Offside is 1 rule in soccer which still stands as a challenge for technology. Wrong offside decisions by referees has changed the outcomes of several important matches in soccer history. With all the advancements in Computer Vision, Deep Learning and Internet of things, there is yet to come a solution which can replace manual referees without impacting the game flow. There have been proposals which involve using sensors on the ball and/or player’s shoes to keep track of their positions. Even if that works, it cannot be widely accepted because it will reduce the comfort level of the player. We propose a solution which integrates Computer Vision and Deep Learning and just needs video as the input, with no additional tweaks on the ball or the player’s shoes. From the live video feed, we aim to track the player positions and give a real-time feedback to the referee, like how the response of goal line technology is conveyed. As a first step, we have worked on static images and dealt with simpler cases. We aim to implement this on videos and consider boundary cases of offside in future. Following image shows a sample output:
