
based
tracking
and
the
current
cptv
failed
to
track
a
large
local
deformation.
So
this
is
where
they're,
showing
what
piv
is
doing.
A
You
see
these
tracks
here
they
have
the
displacement
field
and
the
displacement
map
and
so
they're
showing
this
sort
of
how
this
works
in
a
regular
context.
If
you
see
cells
in
this
field
of
view,
they're
moving
around
they're
leaving
this
track
and
they're,
you
know
you
can
visualize
that
the
thing
is
is
when
they're
contracts,
yeah.
A
I
don't
think
so.
This
is
what
is
this
supposed
to
be.
A
E
Just
yeah,
it's
just
a
probability.
A
A
A
These
are
beat
images
where
these
beads,
instead
of
cells
and
they're,
just
showing
this
displacement
and
they're
they're
kind
of
getting
a
sense
of
you
know
what
what
the
the
sort
of
the
parameters
of
the
method,
and
then
they
have
this
displacement
map.
This
placement
field
that
they're
getting
from
these
images,
so
they
don't
yeah.
They
don't
mention
too
much
more
about
it
in
here,
then.
You
know
they
give
some
examples
of
different
methods
and
how
that
you
know
how
they
did
that.
A
It
has
found
been
found
to
make
tracking
more
reliable
and
robust
when
a
cross-correlation
is
used
to
prevent
information
as
a
predictor,
corrector,
so
they're,
using
cross-correlations
to
sort
of
correct
the
paths
and
to
predict
them.
But
if
you
have
they
have
this
problem
on
long
term
or
long
long,
distance
tracks,
where
this
cross-correlation
breaks
down.
A
So
this
is
they're
using
this
cptv
identifies
individual
beads
using
a
gaussian
mixture
model,
then
applying
a
normalized
cross
correlation,
and
then
this
cptv
retracting
algorithm
actually
improves
upon
this.
A
The
algorithm
assumes
that
the
first
time
cptv
is
performed
with
a
high
significance
criterion
and
a
strict
filtering
criterion
from
which
seed
points
consisting
of
failed
tracking
and
filtered
out
locations
are
determined
looping.
Through
these
seed
points,
neighboring
vectors
of
a
certain
search
radius
are
gathered
per
location
from
which
statistics
such
as
the
mean
and
standard
deviation
is
calculated.
A
So
they
do
a
lot
of
like
comparisons
and
thresholding
and
prediction,
and
so
this
is
how
they're,
making
this
cptv
retracting
algorithm
better
for
this,
and
so
then
they
go
through
some
more
methods
and
I'm
not
sure
if
they
get
into
any
more
images.
A
Let's
see
so
here
are
the
results
yeah,
so
they
created
a
see.
Okay
might
not
be
able
to
get
into
these
images,
but
it's
just
you
know.
This
is
a
nice
paper
on
this
method.
I
don't
know
you
know
we.
I
know.
We've
talked
about
cell
tracking
in
the
group
and
we've
had
some
yeah.
So
susan
said
this.
This
tracking
method
might
be
useful
in
my
research
on
mechanics,
so
I'll,
send
the
paper
out
and
by
email
and
slack,
so
people
can
read
more
about
it
all
right.
A
Oh
yeah
yeah,
no.
A
A
Yeah
well,
in
this
case,
you're
looking
at
like
an
object
tracking
through
like
a
medium
like
it
could
be
a
gel
or
it
could
be
like
liquid
and
it's
kind
of
moving
through
and
it's
like.
If
a
motor
boat
goes
over
a
lake
and
you
get
like
like
a
wake
and
you
interact
that
you
can
see
it
because
the
water
kind
of
tracks
it
for
you.
But
in
these
small
things.
B
A
B
A
I
think
in
like,
if
you
I,
like,
you
know
like
an
under
a
cover
slip.
Maybe
not,
but
if
you
were
looking
at
like
a
real
limit
real
time
image
of
like
an
embryo,
even
you
might
have
small
currents
that
you
know
displacement
currents
and
because
the
cells
are
so
small
they're,
you
know
really
hard
to
detect.
But
you
know
it's
possible.
