Computer Vision: Why Now and What’s Ahead | Intel Software

Computer Vision: Why Now and What’s Ahead | Intel Software

Articles, Blog , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , 3 Comments


Hello, and welcome
to Tech Decoded. I’m Henry Gabb, editor
of The Parallel Universe, Intel’s magazine devoted
to software innovation. And today, I’m here
with Charlotte Dryden, who leads a team developing
software for visual computing, from Edge to Cloud. Charlotte has 20 years of
experience in semiconductors in international business. She holds a degree in electrical
and biomedical engineering from Vanderbilt University,
and an MBA from the University of San Diego. Welcome, Charlotte. Thanks for coming. Thank you, Henry. I’ve done some work in
computational linguistics– essentially, teaching
a computer how to read. Your teaching computers how
to see more like humans. What do you see as the
future of computer vision? So, I work in the computer
vision software technology space. And our whole goal
is to enable machines to see exactly as we do. And what that will
allow is many, many things, such as autonomous
vehicles, valet copters, robots to move as humans
move, and see as humans see. So I see a huge set of
opportunities there. Now, computer
vision is obviously a hot topic right now. But you and I may
define it differently. How do you define computer
vision, as the expert? All the science and
technology that’s needed– both hardware and software– to enable what we
are discussing, which is allowing machines
to see as humans see. Do you still need
a PhD to do it? No. I think that this
technology is becoming rapidly more accessible to
developers of all types. The ninja developers
who are the PhD experts, maybe they’re
ninjas without PhDs, and the maker community
and students, and even just people who like technology
and want to try it out. I think that there’s
enough abstraction with the technologies
that we’ve built, that many of the technologies
are accessible to many. The barrier to entry seems to
have come down quite a bit. And I agree, part of that
has to do with the software abstraction. But what else is lowering
the barrier to entry to people who want to get
in to computer vision? Yeah. Yeah, the one
thing that’s helped is the reduction in hardware– hardware cost. So, it used to require a big set
of servers and a lot of storage to develop any computer
vision technology. If you look at deep
learning, it used to require very expensive
hardware and large amounts of time in large pools
of data to train a model to do any sort of
object recognition. But now, the processor
speeds are faster. The price of the
hardware is reduced. And the hardware
is more available to the average person. So, with that, we
see more innovation from many different
types of developers. So Charlotte, Intel
is heavily invested in the area of computer vision. Can you tell us more about what
we’re doing in this segment? Yes. So, Intel has a broad portfolio
of hardware and software for computer vision. And in addition to the CPU and
the GPU for visual computing, we now have a suite of
hardware accelerator options that can allow the
right performance and power for the right
visual computing workload. So we have the Movidius IP
that we recently acquired, the Myriad product. We also have the FPGA that
we acquired from Altera a few years ago. And now, we’ve
recently– last year– acquired Mobileye, so that we
have computer vision hardware for autonomous driving. With that, we’re in the
developer products group. So we’ve designed
software tools to make that hardware accessible to
computer vision algorithm developers. For example, we’ve developed
the computer vision SDK, which includes a large number
of OpenCV library functions that have been finely tuned
for all of the various hardware accelerators. And then, we have deep learning
tools to help with optimizing trained models for
object detection– or facial recognition,
for example– so that they run best
on Intel hardware. And then we have a host of
tools for custom coding. When you bring up OpenCV– which has an
interesting history, because it originated in Intel
labs almost 20 years ago. And as a matter of
fact, a few months after I joined
Intel back in 2000, Gary Bradski, the
creator, had just published an interesting
article in Dr. Dobb’s Journal describing the OpenCV
library, and the things it could do to teach
your computer how to see. And at the time, as a new
Intel employee, I thought, I didn’t know Intel was
doing computer vision. But now, almost 20
years later, it’s almost the de facto industry
standard for computer vision. It’s open source. It has come a long way. What are some of
the opportunities now for developers who are
using OpenCV for computer vision apps? OpenCV is the de facto standard. A lot of expert
technologists over the years have made Open CV a great
starting point for computer vision algorithm
