Hello. My name is Martin
Kronberg, and this is the “IoT Developer
Show,” season 4. In this 7 episode series,
we explore the Intel vision accelerator design products. These software
and hardware tools are focused on helping
you create computer vision solutions in spaces like
industrial and retail. As usual, we discuss
the tools you can use to create your
project, and highlight some great demos to show off
the technology in action. Before we get into the
details about these solutions and designs, let’s take
a look at some reference implementations. Cameras are rapidly finding
their way into most IoT edge applications. Thanks to the
advanced [INAUDIBLE] computer vision
technology, we are not only able to stream
video, but we are also able to process multiple
video streams in real time. For instance, you can
use an AI based camera to detect traffic patterns
in a retail location. In an industrial or
factory environment where human safety is
critical, you could not only have your video
analytic server check for defects on
the assembly line, but that same server can
detect a human getting too close to the safety
zone and issue an alert. We can monitor
traffic congestion, and optimize traffic
patterns, monitor pedestrians for improved
safety, and even use smart video to keep an
eye on the health of city infrastructure. IoT is a broad market, spanning
across many more segments like agriculture, smart
cities, smart homes, and connected buildings. Each of these segments has
a custom requirement when it comes to vision processing. In order to address
this challenge, Intel and its partners
offer a variety of Intel vision accelerator
design product solutions. Let’s look at some ways that
these solutions can benefit your vision applications. Unlike high performance
systems typically found in the data center and
cloud environments, vision analytics
are often deployed at the edge near
the actual sensors, or in our case, cameras. This is necessary,
because you will often face bandwidth issues
when streaming video to a data center. And this problem is
compounded if you stream video feeds from dozens
or even hundreds of endpoints. Edge analytics can also help
you improve data security. Processing image data
close to the source lets you keep the raw
image data private. Only the results are
uploaded to the cloud or data center for further analysis. Processing at the edge has
its own issues to consider. Edge processing capacity is
limited by several factors– space, cooling, and cost. Technologies like video
processing units, or VPUs, and field-programmable
gate arrays, or FPGAs, are available, which
offer unique benefits for vision solutions. We’re going to cover
those in great detail throughout the season. VPUs and FPGAs are
much more power efficient than a CPU or GPU,
fit into a smaller package, and reduce the processing
requirements of the main CPU. Vision analytics
at the edge often requires multiple
layers of processing– image cropping, detection, and
classification, for instance. These tasks can be shared across
multiple processing elements, which are tailored to
improve the performance of the application. With video specific tasks
being handled by specialized hardware, the CPU can be
utilized for operations that it’s most efficient at– networking and database
management to name a couple. Using multiple types of
processing in a single solution is known as
heterogeneous computing, and it is something
that we will talk about throughout the series. Keep in mind that we’re
focusing specifically on hardware acceleration, but
there is an integral software component, as well. The Intel distribution
of OpenVINO Toolkit lets you develop and optimize
your applications quickly on a PC, and then target
the appropriate hardware technology– CPU, integrated GPU, VPU,
FPGA, or some combination. And that’s it for
the first episode. In the upcoming episodes we
cover Intel vision acceleration hardware offerings. We will see the Intel
neuro computer stick 2, multi VPU accelerators, such
as the Mustang V100, and FPGA cards, such as the Mustang
F100, featuring the Intel Arria. Thanks for watching,
and see you next time. [MUSIC PLAYING]

Intel® Vision Accelerator Design Solutions | IoT Developer Show | Ep. 1 | Season 4 | Intel Software
Tagged on:                                                                                                                                                     

One thought on “Intel® Vision Accelerator Design Solutions | IoT Developer Show | Ep. 1 | Season 4 | Intel Software

  • August 1, 2019 at 9:19 am
    Permalink

    Video 2,3 and 4 are not accessible in the list

    Reply

Leave a Reply

Your email address will not be published. Required fields are marked *