Intel developing 3D Athlete Tracking for Tokyo 2020 Olympics

Intel developing 3D Athlete Tracking for Tokyo 2020 Olympics


Ok, so w’re thrilled about our partnership
with Alibaba, and as Chris said in that video we’ve developed a novel technology concept
called 3D athlete tracking. Unlike the way this is done today there’s
no special suits, there’s no special sensors here, what we’re doing is taking the data
from regular cameras. We’re applying an AI algorithm to analyze
that information in many different ways. It’s gonna help the way the audience can experience
the Olympic games, in sort of, real time, but, also we’re finding its going to help the
athletes themselves in the process of training for the olympic games. And so we’re working with a number of partners
to bring this to life. we’re working with Alibaba, as you heard about,
and we’re also working with some interesting new companies that uh, I want to tell you
about now. We’re working with a company called wrnch. This is a computer vision and deep learning
software engineering company, and what I’m gonna do is have the CEO of that company,
Dr. Paul Kruszewski join me here and tell you a little bit about how we’re working together
to bring ideas like that to life, so Paul, would you join us please? Good to see ya. Thanks. Thanks for coming out. My pleasure. So tell us a little bit about wrnch. So wrnch AI is the worlds fastest and most
accurate human pose estimation platform which extracts 3D human motion and behaviour data
from standard video. Our trained model takes 23 skeletal tracking
points from live video in real time which enables us to analyze video data that just
wasn’t possible before. Thats great, and I understand you’ve been
using Cascade Lakes and using some of the AI algorithms that you have on that, tell
us about that. Yes, we’ve been working with some great Intel
engineers to see just what Xeon processors could do for wrnchAI. I’d like to show you what we’ve been able
to accomplish so far. Uh huh. What I’m showing here are 2 systems running
wrnchAI. On the one side is our standard GPU configuration,
and on the other side is with the new Intel processors John, if you could start the demo we can see
how the results stack up. So both systems are now tracking the human
movement in the video, on the exact same video. With our current solution on a GPU we’re inferencing
around 90 frames per second. But, on the Cascade Lake platform, optimized
with Intel DL Boost, we’re inferencing at 450 frames, which blows my mind. (laughs) Now if my math is right there its
5X faster, am I right? It’s 5X and we’ve only really started. So there’s this perception out there that
you can only, you need GPUs, and thats the only way to do AI. What’s your experience with that? That was our perception going in to this project. We felt exactly the same way as everyone else,
but the results speak for themselves. It totally changed our minds. I don’t think its any more about CPU vs GPU,
the results are fantastic and we’re really excited to bring this Xeon compute to our
cloud system. That’s awesome. Thank you so much Paul, we’re looking forward
to it. Thanks a bunch.

1 Reply to “Intel developing 3D Athlete Tracking for Tokyo 2020 Olympics”

Leave a Reply

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