The recent Olympic Games 2020 made us cheer up after long gloomy COVID-19 impacted days. Made us all realise how the athletes trained themselves in restricted scenarios and limited resources all due corona induced changes and restrictions while still giving their best attempts. The trainings for outdoor games were not the same owing to stadiums and grounds being locked up. Coaches came with various innovative strategies to train and prepare the athletes for best performance. 

It might not have been very evident just looking at the games on TV that AI had a critical role in Tokyo Olympics. Olympics have a great history of incorporating technology in various areas of training, monitoring, scoring, performance management, and more.   

AI coupled with various technologies improves data accuracy when it comes to various games. Japan had been at the forefront in terms of embracing the newest technological advancements and innovation. The country has been amalgamating futuristic technologies in every developmental activity and setting up ground breaking implementation across all sectors.   

 In line with this, Intel’s 3DAT made its Olympic debut this year at Tokyo Olympics. 3DAT stands for 3D Athletic Training which is a technology developed by Intel and Alibaba. The technology uses existing and upcoming Intel hardware and Alibaba cloud computing technology to power a cutting-edge deep learning application that extracts 3D forms of athletes in training or competition. 

The AI aspect of the system is being taken care of by WRNCH. It is a computer vision / deep learning software engineering company based in Montréal, Canada. The wrench AI platform enables software developers to quickly and easily give their applications the ability to see and understand human motion, shape, and intent. 

Jonathan Lee, Director of Sports and Performance technology, Olympics technology group said, "3D Athlete Tracking is a technology we developed here at Intel that enables us to capture standard video of athletes and extract information about their shape and movement," 

He added, "We do this with the help of AI and computer vision. With the technology, we can recognize the different parts of the body, from the eyes to the nose to the ankles and toes, and use them to construct a 3D skeleton of the athlete or, in some cases, several athletes. We can then extract information such as speed, acceleration, and biomechanics from these skeletons." 

The idea is to club computer vision with AI deep learning algorithms to generate a 3D Mesh that enables coaches and trainers to extract intricate real-time biomechanical data via multiple standard video cameras without the use of special sensors or suits from which they can analyze performance and introduce new training enhancements. 

During the Tokyo Olympics, 3DAT works by mixing videos captured by multiple 4K image processing cameras with broadcast footage. This fusion creates a three-dimensional model of athletes in action. In simpler words, this technology will monitor and capture very fine movements of the athletes (cameras running at 60 frames per second) which will provide many accurate data from an athlete's performance. This is a revolutionary idea that could transform the sports training scenario.  

The cameras footage will create the 3D structure of the athletes' performance which later can be converted into animated videos or into graphical content. The data produced by the cameras is then analyzed using Intel Deep Learning Boost AI capabilities. During the games, the athletic performances can be captured on a superficial level and not on the granular levels. However, the game changes in milliseconds, and hence it is critical to track every fine detail and movement.  

In simpler terms in a 100 meters race, the 3DAT technology captures the speed of the athlete at run time and creates heat maps of the same. This depicts the speed, acceleration, de-acceleration, and biomechanics patterns during the race. It can also show at which point an athlete achieved the best speed or the lowest and on what body movements the same happened. 

 Trainers can understand the style and pattern of the athlete’s performance for better analysis Ising the data produced by the 3DAT systems. The whole process of capturing the video and processing takes around half a minute which can then be broadcasted. This speed of processing provides a great experience to both the audience and the trainers.  

The challenge with the usage of 3DAT in the Olympics was the fact that unlike other sectors such as movies or entertainment, sensors cannot be fitted on the athletes' bodies. And hence came the need to train the AI algorithm accordingly to work just like it would with the sensors attached. As this was the first such application of AI, the models had to be manually trained at first and late were modified to produce results by studying the data fed. 

The technology is revolutionary and looks promising in sports and athletics. The development on the same technology will lead to even impressive results in the days to come. Digital twins are said to be the future of Olympic training. Hope AI proves its worth again in a fair and bias-free manner.