I
guess
there's
got
to
be
some
displacement
of
the
fluids
in
there.
Yeah.
B
A
I've
seen
some,
I
think
I
went
to
a
couple
talks
on
like
how
they've
looked
at
flagellar
bacteria,
where
they
actually
create
a
wake.
I
mean
they're,
very
tiny
forces
but
they're
like
because
they're
moving
their
tails
around
they're,
creating
this
wake
and
they're
like
dynamics,
but
they're,
so
small
that
you
it's
hard
to
detect
them.
So
it's
yeah.
A
Okay,
I'd
like
to
finish
up
today
and
if
you
have
to
go,
that's
fine,
but
I'm
gonna
share
my
screen
again:
anyways
yeah,
I'm
going
to
talk
a
little
bit
about
something
if
I
can
get
the
thing
to
work,
but
I'm
not
sure
here.
So
this
is
okay
yeah.
So
here
it
is.
We
have
the
3d
printed
embryos
I
wanted
to
show
if
I
can
get
into
it.
A
So
I
think
it
was
morgan
who
set
this
up
and
he
sent
out
some
papers
on
3d
printed
embryos
or
he
sent
like
this
paper.
A
I
think
it
was
in
our
slack
about
this
group
that
was
like
printing
3d
embryos
and
they
were
making
like
little
models
of
of
them
from
like
they
would
take
images
and
then
print
them
out.
I
don't
know
if
this
was
decker
or
morgan,
but
in
any
case
this
is
yeah.
A
Yeah
yeah,
so
this
is
using
nmr
they're
able
to
you,
know,
image
things
they're
able
to
get
information
about
chemical
composition
of
things,
but
they
can
also
image
things
at
the
volume
scale
for
microorganisms
and
single
single
cells.
This
is
hindered
by
the
limited
sensitivity
of
the
detector
and
the
difficulties
in
positioning
such
small
samples
and
proximity
of
the
detector.
A
So
you
know
this
is
stuff
that
we
want
to
be
able
to
image,
but
it's
hard
to
get
good
three-dimensional
images.
So
recently
we
introduced
an
innovative
generation
of
nmr
probes,
based
on
the
combination
of
single
chip,
cmos
integrated
circuits
together
with
high
resolution,
pretty
3d
printed
microfluidic
structures.
A
So
this
is
what
they're
doing
they're
like
imaging
these
embryos
and
then
they're,
using
this
single
chip
c,
cmos
transceiver
with
pre-integrated
micro,
coil
and
assembly,
with
a
3d
printed
micro
channel,
so
they're
able
to
like
get
images
of
this
and
they're
able
to
print
them
out,
and
you
know
they're
using
a
3d
printer.
So
it's
like
a
material
that
they
of
they're
choosing
and
they
print
it
out.
Is
this
like
in
some
sort
of
plastic,
or
is
this
like?
Do
you
know
any
more
about
this
morgan.
A
Yeah,
it's
it's
a
it's
definitely
a
nice
little
paper.
It
looks
like
it's
only
one
page,
it's
like
a
report,
but
they
have
like
they
kind
of
go
through
some
of
this.
They
don't
really
go
through
the
steps
in
detail,
but
they
basically
have
this
embryo
of
the
image.
A
B
Certainly,
I've
been
trying
to
look
at
more,
but
how
microfluidics
is
being
used.
You
know
again,
this
kind
of
relates
to
some
of
the
stem
cell
work
and
in
particular,
kind
of
yeah
of
being
able
to
isolate
or
separate
stem
cells
on
the
basis
of
some
some
function,
and
you
know
again,
it's
like
the
the
kind
of
fabrication
techniques
that
are
available
to
engineers.
These
days,
as
you
know,
requires
a
whole
new
retraining,
yeah.
D
Yeah
bradley
yeah,
I
haven't,
I
mean
an
amusing
con
comment
on
this.
The
you
said
they
make
small
models
of
the
embryos
yeah.
I
did
a
calculation
once
where
is,
if
you
took
a
if
you
took
a
diatom
and
you
could
image
it
at
full
resolution,
okay
and
then
you
try
to
print
it.
