►
Description
Attendees: Bradly Alicea, Mayukh Deb, Ujjwal Singh, Susan Crawford-Young, Jesse Parent, and Krishna Katyal. Preview of DevoLearn ML/DL platform, umbrella website, and Axolotl montaging.
A
A
C
A
A
A
D
A
E
D
B
F
D
Welcome
to
the
meeting,
I
think
we
have
enough
to
start
the
meeting
if
other
people
come
in
we'll
just
let
them
in
and
the
first
part
of
the
meeting
I
wanted
to
spend
on
the
new
devil
or
evil
learn
implementation
and
everything
else.
So
I
kind
of
wanted
to
go
over
some
things
before
we
started
with
updates.
D
So
this
is
foreign
and
people
who
weren't
in
the
meeting.
I
wanted
to
go
over
the
naming
process
and
everything,
so
I
put
up
a
a
survey
in
slack
and
in
our
weekly
email
and
I
got
a
couple
of
responses
and
then
a
couple
of
responses
offline.
So
this
is
the
sort
of
the
voting
pattern
here.
It
gave
a
couple
options.
D
I
have
d
baller
and
I
kind
of
threw
that
in
and
then
some
other
ones-
and
it
seems
like
fifty
percent
of
people
like
vivo,
learn
some
people
like
viva
or
may
I
some
people
like
to
open
devo
cell,
so
that's
good.
That
was
something
that
was
sort
of
a
consensus
there
and
then
I
had
a
couple
other
people
saying
that
evil
learn
wasn't
a
nice
name,
and
so
I
decided
to
go
with
that.
I
just
thought
yeah
that
would
be
kind
of
catchy,
and
so
we
posted
it.
D
D
D
D
B
D
D
This
is
the
organization
here,
it's
github.com,
diva
learn,
and
so
we
have
people
you
can
join
to
be
well.
I
have
like
five
contributors
right
now.
If
it
makes
sense
for
you
to
join,
let
me
know
and
I'll
I'll
invite
you
to
the
group.
Basically,
this
is
just
for
people
to
work
on
the
code
underlying
this
application
and
so
we're
doing
this
from
github.
Well,
we're
doing
it
from
github
and
pip,
which
we'll
talk
about
in
a
minute.
D
F
D
I'll
get
to
that
in
a
minute,
so
the
contribution
guidelines
are
here,
and
this
is
just
if
you
want
to
contribute
to
the
to
the
project
so
contributing
to
diva
learn,
and
this
is
something
that
usually
have
just
come
up
with
this
week.
So
you
know
we
might
make
changes
to
this
as
we
go
through.
You
know
taught
over
time.
It
was
just
something
that
they
got
from
a
template,
so
I
think
my
hook
said
that
they
got
this
from
a
template.
B
B
D
D
D
D
You
know
this
anything
related
to
diva
learn
like
kind
of
working
on
the
platform,
or
maybe
some
sort
of
you
know,
machine
learning,
initiative
that
we
want
to
grow
from
this.
So
there's
going
to
be.
You
know
that
sort
of
conversation
going
on
on
that
channel
and
then
you
can
contact
any
one
of
the
three
of
us
so
we're
pretty
active
on
github.
D
If
you
have
a
question
or
you
want
to
make
a
contribution
like
a
pull
request,
you
know
make
a
pull
request
and
then
maybe
ping,
one
of
us
and
we
can,
one
of
us
will
probably
be
able
to
merge
that.
And
so
that's
that's.
If
you
want
to
contribute
and
then
what
do
I?
What
should
I
know
before
I
get
started
so
it's
always
good
to
before
you
just
start
contributing
to
a
project,
know
kind
of
what
the
aims
are.
D
I've
never
rejected
a
pull
request
necessarily,
but
if
you
think
about
it,
you
know
there's
a
good
chance
that
that
will
happen
at
some
point
if
you
contribute
to
an
open
source
project.
So
what
are
the
diva
learn?
Models
do
so.
The
models
that
devil
learn
can
be
used
to
collect
various
types
of
useful
data
from
videos,
images
of
embryos
which
can
be
used
for
statistical
analysis
or,
if
you're,
feeling
crafty,
you
can
use
the
metadata
for
another
deep
learning
model.
D
So
we
were
talking
about
pre-trained
models,
and
this
is
part
of
what
we're
doing
here,
but
you
can
also
you
know
we
can
reuse
what
we've
got
in
in
terms
of
data
and
in
terms
of
code
in
different
ways.
