Swimming Australia to turn on computer vision at training

When the Australian Swim Team heads to Paris in July, computer vision — tracking their every stroke at training pools — will have enriched the same predictive insights that propelled them to the podium at the 2021 Tokyo Olympics. 

Diving deeper into data after Tokyo 

In addition to nine gold medals, Australia's peak governing swimming body returned from Japan with confidence that its data analytics investments had paid off.

Swimming Australia's general manager of performance support and Olympic campaign Jess Corones told iTnews that, since 2021, AWS has been working to connect Swimming Australia’s data lake, Atlantis, to more data sources and to build more predictive insight applications for coaches. 

“After the success of the Tokyo Olympics, and the work that we've done with AWS improving the impact that the data and the transformation of the data and the usability of it could have on performance, we can see the potential and have started to invest more,” she said. 

In the lead-up to the games, the preparations of every elite swimmer - or ‘Dolphin’ - were guided by Atlantis’s predictions about optimal diet, training schedule, injury management plan, race plan and more. 

It was also the first international swimming meet where both historical data and real-time data on the Dolphins’ and competing teams’ performance throughout the day predicted the optimal combination and order for swimming relays.

Coaches accessed the relay suggestions in real-time and made last-minute decisions on order; the Dolphins medalled in six out of the seven relays that they competed in, making them the highest medal-winning relay team at the Tokyo Olympics.

Expanding computer vision 

Since 2019 — when Swimming Australia released its bespoke athletic improvement system “Swim Performance and Race Tactical Analysis” (Sparta 2) — computer vision has tracked Dolphins, but only at competitions.

Automatically recording each swimmer’s technique and performance during training sessions, and then feeding it to Atlantis’ performance insight models is one of Swimming Australia’s and AWS Professional Services’ current projects. 

“We're in a current project at the moment with ProServ [AWS Professional Services] developing a new system called Training Insights,” Corones said during an interview at AWS re; invent in Las Vegas in December last year. 

“The idea is that we've got really rich competition data, but we want comparable data in the training pool.

“We're developing a piece of software that then can use the machine learning and image recognition to track the athletes in a training pool to give us the same data metrics that we're getting in the competition pool.

“And so that's something we'll be looking to release, hopefully, in the next couple of months.” 

It’s much easier to track a swimmer’s distance-per-kick, breath count, turn times and other stats when every other variable is consistent, which they are in a competition environment, Corones added.

“In a competition, you've got eight swimmers, one per lane and they go up and down, but in a training pool, there are so many more variables that the computer vision has to account for, pick up, [and] understand what to focus on and what to dismiss.

“In a training pool, you could have up to six swimmers in one lane, doing four different stroke ... and they're not all doing the full lap - they might be stopping at 25 or 35 meters; they might be doing different segments. So how do we measure that in an automated way?”

Lane Four 

Swimming Australia engaged AWS in 2018 to turn data it collected on athletes’ performance during training and competitions into decision-making tools for coaches.

The data ranges from statistics from the International Swimming Federation Association (FINA) to various athlete management systems, and data from wearable technologies and Sparta 2.

The Professional Services team used the AWS Serverless Data Lake Framework to integrate the data sources and establish data lake environments.

Corones said that generating performance insights through Atlantis had enabled training to become “much more individualised towards each athlete, and each coach and each training program.

“What the data lakes allow us to do is have that really holistic view and understand each of the different variables and how they interact in performance.” 

As well as connecting Atlantis to more data sources, another of Swimming Australia’s and AWS’ focus areas after 2021 has been making the insights more readily available to more coaches, Corones said.

“So after we create those insights, how can we get those in front of the coaches in a nice visual way to enable them to use those insights to make decisions on athletes and training?” 

In early 2023, Swimming Australia and AWS released an online portal that allows coaches to access their swimmer’s data and predictive insights from Atlantis via an AWS QuickSight dashboard. 

“Lane Four” is a “one-stop shop coaches can access on their iPad from the side of the pool to access everything they need to know about their athlete from strength to conditioning to nutrition,” Corones said. 

“It can also [respond] to their natural language queries and find the answer to the questions that they want. 

“So it might be for example, the coach might want to know who had the top three fastest turns in the women's 100-meter freestyle in the world since 2003. They can ask that question without having to try and find the filters and dates.

"This is so beneficial for the coaches because sometimes in a training session, they need that information at the click of a finger; they might be in the middle of a training set and they need that benchmark data to give the athlete the goal.”

Jeremy Nadel attended AWS re:Invent 2023 in Las Vegas as a guest of AWS.