AUTOMATED FOOTBALL ACTION EVENT DETECTION
Project Overview
Sports is one of the most followed industry of the decade and many new advancements are incorporated to help the audience better interact with the game. In the background, many new concepts are adopted by the sports groups, sports teams and sports institutions to help their players perform well and to come-up with new techniques. Football is one among the most followed game and their are many elite clubs who want their teams to perform well during matches. In the UK alone, there are many clubs engaging players of all ages to train them to be better players of future. These clubs and sports group are constantly looking to adopt new technologies that could help to achieve their goal and one such adoption is the use of Computer Vision and Artificial Intelligence.
Most of the football team (coaches) relay on the metadata generated by the human annotators of a game play to understand the key strengths and weakness of their team during a match. These metadata is generated by group of human annotators by watching the game recorded in a video format. The metadata consist of tracking details of every player, football, different action sequences and all details requested by the sports clubs. This process is time consuming, since human annotator must watch entire game to generate different detailed results. Computer Vision and Artificial Intelligence can be to generated necessary results that indeed reduces the total time required to meaningful results from the games.
Project Description
The overall goal of the project is to detect the players in the video footage and also track every player in a given frame. There are also few additional goals such as background research in the area of computer vision related to sport, develop a new tracking system for player tracking, developing/ customization of the payer detection system and more. The project results will help the Statmetrix company and the sports industry to adopt the new advancements in the field of Computer Vision and AI
Our entire work is divided into 4 sections, player detection, player-audience classification, player tracking and number tracking. Since we are using a custom setup camera system, the object detection models must be optimised to detect person class objects in all the frames. Using static bounding box location and spatial features, the detected person class objects are classified into players and non-players (audience). This helps to reduce the number of objects to be tracked. Using custom tracker, the objects are tracked every frame to produce online tracking results. This result is later used in detecting player t-shirt numbers and to join any gaps in the tracks.
This project is sponsored by Innovate UK, with a partnership between Loughborough University and Statmetrix Company.
