OTT & Media · AI in production
Point a phone at the court, get the player
Player recognition from a phone camera during a live game, with statistics on the screen before the play is over.
- Client
- A U.S.-based sports technology company
- Segment
- OTT & Media, sports
- Timeline
- 2 months, 4 person-months
- Platform
- Mobile app with a backend for sensor data
- OpenCV
- Unity
- Java
- Spring
- Apache Spark
- Tomcat
Before
What was true before
A fan in the arena had to choose between watching the game and looking up a player. Scouts and trainers had the same problem with better seats: the performance data existed, but not while the play they cared about was happening.
After
What we did
A mobile application recognises players through the phone camera during live play and shows their statistics, backed by a service that calculates player positions and collects data from ball and player tracking sensors.
Build log
How it was built
- 01
Designed an ML workflow that recognises only the players moving on court
- 02
Processed phone positions on the backend to predict detection scenarios
- 03
Sent player coordinates to the backend and calculated positions from any spot around the court
- 04
Showed the player’s ID and stats in the app when a fan taps them
- 05
Collected real-time data from ball-tracking and player-tracking sensors
Measured
Results
As reported in the published case.
- to recognise players during a game
- 2-3 seconds
- MBps of player coordinate data generated
- 43
- FPS player detection in the mobile app
- 16
- person-months over 2 months to deliver
- 4
Engagements
The services behind it
Engagement
Where this work sits
Next step
Yours would be the next one.
Tell us where it breaks, and you get a written answer on what breaks first and what it costs.