Soccer On Your Tabletop in action
After 3D television failed to fulfil its early promise as the “next big thing” in sports broadcasting, it has been Virtual Reality (VR)’s turn to carry the burden of the industry on its shoulders.
BT Sport has produced live matches and highlights for several events, including the UEFA Champions League Final, while BBC Sport is making a range of VR content available on a dedicated application for the 2018 World Cup in Russia.
Sports organisations are especially interested in how VR can bring remote fans closer to the action by creating an experience similar to that of being in a stadium.
VR sport hasn’t yet hit the mainstream, but it is gaining in popularity thanks to more powerful smartphones and mobile networks. Indeed, the arrival of 5G networks from 2019 onwards will bring faster speeds, higher capacity and ultra-low latency ideally suited for VR.
But a group of researchers are looking at another technology that could change the way we enjoy sporting events – Augmented Reality (AR). Whereas VR presents a fully immersive image, AR adds digital elements to real life scenes, as best demonstrated by Pokémon Go.
Soccer on Your Tabletop
“Soccer on Your Tabletop” is a prototype AR system that converts any football video on YouTube into a 3D hologram that can be projected onto a flat surface like a table. This can be viewed using a compatible headset.
3D renders are already possible through the use of motion capture and other technologies, but these techniques necessitate the use of numerous cameras dotted around a large area like a stadium. The cost and effort involved limits its widespread use and practicality.
In a paper due to be presented at the Computer Vision and Pattern Recognition conference in Salt Lake City, the University of Washington’s Konstantinos Rematas and his colleagues detail how they wanted to see if it was possible to achieve the same feat simply using a 2D video as an input.
The real innovation of the system is how it is able to locate players on the pitch. The researchers used EA Sports’ FIFA video game to train a machine learning model that can estimate the per pixel depth values of any soccer player from a single frame of footage.
You can see various examples of the system at work in the video below, but there are some limitations that the researchers recognise.
,
Source: Forbes