►
From YouTube: DevoWorm Summer of Code weekly meeting, 8-9
Description
Meeting for Week 14. Asmit Singh, Ujjwal Singh, Paras Mehan, and Vinay Varma. Update on ML for DevoWorm and Digital Bacillaria
A
C
A
A
B
D
So
we
have
started
like
we
have
football
team
versus
Mike,
how
many
and
Bob
said
Mike?
How
many
does
little
things
like
that
and
we
burn
in
the
fire.
We
have
brought
that
cloud.
We
have
taken
like
approximately
93
to
95
fields,
which
acontecimiento
we
have
tried
to
extract
the
position
of.
As
you
know,
in
these
fields,
through
the
coordinates,
which
is
our
best
axis,
and
we
are
trying
to
figure
it
out
like
how
they
are
changing.
We
are
not
you
and
Dad
fighting
as
well
and
on
the
documentation.
We
have
static
documentation.
D
D
D
B
He
wanted
to
have
like
a
common
Google
Doc
further
on
to
contribute
to
he
likes
to
work
with
Google
Docs
and
it's
probably
better
for
writing
papers,
although
I
also
like
if
you
would
get
hub,
but
you
know
I
think
in
this
case,
Google
Doc
it
better
to
collaborate
on
as
you
can
do,
a
lot
of
things
of
formatting
in
Google
Doc,
which
is
you
know
pretty
advantageous
when
you
get
down
the
line
or
like
you
know,
references
and
things.
If
I
you
know
what
you
could
do
is
just
like.
B
You
could
work
directly
in
the
Google
pub
there
in
the
Google
Doc
that
you
shared
you
could
just
put
in
what
you
have
now.
If
you
look
at
the
Google
Doc,
you
know
I've
just
made
it
so
that
there
are
a
lot
of
empty
spaces,
and
so
you
know
you
could
put
it
in
I.
Guess
put
some
in
the
methods
and
put
some
in
the
results
and
then
I
mean
we'll
go
through
it
maybe
next
week
and
then
we'll
figure
out
because
things
will
get
shifted
around
anyways.
B
The
reason
being
is
because
and
I
want
a
little
bit
more
interactivity
between
like
myself
and
Richard,
and
it
does
help
that
if
you
have
like
someone
like
Richard
who's,
you
know
biologist
by
training,
and
you
know
so
physicist,
but
you
know
he
has
like
a
lot
of
expertise
in
vast
area
for
him
to
read
through
it
and
say
you
know,
I,
don't
understand
this
thing
about
machine
learning,
so
I
think
the
better,
a
big
beneficial
to
have
it
like
in
the
public
and
then
in
terms
of
your
results.
They
sound
good.
D
D
C
B
B
E
F
B
A
E
B
B
D
B
Okay,
well,
yeah,
give
me
some
updates
on
that.
As
you're
going
along,
you
know,
I
mean,
even
if
you
don't
have
any
questions,
just
is
kind
of
like
running
things
by
me
and
then
I
can
you
know
if
you'll
get
dick
in
the
loop
on
it?
If,
if
it's
something
that
sounds
like
it's
a
problem,
something
it
sounds
anything
add
to
it,.
D
So
so
like
we
have
not,
we
are
planning
to
do
is
like
we
have
to
fill
coordinates
of
massive
area,
in
fact,
either
like
xmin
xmax,
ymin
ymax,
and
we
are
trying
to
see
that
out
like
have
their
menu
select.
We
are
various
we
are
trying
to
find
the
difference
between
these
four
coordinates
in
the
93
continuous
trail,
so
that
we
have
the
displacement
like
something
after
the
capture
displacement
in
these
fields
and
I
think
like
we
can
calculate
the
velocity
and
all
the
strain.
B
It
stretching
outward
yeah
I,
think
that's
part
of
the
movement,
though
I
mean
like
they're
gonna,
be
like
different
components
of
the
movement.
So
it's
not
just
gonna,
be
like
maybe
a
smooth
path.
There
might
be
some
like.
You
know,
moving
in
one
direction,
but
it's
kind
of
drifting
does
remember.