It
would
be
50
meters
in
diameter,
diameter.
A
D
These
3d
printers
don't
have
that
great
resolution.
I
once
tried
to
make
a
three-dimensional
globe
with
3d
printer
but
two
millimeters
in
diameter
and
it
was
worthless.
A
B
Okay,
I
need
the
detail.
D
Showing
that
in
terms
of
silica
and
diatoms
there's
structure
at
eight
orders
of
magnitude.
A
Yeah,
so
I
think
they
also
talk
here
about
like
how
they
do
this,
like
they
basically
get
this
reading
it's
hard
to
see
here,
but
then
they
use
that
to
sort
of
like
as
the
detail.
So
it's
not
not.
B
A
An
exact
replica
but
they're
getting
like
a
spectrum
of
the
body,
sections
and
eggs,
so
they're,
looking
at
these
heterogeneities
as
like
a
spectrum
of
some
type
and
then
they're,
just
basically
printing
it
out
yeah.
So
I
don't
know
it's
yeah,
it's
you
don't
get
all
the
resolution,
but
it's
you
know
it's
something
that
we.
B
A
Yeah
now
dick-
and
I
did
a
paper,
I
think,
with
a
few
other
people
on
these
images,
like
one
of
his
students
at
one
point-
had
hand-drawn
amanda,
an
imaginal
disc
from
drosophila,
which
is
like
this
structure.
That
would
become
an
eye.
So
you
know
it
is
like
very
a
very
you
know.
A
It's
a
very
small
thing
and
the
resolution
was
kind
of
played
tricks
at
the
resolution,
because
you
had
all
these
fine
structures
that
were
drawn
out
at
a
very
large
in
a
very
large
size
and
so
digitizing
that
then
kind
of
cheated
in
a
way,
because
you
got
higher
resolution
without
having
to
worry
about
like
the
the
imaging
power,
and
so
that
was
kind
of
an
interesting
exercise,
because
you
know
we
got
this
sort
of
resolution
for
free
someone
just
drew
andrew
it,
and
there
was
a
lot
of
work
that
was
involved
in
that.
D
Yeah
it
took
my
mother,
I
think
10
hours
to
correct
the
the
hand
drawing
oh.
A
But
it
actually
did
serve
as
sort
of
a
prior
and
that
it
helped
us
do
the
sort
of
analysis
that
we
wanted
to
do,
which
is
looking
sort
of
at
the
frequency
of
things
across
the
surface.
So
you
would
have
these
things
that
were
sort
of
defined
at
a
decent
scale.
You
could
like
tell
kind
of
what
you
know.
We
had
a
good
prediction.
I
guess
human
expert
prediction
about
what
they
should
look
like,
and
then
we
could,
just
you
know,
measure
it
out
without
like
obscuring
it
into
very
small
scales.
B
Yeah
that
just
makes
me
think
about
a
recent
recent
paper
I
saw
from
from
mgh
on
anastasia
that
has
a
paper
on
like
using
high
resolution.
B
You
know
fixed
postmortem
brain
data
to
to
improve
you
know
imaging
data
that
you
collect
from
a
from
a
person.
That's
you
know
that
you
don't
get
to
scan
for
a
day
sounds
like
a
similar
kind
of
thing.
A
All
right,
we
have
anything
else.
We
want
to
talk
about
not.
D
Yeah
bradley,
do
you
remember,
recall
how
many
cells
were
in
the
drosophila
imaginal
disc.
A
Oh,
I
think
we
were
able
to
segment
like
eight
thousand,
so
that
was
all
the
things
we
could
segment.
So
you
know
there
were
different
types
of
cells
that
they
were
characterized
by,
like,
I
think
the
volume
and
other
things
but
yeah.
I
think
we
got
like
8
000
out
of
it
hello.
So
I'm
following
up
on
a
few
things
from
today's
meeting.
The
first
thing
I'm
going
to
talk
about
is
on
growth
and
form
by
darcy
thompson.
A
It's
a
great
book
and
we
talked
about
it.
I
think,
a
couple
years
ago,
in
the
meetings
the
100th
anniversary
of
this
book
was
in
2017.