developers who want to add any sort
of vision function to their application
or their algorithm. So, Open CV will continue
to evolve, especially as Intel leverages
various hardware accelerators, especially
for low-powered situations– even high performance compute
situations in the cloud. So OpenCV will
continue to evolve, and will continue to have
more and more functions so that machines can
behave like a human eye. For the future for
developers, I see them continuing to leverage OpenCV. I see us continuing to educate
developers of all types on the use of OpenCV, so that
they know that it’s accessible and it’s not as hard
as it used to be. See, one of the things
I love about OpenCV, is that it makes me feel like
a computer vision expert, when I’m not. I love that. Most people don’t admit that. They use OpenCV,
and then they act as if they built the
whole thing ground-up. It’s raised the
level of abstraction to allow that to happen. And I get access to
highly-tuned functions that do computer vision. Exactly. What do you see
coming with OpenCV.js? So, I see a lot of future. OpenCV.js is a big
one, because it makes computer vision functions
accessible to web developers. So, with the Edge to the
Cloud and this whole internet of things, we’re going to
have a lot of web apps. And having strong
vision functions available to web app
developers, those are worlds that didn’t
used to come together. When you combine those worlds
with some of the deep learning technologies that we have, and
other advancements with custom coding– again– I see a bright future
for the internet of things with very
sophisticated vision functions. And now that virtual reality
and augmented reality are meeting computer
vision, what kind of compelling applications do
you see coming in the future? So we have good applications
for AR today that are useful. We can take our smartphones
and overlay our smartphone on an image and get
extra information about that particular image. Or, see video streams or other
images on top of the image that we’ve selected. That’s augmented reality. And for virtual reality, I think
we’re just getting started. We see virtual reality
a lot at trade shows. But the price of the hardware
isn’t accessible to many yet. So I see opportunities
for that area to evolve. When you’re taking
multiple video streams and combining that
with motion detection to create a surreal
environment, that’s very heavy and
complicated technology. But I can see where that would
help medicine quite a bit– or education, or
industrial use cases. And if we change
gears a little bit and think about the
socio-technical issues of living in an era
of ubiquitous cameras, that we’re under
constant surveillance, cameras are everywhere,
there’s certainly good and bad. But what’s your take
on what it means to live in the era of
constant surveillance? And what do you try
to convey to your team as they develop these
software solutions? Some people love that we live in
an era of ubiquitous computers. Some people just love their
selfies, and their videos, and their cameras. I, personally, am
a little stressed out by all of the
cameras everywhere. I value my privacy and I
value the safety of my data. I’m guessing I’m not alone. So, to me, the
ubiquity of cameras brings up the concerns
around ethics. And for my team, who
develops developer products, we need to think about
how to help developers be more responsible when
they’re developing their vision applications. How do we give them
the right controls so that they can give
the end user more choices around
privacy and security, while still having the
convenience that these vision applications allow? Well, thank you, Charlotte,
for sharing your insights on computer vision and
its broader impact. I’m Henry Gabb
with Tech Decoded. Please visit our website
for more detailed webinars on computer vision in
Intel’s hardware and software portfolio for visual computing. [INTEL THEME JINGLE]

3 thoughts on “Computer Vision: Why Now and What’s Ahead | Intel Software

  • oliver lando Post author

    6:50– "Most people don't admit that …then they act as if they build the whole thing.." lol!! but I guess that's what libraries are for.. Differently, I wonder what "visual ethics (camera privacy)" would entail?.. Thank you for the video..very projective!!

  • Calm Energy Post author

    Great interview! Charlotte sounds like a great team lead and I'm glad ethics are considered. Everyone should have a dystopian sci fi nerd (myself) on the team to ask the question "what's the worst that can happen? " And that person should see every episode of black mirror so we can avoid any of those types futures. This camera tech though should have way more pros than cons. OpenCB!

  • Raed M Post author

    That was very good interview and Charlotte did a nice job explaining computer vision. Thank You

Leave a Reply

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