We
have
the
issue
with
them.
You
know
they're.
Of
course
the
videos
aren't
like
excellent
quality.
I
mean
they're,
not
like
real
easy
to
track,
but
then
you
have
like
you
know.
The
thing
is,
actually
you
know
pulsating
and
kinda
like
moving
a
little
bit.
B
D
B
Yeah
I
mean
you
know
you
could
break
it
into
different
components.
Like
you
could
say,
this
is
like
the
the
wideness
and
then
that
you
know
we
could
figure
out
what
that
is
later
and
then
you
could
figure
out
like
a
trajectory
which
is
like
the
main
component
on
the
movement.
So
they're
gonna
be
different,
like
they're
gonna
be
different
parts
of
it.
B
I
mean
that's
yeah
I,
just
we
should
have
more
I
think
she
hasn't
been
more
like
in
slack.
You
know
whatever
just
to
make
sure
that
we
know
and
I'm
like
we're
on
the
same
page
about
it.
It
helps
us
all
think
about
it
more
and
then
we
have
a
description
of
neural
networks
for
perch
or
biologists.
That's
still
standing.
That's.
B
And
than
that
yeah
or
well
I
guess:
we've
talked
about
that
a
little
bit
hasn't
it
in
as
well.
So
you
could
address
that
when
you're
ready,
the
template
methods,
I
think
that's
what
we're
working
on
with
the
methods
like
I
said
that
that
can
go
in
the
Google
Doc
and
we
can
get
that
into
shape
their
documentation
of
procedure.
The
same
thing
you
know
might
take
a
couple
of
rounds
of
edits
to
get
it
really
nice
and
readable,
then
reviews
the
Butley.
B
A
A
B
B
B
Yeah,
it
would
be
like
that
fifteen
to
twenty
minutes,
and
then
you
know
we'll
do
the
15
or
20
minutes
and
sly
in
PowerPoint,
slides
and
then
people
would
ask
questions
about
the
people
about
you
know.
I,
don't
know
if
you
can
answer
them,
because
sometimes
they're
very
opaque
in
the
paper
that
how
they
do
things,
but
we
can
discuss
it.
A
B
A
B
B
B
B
Yeah-
and
it's
just
a
matter
of
like
you
know,
maybe
sharing
plots
and
like
showing
how
the
debt
look
at
data.
Look
like
you're
talking
about
some
of
these
things
that
you're
having
a
little
bit
of
problem
with
just
putting
that
out
and
like
I'll,
give
me
my
feedback
on.
So
it's
that's
good,
we'll
put
that
in
in
progress.
B
B
Ok
and
then
ok,
so
that's
done.
Let's
see
model
features.
Looking
in-between
frames,
we're
kind
of
doing
that
now,
that's
kind
of
in
progress,
pretty
video
resources
for
vessel
area,
biology,
that's
vs9
thing
and
then
define
features
with
end
in-between
frames.
That's
kind
of
what
we're
doing
is
well
not
with
the
in
max,
so
so
yeah
I
look
forward
to
seeing
your
stuff
on
slack
and
we
can
discuss
up
there.
This
is
becoming
weak
and
now
are
there
any
issues
we
need
to
put
on
the
board.
D
B
E
B
B
B
E
So
I
was
talking
about
this,
like
our
chat
in
the
last
week
so
like
when
we
spoke
about
the
in
like
how
the
software
should
be
like
what
will
be
the
inputs
and
what
my
work
will
be
and
what's
the
desirable
output
and
how
a
user
can
interact
with
our
software.
So,
regarding
this,
regarding
this
conversation,
yeah
I
have
a
little
doubt
so
you
are
referring
to
inputs,
would
be
an
image
and
perhaps
a
metadata
file.
Do
you
mean
by
metadata
file
the
CSV
file?
Well,.
E
So
I
want
to
repeat
what
I
understood
like,
so
you
can
point
out
a
file
misunderstood
something
okay,
and
so
our
software
will
be
a
web
app
where
in
the
server
will
be
having
a
CSV
file
like
like
the
data
file.