So
it
was
a
book
that
was
written
in
1917
by
darcy
wentworth
thompson
and
he
was
a
I
guess
he
was
the
developmental
biologist
mathematician.
He
was
kind
of
a
polymath
and
he
proposed
that
we
can
take
animal
phenotypes
and
we
can
warp
them
in
different
ways
that
they
sort
of
the
coordinate
system
will
fit
the
phenotype
of
another
organism.
A
He
also
worked
out
some
really
interesting,
math
with
respect
to
nautilus
shells
and
and
other
kinds
of
sea
shells
that
are
have
a
rotational
morphogenesis,
and
so
he
worked
out
some
things
on
that
as
well.
So
let
me
show
you
some
of
the
things
that
we
have
here
and
then
I'm
going
to
talk
about
this
is
paper
and
egg
shapes.
A
This
is
there
was
a
revision,
a
revised
version
in
1942
that
came
out,
and
it
was
a
little
bit
thicker,
but
the
original
book
it's
it
still
had
the
spirit
of
the
original
book,
and
this
is
a
nautilus
shell,
of
course,
and
you
can
use
things
like
the
golden
mean
to
characterize
this:
the
growth
of
the
structure,
if
you've
ever
heard
of
phylotaxis-
that's
another
thing:
they
characterized
with
mathematical
rules
with
number
sequences
and
so
forth.
A
B
A
A
Things
like
architecture
and
other
things
I
don't
mean
like
intelligent
design,
I
mean
you
know,
design
design
and
in
developmental
biology.
You
know
the
development
of
biology.
Its
use
has
been
a
little
bit,
not
not
necessarily
at
the
forefront
of
the
field
in
in
artificial
life
as
well.
There
were
a
couple
papers.
Several
papers
have
been
written
from
time
to
time.
A
On
this
topic
there
there's
a
book
called
on
growth
form
in
computers,
which
is
where
people
are
looking
at,
how
to
model
this
sort
of
on
growth
and
form
thing
using
rules
using
other
things,
so
it's
very
attractive
to
an
artificial
life
audience.
A
Peter
bentley
is
one
person
who
is
involved
in
that.
So
you
know
you
have
digital
artists
and
you
have
people
interested
in
development,
and
so
that's
where
that
that
has
lived
on
in
that
those
spaces.
A
A
So
there's
an
operculum
which
is
about
here
on
the
fish.
There
might
be
a
an
intersection
of
lines.
So
that's
a
grid
point
there's
another
grid
point
down
here
at
the
caudal
at
the
beginning
of
the
caudal
fin
there's
another,
maybe
a
landmark
here
at
the
middle
of
the
dorsal
fin
and
so
forth.
So
you
have
these
different
landmarks
on
the
fish,
then
what
you
can
do
is
you
can
take
this
square
coordinate
system
and
you
can
say
okay.
A
I
want
to
take
this
coordinate
system,
but
I
want
to
warp
the
coordinate
system
so
that
it
fits
this
other
fish
that
I
found.
So
this
is
a
different
species
of
fish.
It
has
the
same
landmarks
but
they're
in
different
proportions
to
the
original
fish.
So
you
work
the
grid
and
then
you
measure
that
that
warping
of
the
grid-
and
you
evaluate
it
in
some
sort
of
in
maybe
you
know
some
sort
of
math.
You
can
use
some
sort
of
math.
A
You
can
use
geometry,
you
can
use
topology
whatever
you
want,
and
so
this
this
works
for
a
number
of
species
of
fish.
Just
you
know,
all
you
need
are
the
same
landmarks
on
each
on
each
species
and
you
can
go
across
species
and
characterize
these,
so
this
is
a
square
grid.
This
is
a
warped
grid
to
the
to
the
front.
It's
it's
elongated
at
the
front
and
shrunken
now
at
the
back.
A
A
The
caudal
fin
is,
is
shortened
relative
to
the
operculum
and
there's
this
it
basically
characterizes
geometric
or
phenotypic
space
in
some
sort
of
geometry.
A
A
Okay,
so
that's
what
I'm
going
to
talk
about
with
respect
on
growth
and
form?