So
if
we
have
an
idea
of
something
we
want
to
recombine
that's
going
to
be,
you
know
this
is
going
to
be
where
this
platform
becomes
fruitful
at
naval
learn.
D
We
want
to
create
a
collection
of
robust,
deep
learning
models
which
can
be
used
very
easily
for
research
and
analysis
of
various
biological
processes,
so
the
models
are
all
based
on
pie.
Torch
we'd,
prefer
pie
torch
to
be
used
for
your
model
just
because
of
interoperability,
but
I
think
you
know
we
can
expand
to
beyond
pie
torch,
but
we
want
to
make
sure
that
we
keep
everything
it
might
make
sense
to.
F
D
F
A
A
D
D
D
F
F
B
F
F
F
A
F
F
F
A
A
F
D
So
I
mean
that's
yeah,
that's
something
we
can
do
in
the
future.
I
was
it
like.
I
said,
there's
also
another
project,
the
spa
that
we'd
like
to
do
on
documentation.
So
that's
almost
a
separate
thing
for
the
long
term,
but
but
this
is
very
good.
I
think
the
idea
and
then,
if
you
have
other
ideas
on
how
to
improve
the
documentation,
that
would
be
good.
D
D
D
I'm
not
sure,
what's
in
here
right
now,
but
yeah
you
can
put
them
in.
You
know,
put
them
in
the
regular
way,
just
any
sort
of
yeah,
slides
or
images
or
anything
that
you
want
to
put
in
here.
You
can
put
them
in
here,
so
that's
the
diva
learn
organization
and
that's
that'll
be
growing.
You
know
and
we'll
be
revising
it
as
we
go,
but
I'm
gonna
then
go
to
the
pipey.
D
D
You
can
view
the
release
history
and
then
the
description
here,
which
is
where
we
have
our
you
know,
sort
of
a
summary
of
the
github
project
repository
or
the
organizational
repository,
and
this
is
the
readme
from
one
of
the
repositories,
one
of
the
readme's
in
the
in
one
of
the
repositories.
So
this
is
basically
it
looks
like
my
oaks
notebook
that
he
was
working
on
and
just
kind
of
put
into
this
read
me.
D
A
notebook
so
then
you
go
so
basically
just
walks.
You,
through
the
segmentation
of
the
c
elegans
embryo
importing
the
model
running
the
model
on
an
image,
reviewing
the
prediction
and
finding
the
centroids
and
then
generating
synthetic
images.
So
you
put
this
up
with
a
pre-trained
gan,
so
this
is
something
I
think
we
talked
about.
D
I
think
last
week
where
he
had
built
a
gan
that
will
sort
of
generate
a
bunch
of
synthetic
embryos,
and
then
you
look
at
the
ones
that
are
where
the
error
is
minimized
and
you
get
this
embryo,
and
so
that's
that's
in
there
as
well,
then
predicting
populations
of
cells.
So
this
is
the
graph
of
the
different
families
of
cells
or
the
sub
some
sub
lineages
in
the
embryo
and
the
graph
here.
D
This
is
an
animation
of
it,
but
basically
it
allows
you
to
do
that,
making
predictions
from
a
video
and
saving
the
predictions
in
a
csv
file.
So
all
this
can
be
exported
as
a
csv
file,
and
then
we
have
statistics.
We
have
the
license
and
mit
license.
That's
good,
I
think
that's
suitable
for
software.
I
know
we
use
the
creative
commons
license.
More
generally,
but
I
think
for
software
we
can
use
the
mit
license,
that's
what
they
do
in
open
worm
for
a
lot
of
software
applications.
D
A
Yeah
so,
first
of
all
that
looks
very
good.
Actually
the
ppi
by
redmi
file
because
like
when
there's
a
demo,
I
think
more
people
will
be.
It
will
be
more
interesting
for
people
who
just
look
at
a
look
at
it
for
a
short
period
of
time,
and
they
can
they
can
try.
They
can
try
this
model
at
once,
so
I
think
great
work.
There.
D
That's
good
yeah!
I
think
we'll
have
to
make
a
couple
of
tweaks
like
this.
Might
change
this
to
this
or
we
might
yeah.
I
don't
know
there
might
be
some
other
things
that
you
want
to
change
on
it.