Like
you
gave
me
like,
we
discussed
about
what
the
CSV
file
to
have,
which
has
the
XY
coordinates
and
I
present
coordinates,
and
the
cell
names
and
under
timing
at
which
the
cell
is
born.
Yeah.
B
E
Are
you
a
new
user
comes
and
who
actually
uploads
a
video,
since
this
is
the
C
elegans
like
they?
They
suppose
the
time
at
which
the
cell
bonds
is
roughly
at
the
same
time.
So
without
reference
to
CSV
data
we
can
try
and
label.
We
can
try
and
add
the
video
which
the
user
has
a
prototype.
Yes,
so,
oh
so
like
I
suppose,
let's
say
we
have
another
reference
data.
Let's
say
we
have.
A
B
cell
will
be
born
at
depth,
42nd
minute
yeah
so
over.
E
We
can
split
the
42nd
frame
or
Achtung
some
strings
in
batch
the
time
frame,
and
we
can
try
to
predict
where
the
baby's
a
butyl
will
be
in
the
new
user
and
you
can
Marge.
You
can
merge
that
cell
and
you
can.
You
can
again
put
it
back
into
the
video
and
and
let
me
use
the
download
the
video
right
yeah.
B
D
B
Identified
with
their
centroids
and
their
sizes
like
you
could
have
you
know,
you
would
upload
a
video
and
then
you
download,
like
an
analysis,
we're
just
building
the
CSV
file.
So
it
would
be
like
that
one
column
would
be
the
name
of
a
cell.
Then
the
prediction
of
what,
where
that
cell
is
in
the
image,
so
it
actually
be
you're
just
taking
like
an
image
and
you
take
each
cell,
you
segment
the
thing
and
then
using
the
data
and
the
CSV
files
that
we
have.
B
You
would
predict
what
each
blob
is
in
name
of
the
cell
based
on
the
location.
It
would
be
a
prediction
and
then
you
would
have
this.
You
could
report
the
size
of
the
blob,
the
size
of
the
cell,
that
you're
predicting
and
then
the
centroid
position
of
that
prediction.
So
why
does
think
it
would
be
a
prediction?
So
people
wouldn't
think
that
it
was
like.
You
know,
an
absolute
answer.
You
know
cuz,
that's
the
way
they
do
it
and
they
like
bioinformatics.
They
make
the
prediction
based
on.
E
Yeah
sure
so
now,
like
now
I
think
I
had
an
I
have
an
overall
idea
of
what
this
like,
where
this
project
is
going.
So
everything
seems
fine.
I
think
it
will
be
great
if
we
can
do
this,
but
I'm
only
concerned
about
the
segmentation
part,
because
the
segmentation
that
we
did
like
I'm
bit
concerned
about
how
we
can
get
the
centroids
and
how
accurate
would
would
that
be.
E
E
B
B
That
works
a
lot
better.
So,
like
you
know
again,
it's
like
the
the
bioinformatics
example.
You
know
you
have
what
they
do
in
a
lot
of
like
when
they
predict
gene
sequences.
They'll
use
a
couple
of
different
options
for
like
algorithms
and
then
they'll
put
the
sequence
in
and
they'll
do
a
match
and
then
they'll
get
a
certain
like
you
know,
accuracy
rate
and
maybe
report
the
accuracy
rate
of
the
match.
B
You
know
that's
possible
or
some
way
to
to
know
whether
you
know
that's
a
really
good
fit
to
the
data,
but
then
they
might
use
another
route,
so
I
mean
you
know
if
you've
looked
at
the
digital
vessel
area
stuff,
there's
these
barrier
these.
You
know
these
microorganisms
that
swim
around
and
they
move
around
a
lot.
And
so
you
know
one
elder.
That
might
be
great
at
like
capturing
some
one
aspect
of
the
cells,
but
maybe
not
so
great
a
caption,
another
askin,
and
then
it
might
be
a
bit
different.
B
So
that's
a
problem,
maybe
in
some
of
the
input,
videos
or
uploading
things
that
aren't
like
super.
You
know
controlled.