I
think
if
you
read
the
book,
you'll
enjoy
it.
It's
a
very
thick
book.
It's
like
1200
pages
or
something
and
an
interesting
historical
note
about
darcy
thompson
is
that
he
wrote
his
book
in
1917
and
if
you
know
here
history,
you
know
that
darwin
wrote
his
book
in
1859
and
you
know
we
had
people.
We
had
the
german
embryologists
who
didn't
really
want
to
embrace
evolution.
A
You
know
fully
so
they
they
were
doing
a
lot
of
developmental
biology
in
the
absence
of
evolution.
Eventually
they
came
on
board
and
that
became
part
of
developmental
biology.
You
also
had
mendel,
who
was
working
on
his
heredity
experiments
with
peas,
and
that
was
also
in
like
the
1880s.
So
this
is
well
before
darcy
thompson
wrote
this
book
what's
interesting
about
this
book.
Is
he
talks
about
phenotypes,
but
he
doesn't
really
get
into
darwin
and
he
doesn't
get
into
mendel
and
he
doesn't
really
get
into
any
of
this
stuff.
A
That's
going
on
around
him,
and
I
don't
know
why
that
is.
The
development
of
early
developmental
biologists
are
somewhat
resistant
to
evolutionary
theory
and
to
heredity.
Maybe
it's
because
it's
not
something
that's
immediately
applicable
to
what
they're
doing,
although
evolution
certainly
is,
but
it's
an
interesting
sign.
A
Okay,
then
there's
this
new
paper
in
egg
shapes,
and
this
is
a
really
nice-
it's
a
what
it
is.
It's
a
paper
on
looking
at
the
diversity
of
egg
morphologies
and
then
there's
even
a
data
set
that
you
can
download.
So
this
paper
is
by
seth.
Donahue
who's
actually
has
a
decent
presence
on
twitter.
He
works
at
the
university
of
chicago
and
he's
still
the
university
of
chicago
and
yeah,
and
so
this
is
a
paper
on
insect
egg,
morphology,
evolution,
development
and
ecology.
A
It
is
the
single
cell
developmental
stage,
a
resource
investment
in
the
next
generation,
an
unusually
large
and
complex
cell
type,
and
the
protective
vessel
for
embryonic
development
and
a
professor
once
described
a
seed
or
which
is
the
plant
analog
of
a
egg,
is
a
baby,
but
in
a
box
with
its
lunch,
which
means
that
it's
a
very
small
organism,
it
has
some
sort
of
nutrients
inside
and
it's
in
a
box,
a
protective
box.
So
that's
pretty
consistent.
A
I
in
this
review
I
described
the
more
morphological
diversity
of
insect
eggs
and
then
identify
recent
advances
in
understanding
the
patterns
of
egg
evolution,
the
cellular
mechanisms
underlying
egg
development
and
the
notable
aspects
of
egg
ecology.
I
also
suggest
areas
for
particularly
promising
future
research
on
insect
egg
morphology.
A
These
topics
touch
upon
diverse
areas
such
as
tissue,
morphogenesis,
life,
history,
evolution,
organismal,
scaling,
cellular
secretion
and
lv
position,
ecology
and
very
integrative
approach
now
we're
in
cellular
secretion.
I
think
it's
kind
of
interesting
when
I
was
doing
some
work
on
induced
pluripotent
cells
and
on
stem
cells.
A
The
head
of
the
lab
asked
me
if
I
could
do
some
work
on
what
they
called
the
secreto,
which
is
where
cells
get
secrete
a
lot
of
chemical
compounds
into
their
immediate
environment,
and
he
wanted
to
see
if
we
could
characterize
these
as
sequence,
data
and
other
things
and
do
some
bioinformatics
and
as
it
never
really
got
off
the
ground
because,
as
it
turns
out,
this
is
not.
They
do
have
tools
for
this
and
they
do
have
some
data
sets.
A
But
it's
really
a
understudied
area
of
biology,
and
I
just
I
find
it
interesting
that
I
see
some
references
to
it
recently
and
I'm
just
interested
in
what
people
are
thinking.