I
would
go
through
and
make
some
edits
or
make
some
if
you
want
to
go
through
and
give
suggestions
for
edits.
I
don't
know
if
we
caught
everything.
F
D
Yeah,
so
this
is
the
readme
that
is
referring
to
this
is
in
the
evil.
Learn
it's
like
evil,
learn
and
then
there's
a
divolent
repo
inside
that
organization,
and
then
that's
where
a
lot
of
the
code
is
hosted,
but
there's
a
readme
in
here
and
that'll
actually
come
up
when
you
go
to
that
main
repo
page
and
then,
if
you
want
to
edit
it,
you
go
to
this
pencil,
this
pencil,
icon
and
it'll
allow
you
to
edit
the
text.
D
If
you're
outside
of
the
computers,
you
can
make
a
pull
request,
of
course,
and
then
we
can
review
what
the
changes
are
and
accept
it.
This
is
yeah.
This
is
good.
I
don't
yeah.
I
mean
I
think
I
like
the
idea
of
having
the
little
sort
of
animations
and
then
kind
of
giving
people
a
taste
of
what
it
is,
so
that
they
get
a
sense
of
what
it
you
know
what
it
is
before
they
breaking
down
the.
What's.
D
D
A
D
F
F
F
F
D
F
F
D
D
F
B
F
D
F
F
F
E
F
F
F
F
F
B
C
F
F
B
F
F
B
G
B
B
F
F
F
B
F
F
F
F
F
F
B
D
D
B
F
D
F
A
F
A
A
A
A
A
A
It
is
segmented
like,
as
shown
in
the
library-
and
these
are
the
point
of
contacts
that
are
being
the
maintainers
of
this
library,
which
are
there
and
you
can
reach
them
out
like
any
of
us.
If
you
have
any
kind
of
trouble
of
facing,
or
you
can
create
an
issue
if
you
are
familiar
with,
if
you
are
familiar
with
github
or
you
can
just
ping
any
of
us,
we
will
try
to
resolve
it
as
soon
as
we
can.
So.
This
is
basically
about
c
dot
delta
segmentation
models.
A
I
am
going
to
create
a
song
like
this
says
segmentation.
Spelling
is
wrong
here.
These
kind
of
typo
is
going
to
be
fixed
before
it
is
going
to
be
hosted.
So
my
first
question
is
like
what
should
be
the
name
of
this
portal
like
it
should
like
today,
I
have
written
if
you
can
see
dave,
so
it
should
be
day
one
day,
one
day,
one
ai.
So
this
is
the
first
question
that
I
have.
D
F
A
B
F
F
F
A
A
A
Okay,
so
I'm
looking
for
this
next
thing
that
I
have
to
ask
like
or
excuse
so
here
it
is
the
main
portal
which
is
going
to
connect
every
library
implementation
on
the
web
app.
So
what
should
be
the
name
of
this?
This
is
the
main
portal.
This
is
that,
consisting
of
all
the
things
that
we
have
done
like
they
will
learn
basic
area
image
segmentation
general
devils,
do
daily
methods
and
all
these
things
they
should
be
named
as
they
warm
ai,
but
this
should
be
named
as
something
else.
F
A
F
A
Like
you
are
not
purchasing
like
you
can't
you
have
to
get.
If
you
want
to
redirect
someone
from
typing
something,
so
you
have
to
like
purchase
it,
so
you
can
do
whatever
you
want,
but
if
you
are
using
io,
so
you
need
like
it
has
that
protocol
that
your
name
of
the
website
should
be
same
as
a
username.
F
A
F
A
D
D
F
F
F
A
Like
I
think
like
this
is
the
seat
or
delegate
speech,
and
we
can
just
rename
it
as
tableau
and
we
can
keep
david
for
seated
against
itself,
whatever
we
are
doing,
yeah
whatever
we
are
doing
like
in
the
future
itself
on
cda
telegrams
are
going
inside,
they
will
learn
another
models.
We
can
do
it
as
separate
something.
F
No,
no,
not
necessarily
we
can
just
we
can
do
other
models.
Obviously,
inside
default,
yeah
like
there
are
a
lot
of
data
sets
on
human
embryo
like
human
embryos.
We
could
work
on
those
as
well
if
they
are
kind
of
in
on
track
with
deform.
I
don't
know
if
they're
very
on
track,
but
it's
after
all,
this
analysis
right,
we
can
like
devo,
learn,
might
not
always.