You
know
in
their
image
because
people
who
are
images-
sometimes
you
know
they
don't
mouth
the
specimen
as
well
as
like
some
of
the
videos
that
we've
been
working
with,
because
those
are
from
applications.
B
E
Yeah,
yes,
so
for
what
I
understand
from
that?
Is
that,
like
from
the
segmented
from
the
segmentation
part
I,
think
we
can
work
on
it
a
bit
and
we
can
adjust
it
neatly
so
that
we
get
this
encouraged
as
I
create
as
possible,
because
the
tagging
of
the
cell
mainly
depends
on
the
centroids
that
we
get,
that
we
get
from
the
segment
segmentation.
So
reducing
that
greatly
reduces
the
overall
error.
E
I
I
think
I
think
that
greatly
influences
the
overall
order
and
so
I
think
we
can
work
a
bit
more
on
the
segmentation
part
and
ensure
that
each
cell
is
neatly
divided.
Instead
of
a
noisy
thing
that
we
are
getting
right
now.
So
once
once
we
get
that
and
and
then,
if
we
are
able
to
like
get
some
clear
boundaries
and
then,
if
you
can
can
calculate
the
center
of
each
of
the
cells
on
the
segmentation
maps,
then
I
think
we
can.
E
B
B
E
I'll
think
about
that
measure
so
currently
I
think
what's
urgent
is
to
work
on
the
like
refining
the
segmentation,
and
once
we
are
aware,
we
were
satisfied
with
the
segmentation
and
then
work
with
the
center
calculation
of
centroids
and
then
it's
an
easy
I
think
after
that
it
it's
a
bit
easier
to
do,
because
it
suggests
the
process
manner
match
between
this
reference
points
and
for
them
and
for
the
metric
on
how
on
how
good
this
like,
which
helps
you
estimate
how
good
the
software
is.
We
can
just
verify
on
several
new.
C
B
E
The
complete,
like
I'll,
do
that
I've
divided
the
documentation
into
two
parts
at
one
is
for
the
users,
and
one
is
for
the
developers,
like
the
user.
Section
mainly
focuses
on
like
it's
an
abstract
and
how
honest
and
a
software
works.
I
am
giving
steps
on
how
to
get
get
the
jar
file
and
how
to
install
the
image
J
and
get,
and
what
am
I
steps
to
basically
get
that
working
to
happen.
For
the
developer.
E
B
E
E
I
think
yeah
I
think
what
we
try
to
avoid
in
the
beginning
that
came
to
be
is
the
most
important
one,
because
the
time
Ferran
has
the
scheduling
time
for
enhancement.
I've
asked
it
before
because
of
this,
so
this
only
like
I
felt
that
pigmentation
I
can
maybe
we
can.
We
can
put
some
more
time
on
it
and
see
if
we
are
able
to
get
get
it
done.
I
like
get
it
more
refined,
so
I
think
we're
back
to
it
again,
and
this
is
our
main
issue
right
now,.
B
E
B
B
E
C
E
I
have
something
to
tell
about
it.
So
in
the
power
like
in
the
past
two
days
or
three
days,
so
I
will
receive
their
mail
from
Mallon
ma'am,
asking
all
the
district
students
to
tell
her.
If
we
are
interested
in
doing
an
outreach
like
we
have
a
scholarship
like
$500
scholarship,
for
students
who
like
to
go
to
conferences
or
to
go
to
travel
or
like
travel
to
conferences
for
the
sake
of
community
outreach
like
presenting
our
work
or
something
like
that.
E
E
The
word
of
a
word
about
new
and
like
intersection
between
the
neuroscience
computer
vision
and
also
bioinformatics
so,
and
we
can
so
I
was
thinking
about
that
but
like
like
in
the
past
week
like
we
had
been
busy
with
all
like
so
many
stuff
is
half
so
much
stuff
is
happening
right
now,
yeah,
but
we're
also
so
I
wasn't
able
to
write
that
right
and
the
proposal
so
I
think
I'm.
Thinking
of
writing
a
proposal
in
two
or
three
days
and
I'll
send
her
and
and
see
what
what
will
happen.