Building a sports intelligence engine to boost the viewer experience
February 4, 2021

Building a sports intelligence engine to boost the viewer experience

Computer Vision
A leader in Media & Entertainment

A listed sports media company, broadcasting numerous sporting events live via an online platform.

The Challenge

Sports streaming is a rapidly growing industry, covering an increasing number of sports, leagues and teams. However, it is unfeasible for all this content to be directed by a human editor and many matches are therefore streamed from a single static panoramic camera positioned on the half-way line; a poor viewing experience resulting in low viewer retention and severely limiting the profitability of live-streaming platforms. The client therefore wanted to automate the editing of live-streamed sports to improve the viewer experience.

The Solution

Leveraging the latest machine learning models in video analytics, ML6 built an operational streaming pipeline that detects the players and their team, the referee and the ball as well as key ‘events’ such as the start of the game, goals and penalties. Auto-production utilises this feature extraction on the high-definition video footage to automate scene-cropping at the Edge, ensuring the highest quality version of the ‘smart viewing experience’ can be streamed directly to viewers around the world.


In the first three months of collaborating with the customer, ML6 delivered a far superior viewer experience of auto-edited live-streamed sport, and the project was runner-up in the highly contested AIconics Award for Intelligent Automation in 2019. The next phase will see more advanced player tracking to enable statistics of each team-member’s performance to be automatically compiled, such as top speed, time with ball, passes made or missed, goals scored and distance run. The additional value of this from a training and scouting perspective could be significant, especially at amateur level.


Related posts

Want to learn more?

Let’s have a chat.
Contact us