D
D
A
A
Part
where
you
can
just
click
it
and
whenever
the
model
is
hosted,
I
will
post
it
today,
so
we
can
like
visit
that
model
and
do
our
stuff.
Then
this
is
the
bacillary
model,
so
basically
like
fascinating
models,
a
bit
of
difficulty
while
posting
it.
So
I
think,
like
I
have
to
make
or
get
a
repository
for
it
as
well,
because
it
is
a
very
heavy
model
like
there
are
many
heavy
libraries
that
are
being
used
inside
it.
It
is
not
based
on
a
single
library
so
like
hosting.
A
A
So
there
is
a
thing
that
is
like
bugging
me
a
little
and
other
than
that
they
have
standard
is
completely
flexible
for
python
and
flask
apps,
like
it
is
most
suitable
for
that.
So
this
is
the
thing
that
I'm
considering
right
now
so
chandra
segmentation
model
is
some.
I
think
someone
in
this
group
has
said
something
about
it.
He
wants
to
do
work
with
it,
but
I
have
not
seen
anything
so.
I
started
working
on
it
right
now
and
I
will
soon
like
host.
A
It
will
either
make
it
a
web
app
or
make
it
google
caller
notebooks,
and
I
will
soon
integrate
it.
This
is
the
thing
that
have
been
done
tables
the
page
has
been
created.
Just
we
have
to
link
it,
they
warm
academy,
so
they
know
academy
is.
I
have
to
just
get
a
quick
update
on
that
like
it
is?
Is
it
still
in
progress,
or
should
I
keep
it
here?
Should
I.
D
Remove
it
no,
this
is
this
is
good.
You
can
keep
it
like
this.
I
think
this
is
good.
Well,
I
have
to
go
through
the
links,
though,
make
sure
that
they're
current,
because
I
had
to
move
some
of
the
content
around,
but
I
think
it
is
already
updated.
So
it's
google
classroom,
so
I
think
it
is
and
yeah,
but
I
think
we.
C
A
A
A
C
C
C
C
A
D
D
We
might
not,
we
might
not
say
c
elegans,
we
might
just
say
well.
Let
me
think
about
the
copy
on
this.
What
we
actually
wanted
to
say,
because
it's
a
little
bit
we're
a
little
bit
beyond,
because
we
already
have
basil
area
in
there.
We
have
other
things,
so
we
might
not
use
c
elegans
segmentation,
we
might
say,
like.
A
A
A
B
F
B
A
A
A
B
D
F
A
A
B
A
D
A
A
A
A
A
B
D
D
D
And
so
this
is
an
attempt.
This
will
be
like
in
an
endnote,
so
it'll
be
like
a
bibliography
in
formatted
in
you
know
different.
You
can
format
it
in
different
styles
and
you
can
download
it,
and
this
is
the
start
of
this
bibliography.
So
this
zip
folder
here
contains
papers
that
contains
a
list
of
not
only
papers
from
the
published
by
the
divorm
group
and
poster
posters
and
presentations
but
papers
that
we've
reviewed
in
the
meetings
as
well.
D
D
The
second
item
is
susan
or
sent
me
some
biological
chaos
papers,
and
so
I
know
she's
talking
here.
I
think
she's
thinking
about
chaos
now
for
her
thesis,
and
so
we
have
a
couple
of
citations
here
which
I've
added
to
the
bibliography
I'm
not
going
to
go
into
them,
because
we
don't
have
a
lot
of
time
today,
but
we'll
talk
about
them
in
future
meetings.
So
some
of
these
are
about
chaos
and
cortex.
D
D
D
That
have
been,
I
mean.
Obviously
they
haven't
successfully
generated
montaging,
but
it's
it's
sort
of
like
in
future
directions
that
were
put
it
forth
in
that
paper,
and
so
basically,
what
they're
doing
is
they're
taking
a
bunch
of
images
of
an
embryo
and
the
embryo
is,
you
know
they
put
it
down
on
one
side
and
then
they
use
sort
of
hydrodynamics
to
flip.
The
embryo.
I
guess,
is
that
right,
susan,
it's
like.
D
G
E
B
C
G
G
G
G
G
D
D
So
you
what
you
see
if
you
align
the
images,
is
you
see
that,
like
the
cells
are
kind
of
like
moving
across
the
image,
you
know
you
it
you
it's
well,
it
looks
like
if
you
looked
at
the
earth
as
it
was
rotating
and
you
took
pictures
of
it,
you
know
the
different
features
on
the
planet.
