►
From YouTube: Keynote: Incremental Deep Learning For Satellite with KubeEdge... Xiaoman Hu, Zhipeng Huang, Yue Bao
Description
Don’t miss out! Join us at our upcoming hybrid event: KubeCon + CloudNativeCon North America 2022 from October 24-28 in Detroit (and online!). Learn more at https://kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects.
Keynote: Incremental Deep Learning For Satellite with KubeEdge and MindSpore - Xiaoman Hu, Community Operation Director; Zhipeng Huang, Director of AI Open Source; Yue Bao, Software Engineer, Huawei
Low Earth Orbit (LEO) has become a hot topic in recent years and KubeEdge has managed to bring the cloud native to the space. With several small research satellite equipped with KubeEdge, edge computing with AI was experimented at the furthest frontier. In this talk we will present how we combine KubeEdge Sedna, the cloud native edge machine learning suite, with TinyMS which is a high level API toolkit for MindSpore deep learning framework, to enable incremental learning at the satellite to accomplish tasks like remote sensing and earth observing. We will also discuss how utilize TinyMS that KubeEdge Sedna's semantics could be integrated as an integral part of AI framework.
A
Bungie,
valencia:
it
is
a
great
honor
to
have
the
opportunity
of
keynote
at
2.com
europe.
This
year
we
are
from
the
open
source
cloud
ai
team
in
huawei.
Are
we
are
members
of
the
cuban
volcano
commander
project?
In
addition,
I
was
involved
in
the
early
days
of
forming
kubernetes
policy
working
group
and
the
sincere
security
tag.
A
We
combine
the
space
technology
with
cloud
native
and
ai,
so
in
the
next
two
sections
you
are
going
to
hear
from
two
extraordinary
speakers
on
how
cubatch
and
minds4
helps
the
cloud
native
and
ai
technology
in
space.
Next
up,
please
welcome
value
from
the
cube
batch
team
to
give
a
deep
dive
on
cloud
native
for
space.
B
B
It
graduated
from
sandbox
into
a
cncf
incubation
project
in
september
2020
and
from
2020.
We
have
measured
some
in-depth
attempts
at
a
wide
range
of
fields.
We
now
have
five
six
angle
working
groups,
including
the
aic
device,
iotc
mbcc,
robotic
secondary
wellness
working
group
since
they're
launched
until
now,
cool
badge
has
a
healthy
growth.
B
B
B
We
have
two
components
conclude
and
edge
in
the
cloud
we
have
a
plot
called
it
talks
to
kubernetes
and
process
the
request
to
the
age.
The
age
has
an
edge
pole
and
it
talks
to
container
engine
and
devices
the
edge
takes
down
to
17
megabytes
memory
footprint
to
run
and
also
support
the
oci
conference
container
runtimes.
B
There
are
some
key
features
of
the
coupe
age.
First,
we
support
the
kubernetes
native
apis
for
developers.
We
also
support
seminars,
cloud
age
coordination,
edge
autonomy,
low
resource
environment
with
the
iot
devices.
Kubej
provides
the
device
mapper
to
simplify
the
integration
with
different
device
protocols.
B
Also,
we
provide
a
cloud
view
of
global
metrics
from
the
data
sedna.
The
sentinel
project
is
an
air
tool
cage
over
cube
edge.
It
provides
the
age
cloud,
the
synergy,
ai
framework,
to
support
joint
influence,
incremental
learning
and
federated
learning
in
sentinel
architecture
you
can
find
we
have
a
centralized
global
manager
in
the
cloud
to
coordinate
with
all
the
components
of
different
edge
nodes
on
each
edge
node.
B
B
B
In
addition,
the
device
mapper
of
the
cube
edge
is
used
to
model
and
manage
sensors
of
the
satellites
in
in
a
unified
manner
allowing
management
personnel
on
the
ground
to
obtain
the
working
status
of
onboard
devices
in
real
time.
All
of
these
communication,
with
each
other
and
through
a
highly
reliable
cloud
age
channel
established
by
the
curvature.
C
Students
are
generating
thousands,
even
millions
of
images
every
minute
of
the
day.
