Sparrow: Sport Analytics and Computer Vision Technology

This is a past event

24 people went

The Jay Hurt Hub for Entrepreneurship and Innovation

210 Delburg St · Davidson, nc

How to find us

The talk is in room 208. Please leave a message on or in the Slack channel if you have trouble getting in.

Location image of event venue


Todd Eaglin from Sparrow will show us the challenges his team faced in applying computer vision to solve sports analytics problems.

Location: The Hub, Room 208

** Meetup Schedule **
• 6:00 - Networking
• 6:15 - Topic Presentation
• 7:15 – Chat with presenter
• 7:30 – Wrap Up.

Todd's Bio:

I received my PhD in Computer Science at UNCC. During my dissertation I did extensive work in computer vision, GPU and mobile computing, data analytics and visualization.

I’ve worked for many years as a software engineer and a data scientist at Lowes. I am now a co-founder and technology director at Sparrow.

I enjoy playing sports and being active. I played div II lacrosse at UNCC and I now do Crossfit. I also have a passion for flying airplanes. I hold a private pilots license and I fly regularly.


Bringing computer vision applications to the consumer market is a challenging process for businesses. In my presentation I detail the technological hurdles my company has overcome developing novel technologies for sports, physical therapy and health. I will also cover the scalability and economical challenges with bringing these technologies to mobile computing. Lastly, I will cover challenges we faced in acquiring data and areas of future work.

Outline of the talk:
•Who I am
•The team
•How I got here

-The company
•What we do
•What we’ve done

-Technology (Detail 1 example – Soccer kick)
•Problem statement + Challenge
-Analyze ~1500 frames in 5 seconds or less. (6 second video @ 240fps)
-Device limitations (Compute + Camera)
-Bandwidth/Network limitations-Lighting limitations
•Why Pose estimation is bad in production and at scale
•Edge computing + Machine learning + Cost
-Edge computing (Mobile) is challenging!
•How we overcame it (General tips).
•Closing thoughts.

-Generating data
•In some cases collecting data is unsafe.
•Data augmentation at scale
•GANs (Adversarial networks)
•Realtime raytracing and simulations
-Future work
•HMR (Human mesh recovery)