Will
you
know
appear
to
move
across
this
disc,
but
it's
really
a
sphere.
That's
you
know,
you're
just
sampling
the
surface
of
that
sphere.
From
your
vantage
point,
so
that's
what
you
get
with
this
flipping
microscope.
D
Is
you
get
this
sort
of
surface
of
the
sphere
and
you're
you're
sampling
the
entire
surface
as
it's
flipping
and
so,
but
what
they
did
for
this
version
of
it
was.
They
did
manual
image
segmentation,
so
they
took
a
they
sort
of
defined
a
circular
boundary
out
of
these
images
and
that
was
sort
of
enclosed
the
embryo,
but
didn't
include
you
know
the
background
and
then
those
images
were
aligned
so
that
you
can,
you
know,
present
them
in
sequence.
So
you
can
actually
see
the
movie
movement
of
things
across
the
surface.
D
Then
it
was
projected
to
this
3d
projection
from
a
two-dimensional
projection.
So
you
have
these
flat
images
that
are
circular
and
then
they
have
to
be
mapped
to
a
three-dimensional
projection
and
then
the
then
you
have
like.
Then
you
have
to
create
a
reference
frame
for
that.
An
x
y
coordinate
system
actually
an
x
y
and
also
with
an
angular
aspect
to
it.
D
But
that's
that's
something
that
we
can
talk
about
more
and
then
the
3d
image
is
filtered
and
then
you
do
some
sampling
of
the
of
the
space
there
and
you
figure
out
what
the
surface
should
look
like.
And
so
that's
that's
what
was
proposed,
and
I
and
I'm
not
I'm
not
going
to
ask
for
solicit
in
opinions
about
this
now.
D
But
if
people
are
interested
in
contributing
to
this,
you
know
I'm
gonna
continue
working
on
it,
but
we
should
talk
about
how
to
do
this
more
effectively
and
I'm
not
sure
you
know
what
the
best
way
to
do.
This
is
the
state
of
the
art
and
so
forth,
but
this
is
something
that
we
can
eventually
include
in
some
of
the
other
image
processing
tools
that
we
have
now.
This
is
for
axolotl.
So
this
is
a
bit
different.
This
is
a
different
embryo.
G
G
G
G
G
G
D
G
D
G
G
G
D
I
want
to
get
into
the
papers
today,
but
we
have
some
papers
in
our
reading
queue
that
are,
you
know,
they're
pretty
interesting,
but
I
you
know
I
wanted
to
give
people
an
opportunity
to
maybe
if
they
wanted
to
present
something
if
susan
wants
to
present
a
biological
chaos
or
perhaps
if
jesse
wants
to
present
something
or
krishna
here's
the
drive
folder
that
we
have
this
stuff
in.
So
if
you
want
to
take
a
look
at
that
stuff,
you
can,
I
don't
know,
is
krishna.
D
D
So
I
mean
that
those
are
examples
in
that
folder
or
anything
you
want
to
present
krishna
actually
produced
some
tutorials
for
the
for
the
neuro
match
academy
that
this
was
during
the
academy
itself
that
he
pushed
to
another
repository.
So
congratulations
on
that
it
was
actually
on
coding,
so
it
was
a
coding
exercise
that
he
did
a
tutorial
for
so,
if
you're
interested
in
that,
I
can
I'll
actually
maybe
make
that
public
to
the
group
in
the
group
email
this
week
so
and
jesse
welcome
back.
D
E
E
E
E
E
Orthogonal
slack,
but
the
one
about
the
biological
like
kind
of
a
different
take
on
biology.
What's
it
called
scale
free
biology,
integrating
evolutionary
development
thinking
that
seems
particularly
relevant.
So
I
don't
know
if
that's
mentioned
here,
but
that's
something
that
I
would
look
at
for
sure
and
just
be
sure.
Yeah.
D
E
B
E
That,
too,
before
I
think,
but
as
far
as
like
diva
warm
updates,
the
you
know,
the
presentations
earlier
today
were
great
and
I
look
forward
to
kind
of
helping
out
with
those
things
how
I
can,
because
I
look,
you
know
the
people
who
learned
what's
coming
along
with.
That
is
really
great.
I
mean
it's
been
in
some
other
things
that
I
mentioned
like
in
the
season
of
doc
stuff.