Auto-contextual Generation in Football Video Analytics using Light-weight AI

Football, as one of the most popular sports, creates numerous opportunities for the creative industry.

Sports analytics, a branch of data science, involves analysing videos to provide insights for the media including TV show commentary and enhance the competitive edge of coaching teams and players.

For instance, it offers broadcasters and betting houses real-time game statistics and tactical insights, while aiding clubs in competitor analysis, game preparation, and player monitoring.

A crucial aspect of sports analytics is transforming videos into meaningful contexts, facilitating the creation, analysis, and searchable storage of crucial game events. Despite its importance, analysing football videos is often tedious and time-consuming.

This project aims to address this challenge by developing an AI model to improve the productivity and cost-effectiveness of auto-contextual generation.

We utilise subtitles in videos for data labelling, and our state-of-the-art 3D field reconstruction for relevant feature extraction, thereby enabling us to develop a light-weight AI model that can produce realtime contextual generation.These innovative approaches also enable us to establish a sustainable and cost-effective business model, ensuring affordability and accessibility for customers with budget constraints.