Fútbol - The Beautiful Game

Welcome to KeepUP; an innovative application that leverages AI-powered computer vision and machine learning models to provide advanced analysis of players' performance during guided soccer drills. The platform offers comprehensive metrics and insights, empowering players to enhance their skills and reach their full potential. With AI at its core, KeepUP revolutionizes the way players interact with the beautiful game, offering unparalleled opportunities for growth and connection.

The vision for Keep Up is to create an inexpensive and user-friendly application available to all soccer lovers. This will also allow passionate players to come together and create a community in which users are motivated to improve their abilities through the power of friendly competition.  KeepUp uses machine learning computer vision models to track and analyze players' performance while completing specific drills. The platform calculates 6 athletic metrics and displays them in a trading card format. All users' trading cards are ranked on a global leaderboard. Our application will impact soccer players by motivating them to get outside and train more, work on improving their fundamental skills, and help people find a community of like-minded people.

Technical Breakdown

KeepUp comprises a dynamic client-side interface seamlessly integrated with a robust server-side infrastructure. The client interface facilitates user interaction, guiding players through skill-enhancing drills, while the server-side efficiently processes requested activities and stores results in a centralized database. This enables seamless retrieval of performance metrics by the client interface.
The project encompasses four distinct technical modules:

1. Analysis
2. Processing
3. Cloud Computing
4. Depth Sensing

The analysis module employs AI models such as Yolov8 in Python 3 to analyze provided videos. To improve analysis results, the Depth Sensing module introduces an additional dimension to the videos, enhancing accuracy and precision during assessment.

Facilitating seamless execution of these analysis scripts required the Cloud Computing module which orchestrates the fetching of videos, triggering the appropriate Python scripts for processing.

Finally, the Processing module integrates a simple client interface, enabling players to effortlessly upload their videos and receive real-time feedback on their performance.

Metrics

The player is prompted to take videos of themselves performing 4 different guided football drills.

1. Keep Ups
2. Figure Eight
3. Shooting
4. Vertical Jump

Based on the performance of these 4 drills; performance metrics are outputted by the platform in a training card format. Each metric except for Keep Up count is scaled to a score range of 0-10. This allows for a user-friendly platform. The performance metrics are the following:

• Keep-up count and dominant foot
• Speed
• Control
• Vertical
• Accuracy

Takeaways

KeepUp is an exciting end-to-end platform for users to compete with other community members and develop their football skills. It is a fully-fledged application that allows users to analyze their performance in different drills using our innovative machine learning models.

Our cloud computing technology allows for efficient analysis using the latest machine learning tools with no restriction on the user's device. Our depth processing implementation opens the door to another dimension of analysis that further enhances the user's experience.

All of this comes together for an immersive experience where they can better themselves and connect with their community.

Implementing KeepUp taught us an incredible amount about software development, machine learning, and about designing for and optimizing a client's experience. It was a lot of our first time implementing an entire system from scratch from end-to-end and gave us invaluable experience that will be able to take to any of our future endeavours.

To learn more: https://www.youtube.com/watch?v=6ZMBBnZ_W5s

‍