Machine learning and AI were used at the Tokyo Olympics, which took place in July 2021 after a pandemic delay, to track athletes in 3D, analyze posture, guide training decisions, verify identities and enrich broadcasts. Sport is one of the fields where these technologies are advancing fastest.
Key takeaways
- Intel and Alibaba Group were linked to a 3D athlete-tracking network that coaches could use to study movement in detail.
- Computer vision estimates body posture and joint positions from camera footage.
- Machine learning models help coaches individualize training using data on workload, sleep, diet and more.
- A facial recognition system called NeoFace was used to verify people involved in the Games.
- Motion sensors and timing technology supply live data for broadcasts, for example in beach volleyball.
Which technologies were used in Tokyo?
| Technology | Purpose | Example from the Games |
|---|---|---|
| 3D athlete tracking | Analyze athlete movement and biomechanics | Network associated with Intel and Alibaba Group |
| Pose detection (computer vision) | Locate body parts and key joints in video | Training analysis and feedback |
| Predictive models | Individualize training and flag risk | Coaches’ use of workload and recovery data |
| Facial recognition | Identify accredited people | NeoFace, covering about 300,000 people |
| Motion sensors and timing | Track activity in real time | Omega timekeeping in several sports |
How did AI help with athlete tracking?
The Tokyo 2020 Games were postponed by the pandemic until the end of July 2021. The Olympic Games council and Intel set out to provide services for spectators, such as 3D athlete tracking at sporting events.
Data from systems associated with Intel and Alibaba Group fed a 3D athlete-tracking network. It lets coaches examine every detail of their athletes’ movement. The approach relies on algorithms that interpret the biomechanics of athletes captured on camera and estimate the posture of key body joints.
What is computer vision doing for athletes?
Computer vision is the area of AI that lets machines process images in a way that imitates human vision. In sport, a machine learning method examines videos or photos of people and tries to locate their body parts.
Pose detection tools, such as those built with TensorFlow, can recognize facial features, body parts, clothing and posture. These steps often serve as pre-processing for training machine learning models.
How do coaches use machine learning?
National Olympic teams use machine learning to gain an edge over their competitors. Applications range from real-time feedback to athlete data tracking, and models can even help predict sports injuries.
Coaches rely on machine learning to interpret individualized data for each athlete, based on factors such as fitness status and capabilities specific to that person. Not every athlete is the same. Machine learning helps flag significant, high-risk changes that could affect performance.
Workload data, along with data on sleep, diet, hydration, mood, stress and perceived muscle soreness, can now be analyzed by machine learning algorithms. Coaches can use the results to decide how much training an athlete should do. Teams also use models to optimize formations and improve their chances of winning medals.
What about wearables and equipment?
Technology also reached equipment. Nike’s Vaporfly 4% running shoes were reported to reduce the energy cost of running by about 4% compared with other marathon shoes. Analytics on the data these products generate gives teams a way to quantify athlete performance.
Ralph Lauren’s “RL COOLING” apparel was described as sensing body temperature and releasing heat from the wearer’s skin.
How was AI used in beach volleyball and other events?
Omega’s timekeeping technology, with motion sensors and computer vision, was used in sports such as gymnastics, swimming and beach volleyball. It lets athletes track their activity in real time and review it right in the competition.
Beach volleyball is played by two teams on a sand court divided by a net. Training AI for it requires motion and positioning technology. The systems learn to recognize different shot types, from spikes and smashes to passes and blocks, and to follow the flight path of the ball.
This data is combined with readings from gyroscope sensors in athletes’ clothing. Motion sensors report direction of movement, jump height and acceleration. The processed data goes live to broadcasters for on-screen graphics or commentary.
How was facial recognition used?
A facial recognition system powered by AI called NeoFace was used in Tokyo to identify about 300,000 people involved in the Games. They included athletes, journalists, volunteers, camera crews and security staff.
Frequently asked questions
When were the Tokyo Olympics held?
The Tokyo 2020 Games were postponed by the pandemic and took place from late July 2021.
How does machine learning help athletes?
It analyzes data such as movement, workload, sleep and recovery to help coaches personalize training. It can also flag risk factors, including those linked to injuries.
What is pose detection?
It is a computer vision technique that finds a person’s body parts and joints in photos or video. It is used to study posture and technique.
What data reached broadcasters?
In beach volleyball, processed sensor data on movement direction, jump height and acceleration was provided live for graphics and commentary.
Conclusion
Sport offers some of the most interesting applications of AI and machine learning, both for improving athlete performance and for enhancing the experience of viewers at home or in the stadium.
The Tokyo Olympics showed what these tools aim to do: support people and improve their lives in many ways.