It's
a
very
interesting
thing
because
we're
dealing
with
a
lot
of
signaling
in
development
and,
of
course,
that
would
be
a
very
nice
tool
to
have
if
people
really
had
a
good
set
of
tools
for
secretion,
assays
and
secretomes.
A
So
so
this
kind
of
goes
through
a
quote
from
jan
swarmerdam
who's,
a
17th
century
naturalist
about
insect
eggs,
and
so
he
has
this
collection
of
insect
eggs.
He
talks
about
them,
he's
kind
of
fascinated
by
them
and
he
kind
of
describes
where
he's
finding
them
so
they're
kind
of
you
know
doing
they
have
this
box
of
eggs,
basically
that
they're
offering
us
and
it's
all
these
different
types
of
eggs
and
data
and
measurements
and
other
things.
A
A
So
you
can
see
that
you
have
all
these
different
species
and
the
phytogeny
that's
defines
their
evolution,
and
then
you
have
these
color-coded
and
you
see
that
in
polynomial
you
have
some
generally
larger
eggs
hymenoptera
you
which
are
ants.
You
have
some
smaller
eggs
and
some
larger
eggs,
polynomial
crickets,
I
guess-
and
then
you
know
you
have
some
variation
across
the
different
clades
that
you
have
here.
So
you
have
apterigota,
which
is.
I
don't
know
what
that
is,
but
that's
a
moderate
sized
egg
anyways.
A
You
can
see
that
egg
size
isn't
necessarily
awake.
It's
not
really
derived
in
any
one
clade,
it's
it's
what
they
call
poly
poly!
Well,
it's
it
appears
many
times
in
the
tree
so
independently
and
it's
hard
to
say
whether
there's
a
common
ancestor
for
any
one
egg
size.
A
lot
of
this
is
ecological,
so
ecological
pressures
determine
egg
size,
but
sometimes
you
know
you
see
some
clades
here
or
some
subclades
that
have
like
that
are
they
have
smaller
eggs
and
but
of
course
you
have
smaller
eggs
over
here
too.
A
So
it's
polyphyletic
in
the
sense
that
it's
not
there
isn't
one
origin
point.
I
guess
you
could
argue
back
here
as
an
origin
point,
but
then
you
have
different
sized
eggs
that
are
descendants
of
that
as
well.
So
it's
polyphonic
definitely
like.
I
said,
though
it's
yeah,
so
he
says
that
these
these
clades
defining
these
organisms
are
monophyletic.
A
A
You
can
actually
approach
this
in
different
species
and
see
what
is
happening
in
the
you
know.
If
you
can
knock
out
certain
developmental
genes,
you
can
actually
look
at
developmental
conditions
and
you
can
moderate
the
excise
in
that
way.
So
here
they
did.
Let's
see
they
did
an
experiment
with
drosophila
melanogaster,
which
is
the
fruit
fly
in
which
researchers
artificially
selected
independent
lab
populations
for
unusually
small
and
large
eggs.
So
they
did
some
evolutionary
experimental
evolution
here.
A
genetically
mixed
pool
of
flies
was
separated
into
multiple
populations.
A
A
So
these
are
these
three
variables
that
you're
switching
off
and
you're
actually
looking
at
how
they
fit
together,
and
so
these
are
the
mathematical
calculations
here,
along
with
with
and
then
you
know,
describes
what
an
aspect
ratio
of
an
egg
is,
and
so
we
can.
We
can
look
at
these
patterns
once
we
have
them
measured.
We
can
characterize
them
and
if
I
imagine
here,
we
can
characterize
them
in
terms
of
their
ecology,
or
we
can
do
things
like
experimental
evolution
and
then
measure
the
outcome
of
those
experiments.
A
So
there's
a
lot
of
elaboration
on
egg
shape.
It
can
be
running
this
word
operculum,
which
is
actually
just
a
generic
term
for
some
sort
of
feature
on
a
phenotype
but
yeah,
so
their
egg
shape
is
defined
by
a
number
in
a
number
of
ways.
They
have
different
aspects
of
the
egg
and
you
can
look
at
them
and
use
those
as
features,
and
you
know
well.
