TRACAB Unlocks New Match Insights by Enhancing Sports Tracking Capabilities on AWS

TRACAB uses a computer vision system with multiple cameras to collect and deliver data describing the positions of players and the ball during football matches. This generates near real-time insights for media broadcast teams, referees, players, coaches, and fans.

Originally, TRACAB powered its solution using on-premises servers at each arena. But as it grew in complexity, TRACAB decided to migrate to AWS to increase scalability and speed up deployment. First, it automated a task that required two onsite operators. By sending tracking data to the cloud, TRACAB could use AI powered by advanced deep learning models on AWS to perform the task with a single remote operator. This reduced manual work and increased productivity.

TRACAB gained the confidence to do even more on AWS. It began by streaming video directly to AWS in 2022. TRACAB next enhanced its solution in 2023 with near real-time skeleton tracking, which creates a three-dimensional visualization of players. “Skeleton tracking has added a whole new level of realism to our data,” says Dr. Gareth Loy, head of computer vision and machine learning at TRACAB. The company used skeleton tracking to build new use cases, such as semi-automated offside technology and virtual cameras, which make it possible to view matches from any angle and point of view.

To power its solution, TRACAB uses Amazon Elastic Compute Cloud (Amazon EC2) G4 Instances, which are powered by NVIDIA T4 Tensor Core GPUs, and Amazon EC2 G5 Instances and Amazon EC2 G6 Instances, which are high-performance GPU instances. TRACAB also uses Amazon ElastiCache for Redis, an in-memory data store, and Amazon DynamoDB, a serverless database, as the data stores for low-latency access. On AWS, TRACAB can collect more than 255 million data points per match while achieving an average latency of 0.85 seconds.

Automation and the capability to collect richer event data have improved TRACAB’s offerings. For example, manually collected event data about a pass position can be up to two meters off. Now, TRACAB’s automated event data perfectly aligns with the video, tracking data and event data, and it can offer more details such as the speed and distance of the pass, how players moved in response, and which foot the player used.

TRACAB has rolled out its solution to nearly 200 stadiums, which collectively run more than 2,000 matches per year for leagues such as Bundesliga, LaLiga, MLS and Eredivisie. On AWS, TRACAB plans to capture more data, expand to more stadiums and leagues, and discover more use cases for its data.

“On the cloud, we can embrace new technologies, especially in AI and machine learning,” says Martin Brogren, general manager at TRACAB. “On AWS, as soon as we develop features and solutions, we can simply deploy and scale them on the cloud.”