These
data
are
useful
for
obtaining
information
about
farming
area,
up-to-date
online
maps
and
lots
of
other
data
having
the
ability
to
optimize
a
vast
amount
of
data
using
artificial
intelligence
and
cloud-native
is
important
for
end
users,
but
how
can
it
work
if
we
like
to
detect
the
following
area
on
using
this
data?
C
We
will
train
the
image
detection
model
by
mysport
with
tremendous
amounts
of
data,
so
my
sport,
eulo
v3
tiny
model,
is
pre-deployed
on
the
satellite
and
served
with
cubase
yamna
via
tiny,
ms,
with
a
size
of
no
more
than
30
megabytes
which
fades
in
with
a
satellite
memory.
The
cells
can
detect
the
farmland
area
by
themselves.
C
So
when
discovering
heart
samples,
cuban
cetina
local
controller
compresses
and
sends
these
datasets
to
the
ground
using
high
precision
image
detection
model
for
inference,
these
hard
samples
can
also
be
used
to
train
a
new
model
and
improve
the
precision.
So
the
keyboard
set,
my
global
manage,
performs
incremental
learning
tasks
using
mysport
to
train
it.
After
refresh
the
model.
C
Using
new
models
in
the
end
with
jetlag
to
receive
the
compressed
partial
model,
my
sporter
decompresses
and
updates
a
model,
and
then
pubescent
already
deploys
a
new
model
so
that
the
accuracy
can
be
better
using
new
models.
In
the
end,
you
heard
the
word:
mice
worry
many
times.
So
what
is
it?
Mysport.
A
C
It
is
a
really
user-friendly
ai
framework
that
only
requires
the
developers
to
master
the
basis
of
tensors
operators,
models
and
python
programming
without
a
steep
learning
curve.
On
many
of
the
underlying
capacities
maneuver
with
the
features
such
as
high
order,
differentiation,
optimization
automatic
parallelization
and
graph
operator
fusion,
my
sport
could
achieve
very
high
performance.
A
C
C
Secondly,
in
order
to
paradism
uses
zero
algorithm
code
to
implement,
distributed
parallel,
training
and
maintain
high
performance,
the
paradigms
of
distributed
parallel
training
include
data
model
and
hybrid
parallelism.
My
sport
uses
a
new
type
of
distributed
parallel
training
that
integrates
all
these
paradigms.
Therefore,
by
the
given
data,
my
sport
shortens
image
classification
model
training
by
23
with.
C
Our
team
also
develop
a
high
level
api
toolkit
for
my
4,
which
will
cause
tiny.
Ms,
you
can
see
the
architecture
on
the
bottom
of
this
slide,
aiming
for
a
now
ai
central
service,
faster
adoption
of
deep
learning
framework
capabilities,
especially
for
new
beginners.
Since
its
inception,
mysport
established
an
open
and
global
community
for
developers.
C
In
nearly
two
years,
mysport
has
achieved
more
than
1.2
million
downloads,
with
more
than
20
invitesport
study
groups
created
all
around
the
world.
The
community
adopted
open
governance
with
a
14
member
technical
theory,
committee
and
26
workshops.
The
community
has
also
pioneered
inequality
and
diversity
with
efforts
like
msu
women,
tech
with
various
community
partners.
Not-For-Profit
collaboration
like
train
train
after
a
red
camera
model
for
natural
protection,
further
helps
the
ai
for
goods.
B
C
My
sport
can
make
incremental
deep
learning
simple.
The
ai
workflow
in
this
case
have
been
explained
in
detail.
The
model
retrained
can
detect
features
that
are
nominal,
such
as
typical
fundamental
shapes
and
differentiate
them
from
user
patterns.
We
can
use
ai
to
determine
which
datasets
are
important
present
to
the
ground
segments
for
processing.
C
C
We
think
a
lot
of
ai
technologies
we
describe
above
will
be
using
many
water
will
be
developed
over
in
the
next
couple
of
years
to
assist
this
mission
and
many
more,
such
as
state-based
learning
and
class
native
policy
and
in
the
future.
We
hope
that
cloud
native
space
computing
that
can
enable
better
audit
grounds
and.