A
Egg
physiology
is
interesting
because
eggs
are
actually
this
permeable
boundary
between
the
environment
and
the
embryo,
and
so
eggs
really
serve
this
purpose
of
this
box,
and
so
the
exchange
of
gas
and
water
across
the
egg.
Shell,
at
the
proper
rate,
is
crucial
for
embryo
development.
Within
the
egg.
An
egg
shell's
morphology
can
include
a
complex
internal
structure
in
pores
such
as
hydropiles
and
aereo
piles.
Collectively
these
determine
the
respiratory
treats
of
an
egg.
So
a
hydropile
is
where
water
is
exchanged.
An
aeropyl
is
where
air
oxygen
is
exchanged.
A
In
some
insects
there
is
a
a
period
of
embryogenesis
during
which
the
egg
absorbs
water
from
its
environment.
Increasing
the
an
overall
size,
the
movement
of
water
across
the
eggshell
can
be
further
modified
by
the
developing
embryo
itself,
which
secret,
when
it
secretes
an
additional
inner
layer
to
the
egg.
Shell
called
the
cereal
cuticle,
so
they're,
you
know
in
embryos,
they're
secreting
things
that
are
building
the
eggshell
they're
secreting
things
out
into
that
local
environment.
A
So
they're,
not
just
secreting
things
that
communicate
with
other
cells,
they're,
actually
building
this
this
box
on
their
own,
and
so
this
is.
This
is
interesting
because
actually
there
are
things
that
modify
this
ability
to
secrete
an
egg,
and
we
know
that,
like
weak
egg
shells
can
result
in
vulnerabilities
for
the
embryo.
So
it's
recently
been
shown,
for
instance,
that
in
several
disease,
vector
mosquitoes,
the
chemical
composition
of
the
serosal
cuticle
can
have
a
dramatic
effect
on
resistance
to
desiccation.
A
A
A
This
is
kind
of
describing
the
data
set.
It
just
describes
this
these
thousands
of
eggs
that
have
been
characterized
and
measured,
and
so
it
goes
through
this.
This
is
just
kind
of
describing
the
data.
So
if
you
want
to
use
the
data,
go
to
this
paper
in
scientific
data
go
over
the
sort
of
the
metadata
that
they
give
you
here
and
then
you
can
download
it.
I
think
it's
on
fixture
actually,
and
so
this
is
the
zip
file,
and
then
it
put
it
into
a
an
excel
sheet
which
is
kind
of
hard
to
read.
A
I
just
wanted
to
dump
it
in
there
to
show
how
much
data
you
have
so
I
think,
actually
I
have
it
in
columns
here.
I
don't
okay,
so
these
are
just
like
this
is
the
id,
and
this
is
the
names
you
have.
You
know
you
have
all
these
references
from
the
literature
and
the
references
are
listed
here,
there's
an
id
and
then
you
have
some
information
on
this
side,
so
you
have
a
lot
of
different
potential
measurements
that
they've
made
on
these
eggs.
So
this
is
like
a
literature
review.
A
It's
like
going
through
the
literature
finding
these
eggs,
finding
the
information
about
measurement
and
it's
all
put
in
one
data
set
for
people
to
use.
This
is
just
kind
of
a
dump
of
data,
but
you
can
like
work
on
work
out.
The
data
put
it
into
whatever
format
you
want,
and
you
know
it.
You
know
there
there's
a
lot
of
missing
data,
but
there's
also
a
lot
of
useful
data
in
here
as
well.
A
So
that's
all
I
have
to
say
about
those
things,
the
the
on
growth
and
forum
and
the
egg
shape
data
set.
I
think
those
are
two
complementary
things
by
the
way.
So
you
know
you
could
like
model
eggs
from
the
data
that
they
give
and
then
you
know,
do
these
transformations
and
see
what
kind
of
interesting
things
happen
or
you
could
just
do
it
on
some
other
phenotypic
data
set
that
yet
to
be
named,
but
I
wanted
to
go
over
that
for
people.
A
Thank
you
yeah,
so
well,
thank
you
for
meeting
today
and
we
we
can
continue
the
conversation
offline
and
see
you
next
week.