How computer vision helps young football players to improve skills

Computer vision and ML models from SberCloud are already helping to assess the potential of young football players, collect objective data and identify talented players. The JuniStat project, which recently attracted $310 thousand from the AngelsDeck Venture Investors Club, thanks to this technology reduces the threshold for entering big sports for young athletes.


Remote selection technology

Each football club wants to have top players. But they need previously to be prepared, raised, and before that – to be selected. So, scouts of sports academies are constantly looking for talented children in specialized schools, and at tournaments. However, doing this nationwide is expensive and difficult, and most of talented children do not come to their attention.

Nevertheless, in Russia and globally there are enough bright players who can reveal their potential with proper training, but you can find such talents only by chance. Many simply do not have access to advanced techniques and coaches to improve skills and come forward. At the same time, the parents of young football players cannot always objectively assess how talented their child is and help him with his studies and the development of his sports career.

But every young football player deserves a chance at least to show himself and thanks to his talent and motivation to get into a big sport. So there was an idea for two applications: JuniStat and JuniCoach. The first app helps novice athletes to train on their own and send online applications to clubs. The second app is aimed at coaches and simplifies team testing and data collection for players in football schools.

What does it look like for a football player

The JuniStat app has a personal account of the player. There he talks about himself – indicates the role on the field, height, weight and other information that he considers important. There you can load a cut of the best moments from the matches. In addition, the app has special tests that the young football player fulfills for his assessment. This works as follows:

  1. Junior performs a task on a smartphone camera from a mobile application.
  2. Video is analyzed by artificial intelligence technology.
  3. The system forms a digital player card that can be viewed by football academies and scouts.

The presented tasks help to evaluate the physical and technical training of the player, knowledge of the key skills of the player – the quality of dribbling at speed, ball control, transmission accuracy, the strength of shots on goal, and so on. With the help of computer vision, the application collects more than 60 parameters of physical and technical training of a football player. As a result, each player has his own digital profile with development statistics in the context of many years of preparation.

JuniStat was initially created not only for juniors who want to build a football career. There are useful functions for enthusiasts who train for themselves. For example, they can compete with friends in holding the ball. The application added the Freestyle test, where players can demonstrate their favorite feints. They are then watched by other members of the community. We also launch various challenges, for example: we offer young players to repeat complex elements. The authors of the videos with the largest number of “likes” receive prizes. Thus, JuniStat is not just a selection tool, but a community where young players can train, express themselves and compete with each other.

How it looks for the coach

In the JuniCoach app the coach sees the dynamics of the player’s development (including within the group). Statistics help the coach evaluate the effectiveness of the training program for each player, if necessary, adjust it and build an individual track of the athlete’s development.

At the same time, the application simplifies the process of evaluating parameters – speed, reaction, acceleration, and so on. Most schools use stopwatches to count time intervals. But in this case, the final result depends on the eye estimation and the coach’s reaction. Large academies use expensive laser sensors, but they also have shortcomings. Lasers are not immune from false positives (the player may accidentally block the ray with his hand during preparation) and introduce an error in the estimate of the starting jerk (if the athlete has slightly moved away from the starting line).

Therefore, the creators of the project decided to use computer vision technologies. They allow trainers to obtain results comparable in accuracy and measure parameters that cannot be calculated using lasers – for example, to build graphs of agility, acceleration and braking.

How does it work

When constructing a rating, about sixty different metrics are taken into account. In this case, the test results are translated into physical and technical training skills scores (from 0 to 100). Therefore, it is easier for players and scouts to perceive the overall picture. It is immediately clear whose skill is higher if the first junior has 80 points for holding the ball, and the second has only 50. At the same time, such a system makes the rating “alive.” New players come and show higher results. Yesterday’s leaders are forced to improve their skills to return to the top.

The application for trainers and scouts, on the contrary, emphasizes the initial data. The coach, working with the team, already knows who leads the ball better or worse. Detailed graphs allow him to understand the reason and what particular element needs to be corrected by the player to improve his skill.

Where we raised money for development

The development of the platform and technology requires quite serious investments, especially a lot of costs go to RnD. The project raised investment several times and received grants. So, in 2020, the project received $250 thousand in investments from business angels. And most recently, in January 2022, the club of venture investors AngelsDeck invested $310 thousand in JuniStat. The syndicator of the transaction was Ilya Partin, investment director of Brayne. Due to the syndication model, we raised not only money, but also expertise from club members, which helps in solving current business problems.

Results and Plans

The JuniStat technology was validated in the Russian Football Union. It is already used by more than 24,000 players and clubs from Russia and the CIS, including Rostov, Yuri Konoplev Football Academy, SSOR Zenit, Rodina, Strogino and others.
A project with the Moscow Football Federation is under pilot launch, within its framework digital profiles of 1000 players will be formed and data collection will begin in the context of many years of training of young football players. The project organizers have already begun to connect football schools and clubs from Europe and the USA to the JuniStat.

In January 2022, the project was selected in the top 10 world accelerators Start-Up Chile – and since March plans to enter the Latin American market. Now JuniStat is undergoing accreditation in the FIFA innovation program.

In the future, there are plans to connect new cloud services – including for the deployment of models and advanced experiment management. They will allow you to more quickly expand the capabilities of applications – to add new tests for tactical thinking and tools for psychological assessment of players.

By Gleb Shaportov, co-founder of JuniStat

Previous ArticleNext Article