Last Tuesday Paul van Herpt and I traveled to Lille for a special Machine Learning and Cycling Masterclass. As data partner of Soudal Quick-Step Pro Cycling Team, these are exactly the applications that touch where we as Transfer Solutions can make the difference. Hence Paul and I followed this special course from the IDLab (UGent – UAntwerpen – imec).
Machine learning is already used a lot in sports. In soccer, for example, a huge amount of statistics is at hand: who has how long ball contact, who usually shoots to whom, makes the most runs, who is the most dangerous? That kind of data is already very easily traceable. And in tennis, it is easy to track the ball, calculate speed, etc..
But for cycling, it’s a bit trickier. The turf in soccer or the gravel in tennis differ little from match to match. But because cycling races take place in public spaces, the image of the race on TV, or the road surface is always different. You can follow from a helicopter who is pulling the sprint in the final kilometers before the finish, but then the peloton just turns into a tree-lined street. And there goes your computer vision.
Riders, on the other hand, like many athletes, wear all kinds of wearables, and the bike is also equipped with sensors such as power meters. Those don’t say everything. Ideally, athletes should do an intensive lactate test on a trainer bike, so that it is clear exactly when the rider becomes exhausted. The problem with this test: it is a really exhausting test and that often does not fit into the training program. But with new applications of machine learning, it is possible to make an estimate based on one or two lactate tests and for the rest of the season using a model that works with data from wearables and sensors.
Computer vision has seen considerable growth in recent years. Even then, recognizing a rider in the peloton in TV footage can still be a challenge. Is the bib number in the picture? That’s easy then? If not, the IDLab speakers showed how they use a combination of team shirt recognition and facial recognition to identify the rider. Tricky with that, though, is if the team changes shirt designs frequently.
That recognition of riders can be used to see if the riders optimally keep their sprinter out of the wind and bring them to the front at the finish. Or, in track racing, whether turnouts are taking place optimally. When it comes to safety, it can detect whether sprinters stick to their line in the final phase. This is all possible with machine learning. These algorithms work with a certain accuracy, which will never reach exactly 100%. So when disqualifying sprinters, for example, a human check will always be necessary.
Much more is possible with computer vision algorithms. For example, pose estimation is used for bikefitting. That is, is the bike optimally sized for the rider?
A session on smart prompting what intended to give those without extensive Python experience a leg up: Large Language Models can help achieve initial successes in machine learning.
Finally, the session featured hands-on assignments and CoLab notebooks to try out later. For example, we did an assignment that used data from smartphones to identify the type of road surface. This can reveal in advance where in the route there are firm cobblestones.
Paul and I were the only ones hooked up not from sports, but from our data expertise. It was interesting to hear what the other participants (coaches, movement specialists) were struggling with. One of the biggest problems for them is that although there is a lot of data, it almost always goes straight into the ecosystem of a company like Strava, Garmin, intervals.icu and the like. And there you can only get it out again with difficulty.
For Soudal Quick-Step Pro Cycling Team, Transfer Solutions tackled that right at the source. Data from Garmin devices is automatically fed into the team’s custom analytics platform. That we then have to do our own interval detection is then still worth it, because the coaches can then work on how they want to visualize the data themselves.
All in all, there was plenty to get a lot of inspiration and to put what we learned into practice. It was well worth braving a hot day in Lille.